Field Investigations and SWMM Modeling of an Undeveloped Headwaters Catchment Located in the Lower Coastal Plain Region of the Southeast USA by Kyle Patrick Moynihan A thesis submitted to the Graduate Faculty of Auburn University in partial ful llment of the requirements for the Degree of Master of Science Auburn, Alabama December 14, 2013 Keywords: Headwaters, Field Investigations, SWMM Copyright 2013 by Kyle Patrick Moynihan Approved by Jose Goes Vasconcelos Neto, Chair, Professor of Civil Engineering T. Prabhakar Clement, Professor of Civil Engineering Xing Fang, Associate Professor of Civil Engineering Abstract The southeastern region of the United States is host to a diverse variety of geophys- ical regions including the Blue Ridge, Piedmont, Inland Basins and Coastal Plains. Each region shows its own distinctive set of hydrological characteristics and understanding the connections between these processes is key to developing responsible watershed management practices. This thesis presents a study performed in the undeveloped headwaters of an inter- mittent watershed. Containing an area of 2.9 km2 the study site, referred to as WS-AGC, is located in the Coastal Plains region of Alabama. With collaboration between Auburn Uni- versity and the Alabama chapter of the Associated General Contractors of America (AGC), this work intended to perform a water budget study and to assess the feasibility of sus- taining a pond at WS-AGC. To achieve this goal, two separate tasks were performed. The rst was the construction, deployment, monitoring and maintenance of various eld moni- toring facilities and equipment. These included rain gauges, weirs, groundwater observation wells and a portable weather station. The second objective focused on the development and calibration/veri cation of a SWMM model with respect to various hydrological conditions. Field monitoring studied o ered a glimpse on the hydrological processes related to water motion in the watershed. Such monitoring supported the development of hypotheses on the interactions between these processes at WS-AGC. These dynamics processes included; 1) the observed e ects of the forested land cover on the water table level due to evapotranspiration; 2) stream ows that were either connected or not to the groundwater; 3) variations of runo responses over seasonal uctuations. Also, results from SWMM simulations were generally able to represent the dynamic nature of WS-AGC with regards to mean ow and total volume runo characteristics. However, this could only be achieved with the use of groundwater compartment in the model. ii Acknowledgments I would like to start out by thanking my good friend and adviser Professor Jose Neto Vasconcelos for the support and knowledge he has provided me over the past few years. It has truly been an honor to work with such a dedicated and hard working individual. From the short planned meetings, which never ended under an hour or two, to the blood and sweat that was put into eld studies, I will never forget all the great memories we have created during this research. I would also like to thank everyone that assisted me with my eld work. To Paul Simmons, Cameroun Thomas, George Merriam and Will Brown thank you for all your hard work and strength, without you guys I could have never accomplished all that I did these past two years. To Carmen Chosie and Thomas Weems thanks for all good times in class and during the many hours spent down in the dungeon debugging VBA codes. Lastly I want to thank my friends and family for all the support and encouragement these past two years. To Austin Millwood thank you for all the great times we shared during my brief periods of free time. It really helped having such a great friend around during these challenging times. To my parents, Tim and Sue Moynihan, thank you both for being so incredible as you have done nothing but support and encourage me since I can remember. Without all your love and support I would not be who I am today. iii Table of Contents Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ii Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iii List of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vi List of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xi 1 Introduction and Literature Review . . . . . . . . . . . . . . . . . . . . . . . . . 1 1.1 Literature Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 1.1.1 Water Budget Analysis Techniques . . . . . . . . . . . . . . . . . . . 2 1.1.2 Undeveloped Watershed Modeling Using SWMM . . . . . . . . . . . 4 1.2 Knowledge Gaps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 2 Scope and Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 3 Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11 3.1 Site Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11 3.2 Field Investigations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15 3.2.1 Precipitation Collection . . . . . . . . . . . . . . . . . . . . . . . . . 15 3.2.2 Meterological Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18 3.2.3 Evapotranspiration Calculations . . . . . . . . . . . . . . . . . . . . . 18 3.2.4 Runo Monitoring . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21 3.2.5 Weir Discharge Equations . . . . . . . . . . . . . . . . . . . . . . . . 32 3.2.6 Groundwater Observation Wells . . . . . . . . . . . . . . . . . . . . . 37 3.2.7 In ltration Testing . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38 3.2.8 Soil Sampling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40 3.3 Fundamentals of Operating SWMM . . . . . . . . . . . . . . . . . . . . . . . 43 3.3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43 iv 3.3.2 Internal Mechanisms . . . . . . . . . . . . . . . . . . . . . . . . . . . 45 3.3.3 Interface Objects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52 3.4 SWMM Model Development . . . . . . . . . . . . . . . . . . . . . . . . . . . 53 3.5 Model Calibration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58 4 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62 4.1 Collected Field Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62 4.1.1 Precipitation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62 4.1.2 Temperature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68 4.1.3 Potential Evapotranspiration (PET) . . . . . . . . . . . . . . . . . . 70 4.1.4 Runo . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72 4.1.5 Groundwater . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79 4.2 SWMM Simulation Comparisons . . . . . . . . . . . . . . . . . . . . . . . . 85 4.2.1 Hydrographs Comparison . . . . . . . . . . . . . . . . . . . . . . . . 85 4.2.2 Flow Duration Exceedance curves . . . . . . . . . . . . . . . . . . . . 90 4.2.3 Simulation Error Analysis . . . . . . . . . . . . . . . . . . . . . . . . 92 4.2.4 Statistical Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . 99 5 Summary of Findings and Suggestions for Future Studies . . . . . . . . . . . . . 103 5.1 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103 5.2 Suggested Future Studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106 Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107 Appendices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111 A Soil Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112 v List of Figures 3.1 Left: Areal view of the WS-AGC with the network of intermittent streams. Right: Respective contours lines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12 3.2 Watershed AGC soil composition, provided from (NRCS/USDA web soil survey) 14 3.3 Rain gauge locations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16 3.4 Rain gauge post and meteorological station . . . . . . . . . . . . . . . . . . . . 17 3.5 Construction of Cipolletti weir . . . . . . . . . . . . . . . . . . . . . . . . . . . 22 3.6 Roots causing issues on the lateral walls of the weir . . . . . . . . . . . . . . . . 23 3.7 Support post for the Cipolletti weir . . . . . . . . . . . . . . . . . . . . . . . . . 23 3.8 Upstream view of the Cipolletti weir . . . . . . . . . . . . . . . . . . . . . . . . 24 3.9 Downstream view of the Cipolletti weir . . . . . . . . . . . . . . . . . . . . . . . 24 3.10 Riprap placed at the downstream side of the Cipolletti weir . . . . . . . . . . . 25 3.11 Finished Cipolletti weir looking upstream: completed (June 1st 2012) . . . . . . 26 3.12 Cipolletti weir dimensions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26 3.13 Support post for the rectangular weir . . . . . . . . . . . . . . . . . . . . . . . . 27 3.14 Upstream view of rectangular weir . . . . . . . . . . . . . . . . . . . . . . . . . 28 vi 3.15 Rectangular weir dimensions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28 3.16 HOBO pressure sensor installed upstream at each weir . . . . . . . . . . . . . . 29 3.17 Erosion damage at the Cipolletti weir to the right . . . . . . . . . . . . . . . . . 30 3.18 Upstream view of broad crested weir . . . . . . . . . . . . . . . . . . . . . . . . 31 3.19 Broad crested weir dimensions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31 3.20 Modi ed rectangular weir dimensions . . . . . . . . . . . . . . . . . . . . . . . . 32 3.21 Values of width-adjustment factor, taken from (Dodge, 2001) . . . . . . . . . . 34 3.22 E ective coe cient of discharge, Ce, as a function of L=B and H=P, taken from (Dodge, 2001) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34 3.23 Shallow groundwater wells . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37 3.24 Locations of In ltration Testing . . . . . . . . . . . . . . . . . . . . . . . . . . . 39 3.25 Bore sites at WS-AGC indicated by black markers . . . . . . . . . . . . . . . . 40 3.26 Drill rig entering WS-AGC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41 3.27 Setting up to drill boring . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41 3.28 Soil exiting the bore hole from approximately 4.6 m below the surface . . . . . . 42 3.29 Samples taken from the SPT . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42 3.30 Shelby tubes being prepared to take out of the site . . . . . . . . . . . . . . . . 43 3.31 Conceptual view of SWMM?s runo mechanism, (Rossman and Supply, 2005) . 45 vii 3.32 Conceptual view of Horton?s in ltration capacity recovery mechanism used in SWMM?s computational code, (Viessman et al., 2003) . . . . . . . . . . . . . . 48 3.33 Conceptual view of SWMM?s groundwater mechanism, (Rossman and Supply, 2005) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49 3.34 Discretized Sub-Catchments based on topography and soil types for WS-AGC . 53 3.35 Surveying a downstream cross section . . . . . . . . . . . . . . . . . . . . . . . 54 3.36 Shallow ground water level relative to the surface for the out of stream observation well . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57 4.1 Rainfall recorded at WS-AGC (February 4th - August 14th, 2012-2013) . . . . . 63 4.2 O -site rain gauges used as quality control) . . . . . . . . . . . . . . . . . . . . 64 4.3 Onsite vs o -site monthly rainfall totals (February 4th - August 14th, 2012-2013) 65 4.4 Temperature recorded at Columbus, GA Airport (January 1st - August 14th, 2012-2013) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69 4.5 PET calculations using Hamon?s and Hargreaves Methods . . . . . . . . . . . . 71 4.6 Rainfall runo events at Cipolletti weir used during calibration e orts . . . . . . 72 4.7 Rainfall runo events at broad-crested weir used during veri cation e orts . . . 73 4.8 Rainfall runo events at rectangular weir . . . . . . . . . . . . . . . . . . . . . . 74 4.9 Comparision between runo at Cipolletti weir and rectangular weir (Pre-Construction February 8th-14th, 2013) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75 4.10 Comparision between runo at broad-crested weir and rectangular weir (Post- Construction February 23rd-26th, 2013) . . . . . . . . . . . . . . . . . . . . . . 75 viii 4.11 Representation of the dynamic groundwater processes of a gaining/losing stream such as the one at WS-AGC, (Winter, 2007) . . . . . . . . . . . . . . . . . . . . 80 4.12 Instream ground water level vs rainfall intensity . . . . . . . . . . . . . . . . . . 81 4.13 Instream ground water level vs cumulative rainfall . . . . . . . . . . . . . . . . . 81 4.14 Out of stream ground water level vs rainfall intensity . . . . . . . . . . . . . . . 82 4.15 Out of stream ground water level vs cumulative rainfall . . . . . . . . . . . . . . 82 4.16 Rainfall event without runo event during August 6th, 2012 at Cipolletti/Broad- Crested weir . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83 4.17 Smaller rainfall event yielding runo during July 20th, 2013 at Cipolletti/Broad- Crested weir . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84 4.18 Signs of ET from the out of stream ground water well . . . . . . . . . . . . . . . 85 4.19 Output hydrographs produced with multiple aquifer con guration . . . . . . . . 87 4.20 Output hydrographs produced with single aquifer con guration . . . . . . . . . 88 4.21 Output hydrographs produced with no aquifer con guration . . . . . . . . . . . 89 4.22 Veri cation Flow duration exceedance curves produced with multiple aquifer con- guration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91 4.23 Veri cation ow duration exceedance curves produced with single aquifer con g- uration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91 4.24 Veri cation ow duration exceedance curves produced with no aquifer con guration 92 4.25 Veri cation error analysis for max ow during multiple aquifer simulation . . . 93 ix 4.26 Veri cation error analysis for max ow during single aquifer simulation . . . . . 94 4.27 Veri cation error analysis for max ow during no aquifer simulation . . . . . . . 94 4.28 Veri cation error analysis for mean ow during multiple aquifer simulation . . . 95 4.29 Veri cation error analysis for mean ow during single aquifer simulation . . . . 96 4.30 Veri cation error analysis for mean ow during no aquifer simulation . . . . . . 96 4.31 Veri cation error analysis for total ow during multiple aquifer simulation . . . 97 4.32 Veri cation error analysis for total ow during single aquifer simulation . . . . . 98 4.33 Veri cation error analysis for total ow during no aquifer simulation . . . . . . 98 x List of Tables 1.1 Comparison of annually measured precipitation and evapotranspiration for forested ecosystems in the Southeastern USA, (Sun et al., 2010) . . . . . . . . . . . . . . 4 3.1 USGS web soil survey results for WS-AGC study site . . . . . . . . . . . . . . . 13 3.2 Kestrel 4500 meteorological station measurements and accuracy . . . . . . . . . 18 3.3 Coe cient and constants used when determining the e ective coe cient of dis- charge, taken from (Dodge, 2001) . . . . . . . . . . . . . . . . . . . . . . . . . . 35 3.4 Coe cient, C, values used for broad crested weirs, taken from (Brater et al., 1996) 36 3.5 Literature values of maximum and minimum in ltration rates for Horton Equa- tion, (Akan, 1993) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55 3.6 Minimum and maximum ranges used for Horton?s in ltration input parame- ter,(Akan, 1993) and (Rossman and Supply, 2005) . . . . . . . . . . . . . . . . . 59 3.7 Aquifer properties for various soil types, taken from (Rossman and Supply, 2005) 60 3.8 Minimum and maximum ranges used for aquifer component input parameter, (Rossman and Supply, 2005) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61 4.1 Rainfall statistics for (February 4th - December 31st, 2012) . . . . . . . . . . . 66 4.2 Rainfall statistics for (January 1st - August 14th, 2013) . . . . . . . . . . . . . 66 xi 4.3 Normal Monthly Rainfall Totals (Period of Record 1981-2010)-Location: Colum- bus, GA (Airport) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67 4.4 Monthly rainfall average and yearly totals for (February 4th - July 31st, 2012-2013) 67 4.5 Percent di erences between monthly rainfall average (February 4th - July 31st, 2012-2013) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68 4.6 Monthly Recorded Temperatures vs Normal Monthly Temperatures (Period of Record 1981-2010)-Location: Columbus, GA Airport . . . . . . . . . . . . . . . 69 4.7 Yearly PET totals calculated from Hamon and Hargreaves Methods . . . . . . . 71 4.8 Descriptive rainfall runo statistics . . . . . . . . . . . . . . . . . . . . . . . . . 76 4.9 Descriptive rainfall runo statistics and R/P ratios for wet season events 2012-2013 77 4.10 Rainfall runo statistics and R/P ratios for dry season events 2012 . . . . . . . 78 4.11 Rainfall runo statistics and R/P ratios for dry season events 2013 . . . . . . . 78 4.12 Performance ratings for NSE, (Moriasi et al., 2007) . . . . . . . . . . . . . . . . 100 4.13 Calibration error analysis for all simulations performed . . . . . . . . . . . . . . 101 4.14 Veri cation error analysis for all simulations performed . . . . . . . . . . . . . . 102 xii Chapter 1 Introduction and Literature Review Developing regions of the United States face a tight balancing act between sustaining city growth and protecting the quality of surrounding natural resources. Growing popu- lations and cities create large strains on local resources and increase potential pollution hazards throughout the watershed. The southeastern region in particular has become a ma- jor concern with an increase in timber production, poor timber management practices and urban development(Harder et al., 2007).Forests play a great role in regulating the regional hydrologic patterns of the southern United States where 55% of the region is covered by forest (Sun et al., 2002). According with Wear and Greis (2002) the timber production has more than doubled from 1953 to 1997 in the southeastern United States. However, timber production is not the only factor a ecting this region. The southeastern U.S. is expected to lose about 4.9 million forest hectares (ha) to urbanization between 1992 and 2020, with a substantial part of the loss concentrated in the Atlantic Coastal Plains (Harder et al., 2007). Despite such growth tendencies, maintaining the natural ecology and resources is critical for creating a healthy sustainable environment. In order to promote better industry practices, hydrological processes across the diverse geological regions of the Blue Ridge, Piedmont, Inland Basins and Coastal Plains must be better understood. Field investigations have been conducted for several decades providing sites across these regions with long term hydrological data. These studies have helped describe the hydrology of dynamic forested watershed while providing evidence of the diverse geophysical regions in the southeastern USA (Sun et al. (2002); Amatya et al. (2007); Davis et al. (2007); Harder et al. (2007); Sun et al. (2010); La Torre Torres et al. (2011)). Inter- site eco-hydrological comparison studies have the potential to predict more accurately the 1 hydrologic e ects of headwater forest management under di erent environments (Sun et al., 2002). However, further research is needed to understand the natural dynamics of water balance components in such forested systems along the Coastal Plains to accurately assess impacts of anthropogenic disturbances and for improving forest management strategies related to water quality (Harder et al., 2007). Long-term hydrologic data is essential for understanding the hydrologic processes as base line data for assessment of impacts and conservation of regional ecosystems as well as for developing and testing eco-hydrological models (Amatya et al., 2007). The need for continued research e orts focusing on forested watersheds of the southeastern USA has led to investigations described in this thesis. 1.1 Literature Review This section provides discussions and a summary of current watershed analysis tech- niques found in literature. Initial discussions focus on the water budget approach that incorporates a simple mass balance technique to identify key hydrological characteristics. Then a summary of case studies utilizing and testing the Storm Water Management Model SWMM ability to simulate natural undeveloped watershed dynamics are presented. 1.1.1 Water Budget Analysis Techniques The water budget method was a major breakthrough in qualitatively depicting natural undeveloped watersheds. One of the pioneers into this development was Thornthwaite (1948), who provided detailed descriptions of the driving forces behind watershed dynamics. Seven years later, in 1955, Thornthwaite and Mather (1955, 1957) presented two contributions that laid the foundation for the standards in water budget studies. With its origins predating the computer age, this technique allows for the researcher to simplify and visually track the propagation of water through separated mechanisms. 2 Some components within the watershed dynamics pose a great challenge for researcher to quantify. These challenges mainly lie within the groundwater component of a water budget analysis. Generally, in long term site studies, most researchers ignore in uences from the groundwater in ow and out ow. This is assumed to have a very small net uctuation throughout the year (Harder et al., 2007). After all simpli cations are made, main driving forces are typically compartmentalized into the following mass balance formula Equation (1.1) (Harder et al., 2007). S = P ET Q (1.1) Where S is the change in the water storage within the soil column, P is the amount of rainfall, ET is the actual evapotranspiration (AET) and nally Q is the amount of runo . All of the prior are normalized with the study regions area and measured in terms of millimeters or inches. In order to conduct an e ective investigation using this technique, study areas must have long term local data. Having local data reduces any bias e ects from drastic weather changes over temporal and spacial variations. Typically a water budget analysis is performed with hydrological data ranging between (2 to 30+) years (Sun et al., 2002; Amatya et al., 2007; Sun et al., 2010; La Torre Torres et al., 2011). Studies including Sun et al. (2002), Harder et al. (2007) and Sun et al. (2010) have used this simple approach to investigate the connections between rainfall runo , AET and potential evapotranspiration (PET) as well as S through headwaters undeveloped watersheds. One of the most di cult variables to determine from Equation 1.1 is AET. Several methods have been developed to calculate PET. PET calculations can be performed using various input parameters and they are one valid way to estimate AET. Some of the most common methods rely on temperature based inputs. The interested reader can nd in-depth descriptions from Federer and Lash (1978) ,V or osmarty et al. (1998) and Hargreaves and Allen (2003). Sun et al. (2010) summarized the average precipitation (P), ET and ET/P values for 18 separate watershed studies around the southeastern region of the USA. This summary is presented in Table 1.1 where the importance of ET can be seen. The range of 3 ET/P ratios vary from 0.41 to 0.93 providing evidence that ET is one of the most in uential loss components in the watershed system. Table 1.1: Comparison of annually measured precipitation and evapotranspiration for forested ecosystems in the Southeastern USA, (Sun et al., 2010) Other methods are based on more sophisticated techniques that make use of local solar radiation inputs. These methods are di cult to employ in most areas since the availability of long term localized solar radiation data is very limited. The sensitivity analysis performed by Bormann (2011) provides a detailed in-depth description into the uses of 18 di erent PET model strategies. Bormann (2011) concluded that selecting a PET model depends solely on the type of data available and the climate on the region of interest. 1.1.2 Undeveloped Watershed Modeling Using SWMM Many numerical models exist for simulating hydrological processes of a watershed. Most of which incorporate mechanisms that simplify observed processes such as runo , in ltration, 4 inter ow, and groundwater. The simpli ed structure is typically a collection of subsystems each representing individual hydrologic processes (Axworthy and Karney, 1999). Numerical models range from custom-built algorithms to government supported user interface software packages. Distinguishing which model is the best t for a speci c project can be a challenging task. Work performed by Borah et al. (2009) provides detail descriptions into capabilities and limitations of 14 currently o ered watershed modeling packages. However, the focus in this section is a review of studies which have analyzed the ability of the Storm Water Management Model (SWMM) to model undeveloped watersheds. Also included is a review of algorithms incorporated into SWMM to make the calibration processes more e cient. Davis et al. (2007) analyzed SWMM?s capability to model four rural watersheds with areas ranging from 3.22 to 8.94 mi2 located in the Piedmont ecoregion of North Carolina. In order to develop physical properties of the catchments, many sophisticated techniques were utilized. Using Geographical Information Systems (GIS) software, a 10 m Digital Elevation Map (DEM) and aerial photos, the local catchment characteristics such as area, slope, runo length and imperiousness were determined. Channel dimensions were found from eld surveys and soil hydrologic groups provided estimations of in ltration rates. Finally, Horton?s in ltration equations were utilized within all catchments. Once each model was constructed, calibration e orts were initiated. Results indicated that calibration by ow duration curves cannot be achieved for all events of record through the adjustment of watershed parameters like percent imperiousness, in ltration, overland roughness, and conduit roughness alone (Davis et al., 2007). This led to development of a single aquifer component with one receiving node. During this e ort default aquifer param- eters were initially used. Then soil texture classes properties inputs were altered to create the best calibration results. The best t calibration results were seen when incorporating the sandy clay aquifer properties. Flow duration exceedance curves were unable to match peak ow rates but did provide good results in the mid to low range ow rates. However, during the study no 5 comparisons between local soil data were used to con rm the sandy clay conditions. The study also relied only on ow duration exceedance curves and single event hydrographs to determine the accuracy of simulations. Statistical analysis between calibration and observed data was not performed during the calibration e orts. In order to perform a comprehensive analysis Legates and McCabe (1999), Krause et al. (2005) and Moriasi et al. (2007) suggest that not only should graphical methods be used to evaluate a models performance but a compilation of statistics as well. This suggest that results from (Davis et al., 2007) were not analyzed su ciently. To truly suggest that SWMM can be used for undeveloped watersheds, a more in-depth statistical study must be performed. Work performed by Jang et al. (2007) focused on developing SWMM simulations for pre- and post-development conditions. Models were created for three natural watersheds and four disaster stricken areas in Korea, where impact assessments have already been conducted. Studies were split into two phases of SWMM modeling, the rst phase of testing focused on the three undeveloped headwaters watersheds and the second phase examined the four separate disaster stricken watersheds. Since the proposed approach was to use SWMM both for pre- and post- development condition, it is necessary to verify the applicability of SWMM to natural watershed condition (Jang et al., 2007) During the rst stage, research sites at the Seolmacheon, Weecheon and Pyungchang River watersheds provided observed rainfall runo data to compare SWMM simulation results against. The catchment areas ranged from 8.51 to 55.93 km2 with slopes between 5.45 to 36.96 %. When constructing the model a single sub-catchment compartment was selected as this produced the closest results to observed data. Along with SWMM modeling of the researched watersheds other methods of runo estimation were applied to the natural catchments. The Soil Conservation Service (SCS) or Clark method is the standard method for runo estimations in Korea (Jang et al., 2007). Comparisons were then made between the Clark method and SWMM model outputs. 6 Each SWMM model created for the three study catchments were only tested with three rainfall runo events. Parameters were estimated from either physical information available or the suggested values from the tables in the SWMM manual or existing literature (Jang et al., 2007). The models were left uncalibrated in order to simulate the typical process that would be encountered with an ungauged pre-development site. Maximum ow rate and time to peak results from the SWMM simulations during the three separate rainfall runo events showed close results to observed data. The Clark method also produced good agreement but underestimated peak runo for most rainfall runo events analyzed. Evidence presented from this study shows that SWMM was able to represent the behavior of the undeveloped watershed. However, to provide a stronger case for the ability of SWMM to handle these conditions more events must be analyzed. Also Jang et al. (2007) did not incorporate any sophisticated error analysis techniques. The determination of the models performance was based purely on hydrographs, peak ow rates and time to peak ow rates. Although these results have shown good agreement more analysis is needed to truly deem modeling results as acceptable. Davis et al. (2007) used manual techniques to calibrate four watersheds and Jang et al. (2007) did not incorporate calibration in their modeling e orts. Manual or no calibration e orts severely limit the ability of a user to achieve an optimal set of defensible parameters. A poorly calibrated model might lead to poor designs resulting in four serious impacts: ooding, stream erosion, water quality violations and habitat destruction (James et al., 2002). In order to improve future studies involving SWMM, James et al. (2002) has developed a computing evolution-strategy called a genetic algorithm (GA). Generally, the evolution-strategy is an optimization method based on strategies encountered in biological evolution (James et al., 2002). The strategy also incorporates a limit or range of uncertainty for the calibration parameters of interest. This allows for parameters to remain within meaningful ranges. 7 Using a related theory, PCSWMM software developed a variation of SWMM which incorporates an automated calibration tool called the Sensitivity-based Radio Tuning Cali- bration (SRTC) function. Simulations are run from the upper to lower limits and users are able to tune each parameter. With the assistance of a hydrograph interface, e ects from cal- ibration tuning can be seen in real time. This advancement brings a signi cant contribution for current and future modeling e orts. If watershed models are better accurately depicted then results from simulations can be more representative of actual observed conditions. 1.2 Knowledge Gaps As presented in the literature review current knowledge is limited in regards toSWMM?s ability to simulate the undeveloped watershed. Studies including Davis et al. (2007) and Jang et al. (2007) have developed SWMM models to simulate undeveloped watersheds but were unable to comprehensively examine simulation results. Presented in this section are the current knowledge gaps that exist among the ability of SWMM to e ectively simulate the undeveloped watershed. The identi ed knowledge gaps may be summarized as following: 1. How do processes such as rainfall, in ltration, evapotranspiration a ect the behavior of surface water and groundwater in the headwaters of an intermittent watershed? 2. What is the ability of SWMM to simulate the hydrology of an undeveloped, intermittent watershed, over a hydrological year? 3. Which runo characteristic is SWMM able to simulate best (e.g. maximum ow, mean ow or total volumes)? 8 Chapter 2 Scope and Objectives Beginning in early 2012, the Alabama chapter of the Associated General Contractors of America (AGC) partnered with Auburn University. AGC?s goal was to engage students at Auburn University and provide an unique opportunity of working on a real-life land develop- ment project. This involved students developing new ideas to improve the value and quality of land owned by AGC . Several disciplines were involved in this study including: building science; landscape architecture; biosystems engineering and civil engineering. Each disci- pline was tasked with di erent objectives but all worked simultaneously to achieve AGC?s goals. The civil engineering team was given the task of developing a water budget study and providing a feasibility assessment of a pond to be sustained on site. This task o ered an opportunity to provide the needs of AGC while also allowing to study and model local hydrological processes. To study watershed behavior and provide a feasibility assessment several di erent hy- drological monitoring devices were installed. During this initial phase of the eld monitoring program two in-stream weirs, two shallow groundwater wells, three automated rain gauges and a Kestrel 4500 micro meteorological station were installed. Data collected from these devices was used to analyze and develop relationships between various hydrologic compo- nents. After 15 months of data collection a SWMM model was developed and calibrated to replicate the watersheds behavior. Geographic Information Systems (GIS) remote sensing data and local survey information was obtained and used to provide input parameters for the SWMM model. Locally recorded rainfall, temperature, atmospheric pressure and runo 9 were used as input data sets. Then during calibration and veri cation e orts, runo hydro- graphs from installed weirs were used to verify simulation output. Finally, several graphical and statistical methods were utilized to determine the models capabilities to replicate wa- tersheds behavior. The speci c objectives of this research can be outlined as follows: 1. Begin initial phase of the eld monitoring program 2. Analyze data and develop hypotheses in regards to local hydrological relationships 3. Develop a SWMM model for WS-AGC with the use of detailed local survey and remote sensing GIS data 4. Calibrate and validate the SWMM model with respect to observed hydrograph data sets over various hydrological conditions 5. Assess the capabilities of SWMM to model a undeveloped, intermittent watershed using graphical and statistical techniques 10 Chapter 3 Methodology 3.1 Site Description The study watershed referred to as (WS-AGC), is located in Pittsview, AL (32o 9? 29.30"N, and 85o 10? 8.09"W). Covering an area of 2.90 km2 in Russell County, WS-AGC drains a rst order intermittent stream positioned at the headwaters of a complex series of tributaries that connect and discharge into Hatchechubbee Creek. The Hatchechubbee Creek then continues south and eventually discharges in the Chattahoochee River just north of Eufaula, AL. Figure 3.1 provides an aerial imagine of the WS-AGC, its stream development predicted from GIS and topographic lines. This area has been used as a hunting preserve over the past decades and remains relatively untouched except for a few trails and green elds. The e ects of these areas were assumed to have a negligible e ect on the hydrology of WS-AGC. 11 Figure 3.1: Left: Areal view of the WS-AGC with the network of intermittent streams. Right: Respective contours lines Since WS-AGC?s location is at the transition zone between the Piedmont and Lower Coastal Plains (LCP) regions, the study site possesses characteristics that are particular to both regions. With respect to the Piedmont region, the watershed is in an area of many low rolling foothills and contains various patches of clay-like soils. However the area also re ects the LCP region at its low laying central areas. Composition in the LCP environments plays a major role in runo responses due to the presence of very well to poorly drained soils in low- topographic relief areas (La Torre Torres et al., 2011). Similarly to other LCP watersheds, groundwater eld measurements have indicated that WS-AGC hydrological responses are in uenced by evapotranspiration. Observations of this phenomenon show distinct di erences between the dry (May-November) and wet (December-April) seasons. These seasonal separations were selected based on the study conducted by La Torre Torres et al. (2011) on a LCP watershed in the southeastern USA. 12 Finally, the site also displays some hydrological characteristics of uplands regions. Hy- drologic processes in upland areas are mainly in uenced by steep gradient pro les and hill- slope processes (i.e. inter ow, sheet ow and overland ow) and less in uenced by soil com- position (La Torre Torres et al., 2011). These characteristics provide an opportunity to study a distinct watershed type that to our knowledge has not been examined in literature (Sun et al., 2002; Czikowsky and Fitzjarrald, 2004; Harder et al., 2007; Sun et al., 2010; La Torre Torres et al., 2011; Davis et al., 2007). WS-AGC has an average slope of approximately 12:4% and is comprised of several soil types. After applying the NRCS/USDA web soil survey tool (http://websoilsurvey. nrcs.usda.gov), the site was determined to consist of mostly Hannon Clay and Trout- Springhill-Luverne (33 and 36% of the watershed respectively) as shown in Table 3.1. The drainage classes are classi ed as moderately well drained (Hannon clay) and well to exces- sively drained (Trout-Springhill-Luverne) soils. Toward the lower portion of the catchment the soils transition to Kinston Mantachie and Luka (KMA), see Figure 3.2. This soil type only makes up around 4.6% of the area of interest (AOI). Table 3.1: USGS web soil survey results for WS-AGC study site 13 Figure 3.2: Watershed AGC soil composition, provided from (NRCS/USDA web soil survey) Near the southern end, which constitute in the upstream portion of the watershed, a swamp-like collection area exists. Water during the rainy season collects here and discharges into the stream through a series of natural weir-like outlets. As the stream progresses northward from this point. Downstream the channel cross section begins to dramatically increase in size. Originating with channel dimensions of 0.5 m deep and 1 m wide, erosive processes have caused the farther downstream channel segment to reach dimensions of 3.5 m deep and 3 m wide at the base, each with side slopes of approximately 4:1 (V:H). The highly erodible sandy stream bed soil has also formed many abstraction diversions throughout the channels length. For instance, at one location the stream channel completely disappears into a large mound of sandy soil and then reappears about 20 m downstream. Abstractions 14 have also been created throughout the channel. Large amounts of tree branches and debris that have been conveyed downstream from large runo events have caused natural dam like structures to form. Beavers also have created dams within the watershed. Vegetative cover in WS-AGC consists of a mixture between Slash (Pinus elliotii) and Loblolly Pine (Pinus taeda) throughout, similar to the study site of Harder et al. (2007). The density of trees begins to increase closer to the stream bed and into the riparian-zone. The riparian-zone also contains a bamboo, and vines which are mixed with the pines and becomes very dense in many areas. This area is where evidence of a shallow water table has been found which is supported throughout most of the year depending on the rainfall totals experienced. 3.2 Field Investigations The investigation began in early February 2012 with the installation of three rain gauges and a portable meteorological station. As the year progressed the eld monitoring program grew to incorporate an entire system of monitoring equipment that collects data on local runo , groundwater levels, rainfall, temperature and pressure. Data from this equipment is provided at a high resolution, ranging from 15-30 minute sampling intervals. This sec- tion incorporates details into the construction processes and placement of the hydrological monitoring equipment. 3.2.1 Precipitation Collection Precipitation at WS-AGC was collected using three automated recording Onset RG3-M rain gauges. Before deployment, calibration of each gauge was performed. This calibration was completed in the Harbert Engineering Center Hydraulics Laboratory following guidelines provided from the User?s Manual, (Onset, 2011). The calibration test used a known volume of water, 473 ml, that was dripped upon the top of each gauge. Water funneled down into a two bucket tipping mechanism, where each bucket represented 0.2 mm of rainfall. The number 15 of tips were recorded with a goal of reaching 100 2 tips. Screws located at the bottom of the gauge were then adjusted either clockwise or counter clockwise to increase/decrease the number of tips recorded. Once the test provided consistent results for each gauge it was considered fully calibrated. After the calibration period the gauges were installed on February 2012 in the open grass elds located throughout the property, see Figure 3.3. Data at each location was recorded at regular 30 minute intervals with a capability of a ner resolution as an event is occurring. These gauges provide data that is accurate to 1.0% of the readings and can record maximum rainfall rates of 12.7 cmhr . In order to capture rainfall for the entire watershed, gauges were strategically placed in such a way to divide the watershed into three nearly equal areas. The Thiessen polygon method was then used to average each rainfall collection area into a single precipitation time series le, later to be used in numerical modeling investigations. Figure 3.3: Rain gauge locations 16 Rain gauges were installed using (10cmx10cmx5m) pressure-treated wooden posts an- chored with concrete. The bases of the posts were placed approximately 1.5 meters below the surface and supported with vertical stabilizing arms during the drying process. The posts were extended 3.4 meters o the ground to limit interference from the surrounding tree cover as well as human and animal interaction. Figure 3.4 shows the two posts installed at the west gauge location. On the left is a rain gauge and on the right is the Kestrel 4500 pocket weather tracker. To promote Quality Assurance/Quality Control (QA/QC), collected data was compared with two o -site gauges managed by (Community Collaborative Rain, Hail Snow Network- CoCoRaHS) that are located within a 10 mile radius of the site. CoCoRaHS involves many trained volunteers across the country and is supported by the National Oceanic and Atmo- spheric Association (NOAA). Recorded rainfall at these locations were compared with eld data for each month during research e orts. Figure 3.4: Rain gauge post and meteorological station 17 3.2.2 Meterological Data Located along side of the west rain gauge is a Kestrel 4500 portable weather station, see Figure 3.4. Recording continuously at 30 minute intervals, measurements of various climate information including temperature, barometric pressure, wind speed and humidity are captured. In Table 3.2 the accuracy of each measurement recorded is shown. One of the most in uential measurements taken with this device was the barometric pressure. It provided critical local barometric pressure that was used in calibrating the other pressure transducers placed throughout the site. Temperature data was also a critical measurement. It provided the input data for calculating potential evapotranspiration rates through di erent methods which are discussed later in this chapter. Table 3.2: Kestrel 4500 meteorological station measurements and accuracy 3.2.3 Evapotranspiration Calculations Evapotranspiration (ET) is one of the major components of the hydrologic cycle (Tra- jkovic, 2005). ET accounts for the catchments water losses through plant transpiration, ponded water and upper soil zone evaporation. This study incorporated estimates of ET with assistance from two temperature based methods for predicting potential evapotranspi- ration (PET). Temperature based PET methods may over or under estimate ET but when 18 local pan evaporation or solar radiation data is unavailable these methods provide the closest estimate possible. The rst temperature based PET method used in this study was the Hamon method, (Hamon, 1963). This empirical method incorporates the following equations to calculate the daily PET (mm). PET = 0:1651 Ld RHOSAT KPEC (3.1) RHOSAT = 216:7 ESAT=(T + 273:3) (3.2) ESAT = 6:108 e17:26939 T=(T+237:3)) (3.3) Where: PET is the daily PET (mm) Ld is the daytime length from sunrise to sunset in multiples of 12 RHOSAT is the saturated vapor density ( gm3 ) T is the daily mean air temperature ESAT is the saturated vapor pressure (mb) KPEC is the calibration Coe cient, set to 1 in this study Values of Ld in Equation 3.1 were found at (http://www.orchidculture.com/COD/ daylength.html). The site is versatile and can be applied to any where in the world by selecting the closest line of latitude nearest the study region. Temperature data for Equations 3.2 and 3.3 was provided by the local Kestrel 4500 meteorological station. The second temperature based PET method applied was the Hargreaves-Samani ap- proach. This empirical method involves the use of a more complex series of equations. These include equations to solve for parameters including temperature reduction coe cient, rela- tive distance between the earth and sun, solar declination, sunset hour angle, extraterrestrial solar radiation and nally PET. 19 PET = 0:0075 Ra Ct 12 Tavg (3.4) In order to calculate daily PET from Equation 3.4 the series of equations below must be evaluated. dr = 1 + 0:033 cos 2 J 365 (3.5) Equation 3.5 is the rst step in this process. It uses the Julian date (J) to solve for the relative distance between the earth and the sun, dr. = 0:4093 sin (dr 1:405) (3.6) Then the solar declination is calculated with Equation 3.6. This again uses J to solve for (radians). ws = arccos( tan( ) tan( )) (3.7) Next Equation 3.7 is used to determine the sunset hour angle ws (radians). This involves inputting the previously solved variable and the latitude of the study site . Ra = 15:392 dr (ws sin( ) sin( ) + cos( ) cos( ) sin(ws)) (3.8) Now all prior solutions of Equations 3.5-3.7 can be incorporated into Equation 3.8. This value of extraterrestrial solar radiation Ra (MJ=m2=day) can then be used in Equation 3.4 to solve for daily PET. Ct = 0:035 (100 wa)13 (wa 54%) (3.9) Ct = 0:125 (wa< 54%) (3.10) 20 Lastly Equations 3.9 and 3.10 must be solved. These two equations are dependent upon the value of relative humidity wa. If values of wa are greater or equal to 54% then Equation 3.9 is used and if wa is less then 54% then Equation 3.10 must be used. 3.2.4 Runo Monitoring Several options exist when attempting to monitor a streams ow rate. The United States Department of Agriculture (USDA) Water Measurement Manual (Dodge, 2001) provides numerous options. Such as available include weirs, umes, ori ces and venturi meters. The choice of ow monitoring structures is site dependent and not every type may be right for the situation. Runo monitoring in this study was completed with the use of weirs due to the simplicity of the construction and their conformity to the cross sections faced at the site. The interested reader should examine Dodge (2001) for more detailed descriptions into the other options listed above. The rst weir installed was a Cipolletti type that was constructed instream at the lower downstream portion of WS-AGC. It was selected because it conformed the best with natural vertical slopes of the channel. The weir also provided data that has been proven to be within 5%, Dodge (2001). Construction did not begin until late Spring 2012 since the streams were still owing. Wet season conditions had supported high water table levels and it was not until several consecutive weeks of dry weather that construction e orts could begin. Figure 3.5 shows the conditions faced during the construction process. Even with a few weeks of dry conditions the high water table levels caused the base of the weirs foundation to become ooded as the trench was dug. This slowed the construction down considerably. 21 Figure 3.5: Construction of Cipolletti weir The dense and thriving forest root systems surrounding the construction area became a major issue to excavate through by with shovels. Figure 3.6 shows the dense root systems in the vicinity. This issue is revisited later in this section as it also became a major issue with erosion control around the laterals of the weir. In order to support the weir during the potential ve-to-six feet backwater water ele- vations, ve (10cmx10cmx5m) post were placed at a depth of 1.5 meters below the streams bed elevation, Figure 3.7. The voids around the post were then back lled with concrete and allowed to set. The walls were prefabricated in a workshop with guidance from Dodge (2001) for Cipolletti dimension requirements. They were brought out to the eld and installed across the stream bed. 22 Figure 3.6: Roots causing issues on the lateral walls of the weir Figure 3.7: Support post for the Cipolletti weir Shown in Figures 3.8 and 3.9 are the walls installed against the post. The trench was rst lined with a thin layer of limestone gravel to create a solid base. Then it was over topped with twenty centimeters of concrete all the way across. The rest of the trench was back lled with soil from the construction process. 23 Figure 3.8: Upstream view of the Cipolletti weir Figure 3.9: Downstream view of the Cipolletti weir 24 Figure 3.10: Riprap placed at the downstream side of the Cipolletti weir Continuing the nishing process, plastic lining was placed along the upstream portion to help prevent the seepage beneath and along the sides of the weir. Lastly, large rip-rap was placed to provide energy dissipation on the downstream channel side, see Figure 3.10. Figure 3.11 shows the nished Cipolletti weir at the downstream location. 25 Figure 3.11: Finished Cipolletti weir looking upstream: completed (June 1st 2012) The stream bed at this location is approximately 4.6 m deep and 3 m wide with irregular side slopes of roughly 4:1 (V:H). Stretching across the channel at 3.0 m, the Cipolletti weir has a crest length of 1.1 m and two 1.8 m vertical sides, cut at the 4:1 slopes, conforming to the natural slopes. Figure 3.12 provides a schematic of the Cipolletti weirs dimensions, where (H) represents the height of water owing over the crest. Figure 3.12: Cipolletti weir dimensions 26 Further upstream, the second runo measurement device a fully contracted rectangular weir was constructed. Completed in September 2012, it measures ows from approximately 60% of the contributing watershed area. The construction process followed the same proce- dure as the Cipolletti weir discussed previously. However, this fully contracted rectangular weir provides runo data that is accurate between 1:5% to 2:5% according with the Water Measurement Manual, (Dodge, 2001). Figure 3.13 shows the (10cmx10cmx5m) post placed in the stream to support the weir walls. Stream cross-section dimensions here are smaller, measuring approximately 3.4 m across and 1.2 m deep, with irregular side slopes. The nished weir has crest length of 0.67 m and height of 1.5 m, see Figure 3.15. Figure 3.13: Support post for the rectangular weir 27 Figure 3.14: Upstream view of rectangular weir Figure 3.15: Rectangular weir dimensions At both locations, wooden posts with level loggers (HOBO U20, pressure head range 4 m H2O, accuracy 0.014 m H2O) have been positioned at a distance of four times the weirs maximum measured head upstream from the crest of the weirs, following the recommenda- tions of Dodge (2001). This was determined to be approximately 2.44 m upstream from both weirs. The large distance ensures that the readings are not a ected by draw down associated 28 with the discharge over the weirs. Sensors were placed on the lateral of the posts parallel to the ow, see Figure 3.16. To mark the start of an event, 6.1 cm of head is required to consider the ow fully-developed over each weir, Dodge (2001). Level logger data collection was continuously measured at 15-minute intervals throughout the study period. Figure 3.16: HOBO pressure sensor installed upstream at each weir As mentioned, the stream bed is mostly comprised of a highly erodible soil that has caused many challenges in maintaining the integrity of weirs. Throughout the rst year of service, the Cipolletti weir experienced a few large peak discharges which caused erosion around the side walls. One example of this can be seen in Figure 3.17, where the right side wall experienced large amounts of erosion after a runo event. 29 Figure 3.17: Erosion damage at the Cipolletti weir to the right After several attempts to repair this weir, the decision to convert it to a broad crested weir was made in February 2013. This involved cutting across the level of the crest, creating a weir with a length of 3.7 m and adding a width of 45 cm, see Figure 3.19. Figure 3.18 depicts the changes made to the Cipolletti weir. The remodeling has signi cantly decreased the erosion seen from large runo events insuring it will be able to withstand a longer period of recording. 30 Figure 3.18: Upstream view of broad crested weir Figure 3.19: Broad crested weir dimensions The upstream rectangular weir experienced some minor damages during the monitoring time. Lateral seepage resulting from animal interference began to erode soils out and cause the channel walls to expand. This was repaired by extending the side wall of the weir and expanding the crest length from 0.67 m to 1.2 m. Figure 3.20 shows the modi ed crest dimension and (H) which represents the height of water measured over the weir. Along with 31 the remodeling of both weirs the governing ow equations associated with each were adjusted accordingly. Figure 3.20: Modi ed rectangular weir dimensions 3.2.5 Weir Discharge Equations Both the Cipolletti and fully contracted rectangular weirs were calibrated by referencing methods provided in Chapter 7 of the Water Measurement Manual, Dodge (2001). The mod- i ed broad crested weir was calibrated with assistances from the Handbook of Hydraulics, Brater et al. (1996). Each calibration process is presented below. The Cipolletti weirs discharge calibration was similar to the suppressed weir since the side walls contract the ow over the crest. Dodge (2001) has developed the governing ow equation ignoring the e ects of the approach velocity. Equation 3.11 shows the generic formula available to calculate owrate Q over the Cipolletti weir. Q = 3:367LH32 (3.11) Cipolletti calibration method limitations: Crest length must be at least 0.152 meters (L) Crest Height must be at least 0.102 meters (P) Head measurements must be taken at least four times the maximum head upstream 32 Head measurement must be at least 0.061 meters (H) Ratios of (H/P) must be less than 2.4 Downstream water elevation must be at least 0.051 meters below crest elevations Accuracy is within 5% of determined value, Dodge (2001). Calibration of the fully contracted rectangular weir was completed by applying the Kindsvater-Cater method for determining the head discharge relationship. This method allows for the calibration of weirs that may not meet the crest height limits of traditional rectangular weirs. However, this method has limitations, a few of which are listed below. More detailed discusses can be found in Dodge (2001). Kindsvater-Carter calibration method limitations: Crest length must be at least 0.152 meters (L) Crest Height must be at least 0.102 meters (P) Head measurements must be taken at least four times the maximum head upstream Head measurement must be at least 0.061 meters (H) Ratios of (H/P) must be less than 2.4 Downstream water elevation must be at least 0.051 meters below crest elevations Accuracy is between 1:5% and 2:5% of determined value, Dodge (2001) The Kindsvater-Carter calibration method begins with the use of the basic weir formula, Equation 3.12. Note: Inputs for this method are in U.S. units, but for this study all values were converted into SI units. Q = CeLeH32e (3.12) Where: Q is discharge (ft3s ) Ce is the e ective coe cient of discharge (ft 12 s ) B is the average width of the approach channel Le = L+kb 33 He = H +kh In order to determine the value of kb speci c to the weir of interest, Figure 3.21 must be used. Using the ration of L=B and following from the x-axis up to the plotted line, a value of kb can be obtained. Figure 3.21: Values of width-adjustment factor, taken from (Dodge, 2001) Figure 3.22: E ective coe cient of discharge, Ce, as a function of L=B and H=P, taken from (Dodge, 2001) 34 Then, from Table 3.3, the ratio of L=B is used to determine C1 and C2 found in Equation 3.13. Linear interpolation may be used if necessary. Table 3.3: Coe cient and constants used when determining the e ective coe cient of dis- charge, taken from (Dodge, 2001) Ce = C1 H P +C2 (3.13) Once the coe cients are determined then Equation 3.13 can be updated with varying values of H and Equation 3.12 is continuously calibrated. Since the weir was modi ed due to erosion factors, this process was followed twice to provide a new head discharge relationship. Provided below are the nal calibrated head discharge equations used in this study. Equation 3.14 represents the calibrated equation at original geometry and Equation 3.15 represents the calibrated equation after expansion. Q = 0:0115 H 1:21 + 3:158 2:216(H32 ) (3.14) Q = 0:02059 H 1:21 + 3:161 3:791(H32 ) (3.15) Finally, the broad crested weir was calibrated with reference to the Handbook of Hy- draulics, Brater et al. (1996). Experiments on broad crested weirs have been performed by Blackwell, Bazin, Woodburn, the U.S. Deep Waterways Board, and the U.S. Geological 35 Survey (Brater et al., 1996). Equation 3.16 provides the basic equation used to determine discharge over a broad crested weir. Q = CLH32 (3.16) Where: C is the calibration coe cient L is the length of the crest H is the head measured over the crest The previously mentioned experimental results were combined into Table 3.4 which has been made available by Brater et al. (1996). Using this table, values of C can be determined based on the weirs breadth (B) and the measured head. The broad crested weir in this study had a breadth of 0.45 m so the C values in this column were used to continuously update the ow formula, Equation 3.16. By incorporating a Visual Basic code in a Excel spreadsheet, C values were found using linear interpolation in Table 3.4. Table 3.4: Coe cient, C, values used for broad crested weirs, taken from (Brater et al., 1996) Broad crested calibration method limitations: Crest length must be at least 0.152 meters (L) Crest Height must be at least 0.102 meters (P) 36 Head measurements must be taken at least four times the maximum head upstream Head measurement must be at least 0.061 meters (H) Ratios of (H/P) must be less than 2.4 Downstream water elevation must be at least 0.051 meters below crest elevations Accuracy is between 1:5% and 2:5% of determined value, Dodge (2001) 3.2.6 Groundwater Observation Wells Two shallow groundwater wells were installed, in October 2012, next to the upstream rectangular weir. The rst well is located approximately one meter outside of the stream and reaches a depth of three meters below the surface. Five meters away and located in the center of the stream is the second well which has been drilled at the same depth, see Figure 3.23. Figure 3.23: Shallow groundwater wells 37 Both wells were drilled with an 8.25 cm diameter hand auger. This allowed for the installation of 3.8 cm diameter PVC with screens lengths of 61 cm attached at the ends. A sand pack was placed up to one meter to act as a lter and keep nes from entering. Finally, each well was back lled and capped with a layer of sodium bentonite. Doing so prevented any water owing on the surface to interfere with the shallow groundwater levels. Pressure transducers or level loggers (HOBO U20, range 9 m, accuracy 0.021 m H2O) have subsequently been deployed at these wells. These devices provide ne resolution water level data at 15 minute intervals. During the dates between December 19th, 2012 and January 25th, 2013, various large rainfall runo events hit WS-AGC. Consistent rainfall along with the largest intensity event 45 mmhr occurred in a short period of time and caused large amounts of debris from the surrounding areas to enter the stream. The two groundwater wells su ered major damage as the PVC pipes, which extend about two feet above the surface, were broken. Forces within the stream were so strong that the sensor 3 meters below the surface was lifted out and brought downstream approximately 61 meters. This event is represented by the missing data in Figure 3.36. 3.2.7 In ltration Testing Several locations were chosen throughout WS-AGC to conduct in ltration tests using a AMS 24-inch double-ring in ltrameter. These locations were selected based on topology and soil characteristics provided from the web soil survey (http://websoilsurvey.nrcs. usda.gov). Figure 3.24 shows the location at four separate sub-catchments. At each sub- catchment 4-7 tests were run. The average of each sub-catchment was made and the values were used as comparison to the SWMM models maximum in ltration rate (mmhr ) input. 38 Figure 3.24: Locations of In ltration Testing The procedure followed to run these tests are as follows: Insert double ring into the soil at the test area by pushing handles and rotating back and forth After the double ring has reached at least two inches, ll both inner and outer rings with clean water until they start to over ow Using a ruler or a pre-installed tape measure, record the initial water level within the inner ring and start the timer Note the water level and time of sample as the water begins to drop Continue this test and ll ring as needed until 15 minutes has pasted Take the amount of in ltrated water by 4 to obtain the in ltration rate per hour Depending on the soil types in the area this test may be shorted due to large amounts of water that would be needed to run a 15 minute test Note: Be sure to mind the amount of time the test was run since this will a ect your in ltration rate multiplier 39 3.2.8 Soil Sampling On July 10th and 11th, 2013, soil sampling was conducted across the potential earth embankment location. Collaborating with AGC, a subcontractor from TERRACON Inc was hired to consult and perform the soil sampling and analysis. The site of interest was rst clear cut across the approximate 137 m-long embankment centerline. Then four locations on the centerline and two others approximately 30 m upstream and downstream of the line were chosen for soil sampling, see Figure 3.25. Figure 3.25: Bore sites at WS-AGC indicated by black markers On the trip into the site the 25 thousand pound CME-550 rotary drill rig faced some challenges with steep grades and narrow cut rebreak paths. The weather conditions also provided a saturated upper layer of soil which increased the di culty of the trip, Figure 3.26. 40 Figure 3.26: Drill rig entering WS-AGC Once the rig made it to the sampling location, drilling was commenced. The sampling was conducted using a 15.2 cm diameter hollowed stem auger. Figure 3.27 shows the drill crew setting up the rig to begin drilling and sampling. Figure 3.27: Setting up to drill boring 41 Figure 3.28: Soil exiting the bore hole from approximately 4.6 m below the surface At each drill site standard penetration tests (SPT) were performed at speci ed depths as the auger progressed downward. All samples were collected with accordance to ASTM D1586. Samples taken from these tests were then used for soil analysis, Figure 3.29. At drill site B-1 two bulk samples were taken to conduct grain size distributions, compaction and Atterburg limits. Figure 3.29: Samples taken from the SPT Along with the SPT test and bulk samples taken across the centerline, three Shelby tubes were used in this analysis. At location B-2 just o set from the creek, samples were 42 taken at approximately 1.5, 3.0 and 4.6 m intervals with accordance to ASTM D4220. Figure 3.30 shows one of the Shelby tube samples that was taken to the professional laboratories operated by TERRACON Inc. There the samples were analyzed with a Mercury Permometer test to determine permeability rates. Results from these sampling e orts can be found in the appendix of this document. Figure 3.30: Shelby tubes being prepared to take out of the site 3.3 Fundamentals of Operating SWMM 3.3.1 Introduction The storm water management model (SWMM) was developed in 1971. Since then it has evolved into the current version, SWMM 5.0. Completely rewritten from the previous 43 version, it incorporates a user-friendly interface and new visual e ects to assist with analysis. This version has been used for numerous water-related projects and analysis throughout its operational history, the majority of which has been focused on urban storm water system modeling. It has the capability to simulate single and long-term (continuous) rainfall runo events, an its intuitive user interface has opened the door for less experienced modelers due to its shallower learning curve. Driving the hydrological processes in the program is a computational engine that routes excess runo through a series of sub-catchments, links, nodes, weirs, storage devices and pumps. The engine tracks water quality as well as quantity throughout the constructed model. More details of the internal components of the program are discussed in the following sections below. SWMM is capable of modeling system networks of very large complexities. Channels can be modeled as either opened or closed systems with a variety of size and shapes that the user may input manually or select from prede ned list. Generally a system is comprised of various ori ces, weirs, storage/treatment units, ow dividers and pumps. The program is able to handle all of prior network components using an array of prede ned objects located in the interface. Flow can be routed with three separate user de ned methods. These include steady- state, kinematic wave and full dynamic wave. The user must select a routing method based on the system and outcomes they wish to achieve. For example, if the user selects the full dynamic routing method, the program will take in consideration e ects from backwater, surcharging and reverse ow in the model. Each routing method provides its own unique advantages and are discussed in more detail later in this section. 44 3.3.2 Internal Mechanisms The underlying computational engine of SWMM uses physically-based information to simulate the hydrological processes discussed in the following pages. Principles of con- servation of mass, energy and momentum are used to account for the transport of water, contaminants or sediments through the model. This allows the model to accurately and e ectively simulate runo quantity and storm water quality through the system. Presented in this subsection is background into the algorithms and theory which SWMM uses to simulate the hydrological processes. Surface Runo Figure 3.31 presents a conceptual view of the surface runo process used by the com- putational engine of SWMM. It illustrates the various components, treated as non-linear reservoirs, that water contaminants or sediments travel through once they have been deliv- ered to the sub-catchment by rainfall or another upstream catchment. These components include in ltration, evaporation and surface runo . No matter how water is delivered to the sub-catchment it is rst collected as maximum depression storage. This includes the catchments abstractions such as ponding, surface wetting and interception. Figure 3.31: Conceptual view of SWMM?s runo mechanism, (Rossman and Supply, 2005) 45 Then once this component has reach full capacity, surface runo begins. Runo is routed according with Manning?s equation, Equation 3.17. The depth of surface runo is updated with time by use of a water balance equation. Q = 1nAR23pS (3.17) Where: n is the Manning?s roughness A is the area of channel cross section R is the hydraulic radius S is the slope of channel from one node to the next In ltration Water that is lost from the sub-catchment components of SWMM is generally in the form of in ltration. As the precipitation begins to fall into the sub-catchment, it is percolated through the unsaturated soil zone of the pervious area. To model this phenomena SWMM o ers three choices from which the user can select from. These include the Horton?s, Green- Ampt?s and Curve Number methods. However, Horton?s method was selected for use in this investigation and is descried in detail below. Further details on the other methods can be found in Rossman and Supply (2005). Horton?s Equation: The Horton?s in ltration method is simply based on empirical observations which show that in ltration decreases as an exponential function from the maximum to minimum rate over a rainfall event, see Equation 3.18. The most di cult inputs for this type of in ltration are initial in ltration or maximum in ltration fo and the decay constant k. They dictate the initial in ltration or maximum rate and the rate at which the in ltration will decay over 46 the rainfall period. It is critical to get the closest estimates as possible. Values are usually found in literature de ned by soil characteristics or derived from experimental test. fp = fc + (fo fc)e kt (3.18) Where: ft is the in ltration rate at time (t) fo is the initial maximum in ltration rate fc is the nal or constant in ltration rate once the soil column has become fully saturated k is the decay constant speci c to each soil type SWMM applies a modi ed version of Equation 3.18 that is represented below in Equa- tion 3.19. The modi cations help to account for the recovery of in ltration capacity during dry or no surface ponding time periods within a continuous simulation. These modi cations were created by SWMM developers and Figure 3.32 shows the approach they followed. Values of kd are assumed constant or a scaled value of k from Equation 3.18. fp = fo (fo fc)e kd(t tw) (3.19) Where: kd is a decay coe cient for the recovery curve tw is a hypothetical projected time at which fp = fc on the recovery curve 47 Figure 3.32: Conceptual view of Horton?s in ltration capacity recovery mechanism used in SWMM?s computational code, (Viessman et al., 2003) Groundwater SWMM uses a two-zone groundwater component to model subsurface ows. These two compartments consist of the upper unsaturated and the lower fully-saturated zone. The main di erence between the two being that the saturated zone is assumed to have a constant moisture content and is equal to the soil porosity . However, the upper zone has a variable moisture content of . Each area shown in Figure 3.33 is representing the ux per unit area and each one is described below. 48 Figure 3.33: Conceptual view of SWMM?s groundwater mechanism, (Rossman and Supply, 2005) Where: fI is the in ltration from the surface fEU is the evapotranspiration from the upper zone, de ned as a xed fraction of the un-used surface evaporation fU is the percolation between the upper and lower zones, dependent on the upper zone and the depth dU. fEL is the evapotranspiration from the lower zone, function of the depth dU fL is the percolation from the lower zone to deep groundwater, function of lower zone depth dL fG is the lateral groundwater inter ow to the drainage system, function of dL and channel/node water depth dTOT is the total distance of the upper zone dU depth and lower zone depth dL In order to link the sub-catchment and aquifer component of SWMM Equation 3.20 is implemented. Using the coe cients (A1, A2, B1, B2 and A3) the user is able to de ned the rate of groundwater ow between the aquifer and receiving node. If the user wishes to model the surface groundwater interaction as a simple proportional relationship then exponents (B1 and B2) should be set to 1. Also, the coe cients (A2 and A1) should be equal and A3 49 should be set to zero. Qgw = A1(Hgw E)B1 A2(Hsw E)B2 +A3HgwHsw (3.20) Where: Qgw is the groundwater ow Hgw is the elevation of groundwater table Hsw is the elevation of the surface water at receiving node E is the elevation of receiving node invert Flow Routing Flow routing takes place within the conduits of the SWMM model and between nodes. It is governed by the conservation of mass and momentum equations for gradually varied, unsteady ow. The user is given three choices with which to run their models, including the steady ow, kinematic wave and dynamic wave routing methods. Steady ow routing represents a uniform and steady assumption within each computa- tion time step. The ow hydrograph inputted into each upstream node is assumed to route with no delay or change in shape. Using the Manning equation a relationship between ow rate and ow is formed. There are many limitations with this routing selection including channel storage, backwater e ects, entrance/exit losses, ow reversal or pressurized ow. Using this method is only advised for preliminary analysis for long-term simulations and it is only valid for networks where each node has only one out ow link. More detail is provided in Rossman and Supply (2005). Kinematic wave routing solves the continuity equation as well as a simpli ed form of the momentum equation between each node through the conduits. This simpli cation of the momentum equation includes an assumption that the owing waters surface is equal with the conduits bed slope. Under these assumptions the maximum ow that can be routed is 50 constrained to the full- ow Manning equation value. If the water surface level is above this, then it is either lost or ponded atop of the inlet node. The latter allows water to re-enter the system once surface levels have subsided and reducing the amount of water lost in the system. Under this selection, the model is limited to a system that is represented by nodes that only have a single outlet conduit. This method allows for the modeling of a ow that varies both spatially and temporally which can depict delays in out ow hydrographs from in ow hydrographs. However, it is not able to simulate the e ects from backwater, entrance/exit losses, ow reversal or pressurized ow. One great feature of this method is relative stability of the code. Users can apply this method to long-term simulations with a temporal scale in the range of 5-15 minutes. More detail is provided in Rossman and Supply (2005). The dynamic wave routing method within SWMM solves the 1-D depth averaged momentum and continuity equations referred to as the complete 1-D Saint Venant equations, see Equations 3.21 - 3.22. The Saint Venant terms are solved along each component of a computational cell, over a network of junctions and conduits that represents the physical characteristics. @A @t + @Q @x = 0 (3.21) @Q @t + @(Q2A ) @x +gA @H @x +gA(Sf +hL) = 0 (3.22) Where: Q is the ow rate through the conduit x is the length of the conduit H is the hydraulic head of water in the conduit A is the cross sectional conduit area t is the simulation time 51 Sf is the friction slope hL is the local energy loss per unit length of conduit g is the acceleration of gravity This method allows the user to represent more interesting and realistic scenarios which may occur within the system. For example, if a closed pipe system is the subject of modeling, then pressurized ows which can exceed predictions from the full ow Manning equation value are able to be simulated. Another key feature with the dynamic wave routing method is its ability to simulate channel storage, backwater, entrance/exit losses and ow reversal. This is particularly important when the system includes various ori ces or weirs that may cause signi cant constrictions on the ow. Finally, this method allows the user to simulate systems with any con guration of loops or multiple downstream diversions. However, the drawback of this exibility comes with the smaller time steps and greater computation e ort is needed to maintain numerical stability. 3.3.3 Interface Objects WithinSWMM are two types of interface objects that can be implemented into a model. These consist of the visual and non-visual objects. Visual objects of SWMM include the components which represent the physical environment experienced at the region of interest. Non-visual objects in SWMM include components of the hydrological cycle as well as the inputs time series that drive many of the processes of simulations. They are components that have tremendous impacts in model results; insight and engineering judgment are paramount in their de nition. Parameters and data inputs for both types of objects can be sourced directly from local eld studies or literature studies on sites of similar description. This subsection does not cover the objects available within SWMM?s interface but an in-depth discussions can be found in the SWMM?s Users Manual, Rossman and Supply (2005). 52 3.4 SWMM Model Development Initially WS-AGC was divided into 7 sub-catchments conforming to the local topogra- phy. As the model began developing, the need for additional discretization to depict diverse geophysical characteristics was recognized. To do this the original 7 sub-catchments were broken down into smaller sections, 15 in total. Each new sub-catchment was selected based on soil types predicted from the NRCS/USDA soil survey Figure 3.34. Figure 3.34: Discretized Sub-Catchments based on topography and soil types for WS-AGC Geographic Information Systems (GIS) were then used to improve the accuracy of phys- ical input parameters such as sub-catchment area, slope, channel lengths and storage reser- voirs. Main channel locations and lengths were also established from a 10 m DEM provided by the USGS online data website (http://ned.usgs.gov/) and manipulated with ArcGIS software. There terrain features were then con rmed from eld investigations. Channel cross sections were input as irregular shapes (transects) determined from eld survey mea- surements, see gure 3.35. The cross sections were chosen while walking the streams at 53 WS-AGC. A survey was completed to use as model input at locations where major transi- tions occured. With such detailed site-speci c information, a well-built foundation for the initial stages of modeling was created. Figure 3.35: Surveying a downstream cross section As model setup continued, estimations of more subjective parameters became neces- sary. This included Manning roughness (n) for channel/overland ows, depression storage, minimum/maximum in ltration rates, saturated hydraulic conductivity and eld capacity. These were estimated based on published values from literature and insights obtained from several eld visits , Table 3.5. 54 Table 3.5: Literature values of maximum and minimum in ltration rates for Horton Equa- tion, (Akan, 1993) Results from the USDA/NRCS web soil survey assisted with determining the range of values for each sub-catchment in ltration parameters, Table 3.6. These ranges provided exibility when performing the model calibration; more discussion is provided ahead. Stream beds at the site have many abstractions including vegetal debris, partial ob- structions and pooling areas where ow is subjected to large head losses and varying ow conditions. Manning equation within SWMM uses n values (roughness) that provide ow frictional losses. Thus a single value of n for each conduit had to be estimated to represent the complex nature of the channels. Mannings n values used for channels simulation ranged from 0.04 to 0.4, following Rossman and Supply (2005). Such values are consistent with natural channels of irregular sections with pools and having a vegetation cover. 55 Sub-catchments estimates for overland ow roughness, depression storage and sub- catchment width were mostly obtained with published values and further modeling assump- tions, (Rossman and Supply, 2005). Mannings n values for overland ow were estimated as 0.8, following McCuen et al. (1996) estimation for dense underbrush in wooden regions. The depression storage value of 7.62 mm (1/4 in) was adopted from the recommended rst estimate value in SWMM 5 Users Manual, (Rossman and Supply, 2005). To calculate each sub-catchments width an initial assumption is made that the overland sheet ow will not occur more than 150 m before reaching or transitioning to channelized ow. To distinguish individual runo events from observed data the minimum inter-event time was selected as six hours and each event?s peak discharge must meet a minimum peak ow rate of 0.1 m3=s; any event below this was not considered in the study. In addition, various events were extended as needed since water levels in the stream uctuated during the lower portion of the recession curve. This caused readings of water level to dip above and below the required 6.1 cm of head required for ow to fully develop over the weirs, (Dodge, 2001). Results from groundwater monitoring in the well positioned in and outside of the stream bed, presented in Figure 3.36, motivated the inclusion of the aquifer component of SWMM. Blue bars in the top chart correspond to rainfall intensity, whereas the blue line in the lower chart correspond to the average accumulated rainfall depth measured by the rain gauges deployed in the site. The red lines in both charts correspond to groundwater elevation, and the interruption corresponds to a period of malfunction described earlier on page 35 groundwater observation wells. 56 Figure 3.36: Shallow ground water level relative to the surface for the out of stream obser- vation well Despite of the interruption on the groundwater level measurement, it can be noticed that the initial rain events up to December 19, 2012 have not resulted in any signi cant changes in the groundwater level. Then after the damage was xed the groundwater level was much larger, and at that point connected to the stream. As is shown, even small rain events caused measurable and immediate increase in the groundwater level. This dramatically contrasts with the earlier condition, which seems to indicate that the groundwater is disconnected from the stream. Such complexities of inter ow and its ability to produce longer recession curves in the wet season made the groundwater module a key addition to SWMM modeling. 57 3.5 Model Calibration Calibration for the developedSWMM models was completed using nine separate recorded rainfall-runo events from 06/11/2012 to 02/14/2013. This time frame provided data from the below average rainfall year, 2012, and the beginning of the above average wet year, 2013. As mentioned, events were distinguished by inter-event periods of at least six hours and a minimum peak ow rate of 0.1 m3=s. The calibration between observed and computed rainfall-runo events was conducted using the Sensitivity-based Radio Tuning Calibration (SRTC) function of PCSWMM. During this process eight parameters were adjusted, lim- iting their uncertainty rankings to 40% and below, with nine individual runo events. The eight calibration parameters can be found in Tables 3.6 and 3.8. Calibration e orts using these guidelines included observed events during the dry season (May-November) and wet season (December-April). Following the work of Davis et al. (2007), peak ow rates were examined in addition to ow duration curves for each model con guration. Both compar- isons provided calibrated models with respect to peak ow conditions as well as the more challenging issue of ow volumes. The SRTC tool works by designating uncertainty percent rankings for each parameter of interest. For e cient calibration, rst estimates of the model parameters should be as close as possible to the true value (James et al., 2002). Also, James et al. (2002) suggests that percent rankings should be limited to 50% as an absolute maximum. This insures that calibration parameters will not become arbitrary values outside of meaningful range for the area of interest, (AOI). Another important aspect to calibration is having a su cient amount of rainfall-runo events for the number of parameters of interest. James et al. (2002) points that the amount of calibration parameters must be limited to the amount of individual rainfall-runo events observed and selected for calibration. Following this rationale a total of nine separate observed rainfall-runo events were used to calibrate eight parameters. In all events, the hydrograph obtained at the downstream (Cipoletti weir) was used as observed reference data. 58 Horton?s method was selected to model the in ltration at the site. The four main input parameters include maximum in ltration rate, minimum in ltration rate, decay constant and drying time. The most sensitive of the four in ltration parameters included mini- mum/maximum in ltration rate and decay constant. Each sub-catchment was assigned an average of the maximum and minimum in ltration rate based on a range found from ex- amining the local soil types. These values were provided from Akan (1993) and are seen in the previous subsection, Table 3.5. Estimates for decay constants and drying times were made from an average of the range provided by the SWMM 5.0 Users Manual, (Rossman and Supply, 2005). During the early stage of calibration the drying time was found to have minimal e ect on the results so it was xed at the maximum value of 8 days. The calibrated in ltration parameters of maximum/minimum in ltration rates and de- cay constants were limited to 40% of the original estimate. Large diversity in soil types observed in the eld led to the assumption that this approach would provided enough exi- bility while keeping in ltration parameter values within a reasonable range. Ranges of values used for Horton?s in ltration method are shown in Table 3.6. Table 3.6: Minimum and maximum ranges used for Horton?s in ltration input parame- ter,(Akan, 1993) and (Rossman and Supply, 2005) Where: fo is maximum in ltration rate (mm/hr) fc is nal/minimum in ltration rate (mm/hr) DC is decay constant DT is Drying time (days) 59 Multiple attempts were conducted to calibrate a SWMM model without using the aquifer component. However, these calibration e orts did not result good agreement between modeled and observed hydrographs. This became the main justi cation to implement the aquifer component of SWMM. More calibration parameters were introduced with respect to the groundwater component including bottom groundwater elevation, saturated hydraulic conductivity, eld capacity and conductivity slope. The range of tested values for calibration parameters were derived from Table 3.7, (Rossman and Supply, 2005). Note that no changes were implemented in the calibrated values for sub-catchment inputs with the introduction of the groundwater components. Table 3.7: Aquifer properties for various soil types, taken from (Rossman and Supply, 2005) During calibration involving aquifer components the SRTC was not able to handle the values of bottom groundwater elevation, so a series of trial and errors had to be performed for each individual aquifer component. The lack of eld data with respect to the location of the aquifer?s bottom elevation and the variability of site conditions rendered this value somewhat arbitrary. Designer?s must use personal judgment to remedy each components bottom depth. PCSWMM sets a default value of 10 for the conductivity slope and during 60 calibration slight tuning of this value provided a better t between observed and computed hydrograph recession periods. Lastly, ranges for saturated hydraulic conductivity, porosity, eld capacity, and wilting point values were determined based on soil types and texture classes from Table 3.7. These ranges are presented in Table 3.8. First inputs values were based on average values in order to provide a comprehensive description of the AOI. SRTC was then used with percent rankings limited to 40% and below to insure results would fall within the prescribed ranges. After the ne tuning of input values, results across the entire spectrum became much more satisfactory with the introduction of the aquifer component. Table 3.8: Minimum and maximum ranges used for aquifer component input parameter, (Rossman and Supply, 2005) Where: K is hydraulic conductivity (mm/hr) is soil porosity FC is eld capacity WP is wilting point 61 Chapter 4 Results and Discussion As previously stated, eld data has been collected at WS-AGC since February 4th, 2012, and the following sections present and discuss such data collected until August 2013. Also included are results from temperature based evapotranspiration calculations and soil analysis performed on cores samples taken at the potential embankment centerline. These allowed for calibration/veri cation of the SWMM modeling e orts. 4.1 Collected Field Data 4.1.1 Precipitation The local precipitation at WS-AGC was collected from February 4th, 2012 until August 14th, 2013. Presented in Figure 4.1 are the rainfall events (mmhr ) experienced at the site during this study. Notable rain events include May 14th, July 3rd, September 3rd, Decem- ber 25th, February 10th, April 11th, July 23rd, July 30th and August 14th, 2012 to 2013 respectively. These events all produced rainfall intensities of at least 40 mmhr and up to a maximum of 78 mmhr . Though these events had comparatively short duration, they produced signi cant volumes of precipitation. For example, the September 3rd, 2012 event produced a rainfall intensity of 80 mmhr lasting 30 minutes. This event alone produced a total volume of approximately 116;000 m3 of water over the entire site. 62 Figure 4.1: Rainfall recorded at WS-AGC (February 4th - August 14th, 2012-2013) 63 While conducting this study two o -site gauges were used as quality control. Each gauges location relative to WS-AGC is shown in Figure 4.2. Furthest north from WS- AGC was the Seale 1.4 W station and opposite was the Eufaula Wildlife Refuge station. Presented in Figure 4.3 is a comparison between monthly rainfall totals from all gauges used in this study. The three on-site gauges show very close agreement throughout the study as expected. Slight variations in total recorded rainfall were most likely caused by non-uniform rainfall events passing over WS-AGC. Examining the two o -site gauges, a tendency of larger uctuations is seen through the study period. This di erence was most likely caused from the 15.2 and 17.5 km distance each o -site gauge was North and South respectively from WS-AGC. However, these gauges produced rainfall amounts which remained comparable in magnitude with on-site observations. Figure 4.2: O -site rain gauges used as quality control) 64 Figure 4.3: Onsite vs o -site monthly rainfall totals (February 4th - August 14th, 2012-2013) In Table 4.1 statistics for the longest period of locally recorded rainfall during 2012 is presented. With a rainfall total of 924.2 mm over the 11 month period, the amount comes in 263 mm less than the normal yearly total for this area, shown in Table 4.3. If the January normal rainfall from Table 4.3 was added to this total, the yearly precipitation for 2012 would still fall 165 mm below the yearly average for the area. With this observation, 2012 can be considered as a below average rainfall year. 65 Table 4.1: Rainfall statistics for (February 4th - December 31st, 2012) Now examining Table 4.2, total rainfall for the 8.5 month period is 1144 mm. Just 43 mm below the yearly average total with 3.5 months left in the year. If normal rainfall totals from Table 4.3 were added to this current amount, then yearly rainfall for 2013 would be approximately 1548 mm. This results in yearly rainfall exceeding yearly average by 361 mm. However, with 3.5 months of missing data it?s impossible to determine what the actual rainfall total will be since there are many other factors to consider. Even with unpredictability of the last few months of precipitation it is likely that 2013 will have rainfall precipitation above an average year. Table 4.2: Rainfall statistics for (January 1st - August 14th, 2013) 66 Table 4.3: Normal Monthly Rainfall Totals (Period of Record 1981-2010)-Location: Colum- bus, GA (Airport) Average monthly rainfall data recorded at each rain gauge in WS-AGC is presented in Table 4.4. As mentioned previously local rainfall data was not available for January 2012 and for the full duration of August 2013. These periods are acknowledged as N/A in Table 4.4. Data provided from the west gauge during July and August 2012 was much lower when compared to the other two locations. This was due to a malfunction in the recording device and it was taken o ine for repairs for a period of time. A closer look at this table indicates that in September and December 2012 each gauge has recorded rainfall over the long term averages. On the contrary February to April and October to November 2012 have seen precipitation totals well below long term averages. Data in 2013 has shown a complete di erent picture for monthly rainfall totals. August of this year will not be considered in this discussion since data was only recorded through half the month. With this consideration January, March and May are the only months which have seen a signi cantly lower then long term average rainfall. These three months showed monthly percent di erence ranging from 0.9 to 44.4 percent lower then long term averages, Table 4.5. Each of the remaining months have seen rainfalls ranging from 5.3 to 194.6 percent larger monthly rainfall totals, as presented in Table 4.5. Table 4.4: Monthly rainfall average and yearly totals for (February 4th - July 31st, 2012- 2013) 67 Table 4.5: Percent di erences between monthly rainfall average (February 4th - July 31st, 2012-2013) 4.1.2 Temperature The local temperature data recorded at WS-AGC has been collected from February 2012 to August 14th 2013. Throughout the period of record there have been several equipment malfunctions. Therefore, o -site records from the Columbus, GA airport (40 km away from WS-AGC) were obtained through the National Oceanic and Atmospheric Administration (NOAA) online data sets. Figure 4.4 depicts that recorded data from Columbus, GA airport which was used in this study. Red bars indicate normal monthly temperatures found from a 29 year period of record (1981-2010) taken at the Columbus, GA airport (NOAA) and gray bars represent daily average temperatures at the same location over the course of this study. Examining daily temperatures in Figure 4.4 a few instances in which the daily recorded temperature has spiked well above the monthly average are seen. These have occurred during January, February and January, 2012 to 2013 respectively. Where daily temperatures have spiked from 5 to 13 degrees warmer then average. 68 Figure 4.4: Temperature recorded at Columbus, GA Airport (January 1st - August 14th, 2012-2013) Table 4.6 displays the monthly average temperatures along with the 29 year averages taken at the Columbus, GA airport. Columbus?s monthly temperature averages during the period of study have not shown any major uctuations from the long term averages. However, an interesting observation occurs during the 2012 records. Temperature stayed slightly above average for the rst ve months of the study. Besides this slight over average time frame the collection of data shows that the study site did not experience any major outlying points with respect to monthly temperature. In fact, the period of study has indicated that long term normal monthly average temperatures could be applied for short-term studies. Table 4.6: Monthly Recorded Temperatures vs Normal Monthly Temperatures (Period of Record 1981-2010)-Location: Columbus, GA Airport 69 4.1.3 Potential Evapotranspiration (PET) As stated in the previous chapter, PET was a paramter that neede to be indirectly computed as an input data for theSWMM watershed model. This calculation was developed using two separate well-known, temperature based PET methods. After examining Lu et al. (2005) and Sun et al. (2002) studies on watershed hydrology and PET methods, it was anticipated that Hamon?s method would produce accurate PET data. Lu et al. (2005) showed that when comparing advanced radiation-based PET methods (i.e., Turc (1961), Makkink (1957) and Priestley and Taylor (1972)) to temperature-based methods, Hamon?s method produced the highest coe cients of correlation. This study has also shown two temperature based PET methods, Thornthwaite and Hamon, to have the highest correlation coe cient (R) value of 1.0, (Lu et al., 2005). Hargreaves-Samani method produced the lowest correlation with (R<=0.89) between the three temperature PET strate- gies, (Lu et al., 2005). The latter was selected as the second calculation alternative for PET. Figure 4.5 shows daily PET (mm/day) calculations performed using both methods pri- orly discussed. As seen the Hamon method has produced daily PET values which are signi - cantly higher than the those produced from the Hargreaves method. However, both methods have a tendency to follow the same patterns throughout the year. Trends show lower PET values during the colder shorter Winter days and higher PET values during the warmer longer Summer days. 70 Figure 4.5: PET calculations using Hamon?s and Hargreaves Methods In Table 4.7 the full yearly 2012 total for PET from both PET methods used in this study are shown. Since 2013 data was only available until August 14th and not an entire year it can?t be included in this table. The Hargreaves method produces approximately 60 % less PET estimates than the Hamon?s method during 2012. Also, yearly Hamon PET totaled to a value that is the most consistent with other watershed studies in literature. Studies including those of Sun et al. (2010) have found PET values in the Southeastern USA LCP regions to be in the range of 575-1792 mm/year. Table 4.7: Yearly PET totals calculated from Hamon and Hargreaves Methods This once again reinforces the choice in using Hamon?s method to represent the PET at WS-AGC. Continuing research e orts at WS-AGC should provide more con dence in this methods ability to produce reliable results. However, at this point in the study Hamons has proven to be the most reliably temperature based PET method. 71 4.1.4 Runo During the SWMM calibration e orts, rainfall runo data from the rst seven months (06/11/2012 to 02/14/2013) of this study were used. This included nine separate rainfall runo events which fell into the requirements of a six hour inter-event period and a maximum peak ow rate of at least 0.1 m3s . These nine events are represented by the blue lines seen in Figure 4.6. Runo events were scarce during the rst six months of this study as this was the dry season (May-November). Only three events were recorded at the Cipolletti during this period with the highest peak ow recorded at 0.7 m3s . Even with a strong intensity storm of approximately 80 mmhr recorded in September, the dry conditions allowed for the majority of rainfall to be in ltrated. Figure 4.6: Rainfall runo events at Cipolletti weir used during calibration e orts Once the wet season (November-April) began, a dramatic increase in the response of runo was observed. Storms during this period did not show much variation in intensity or duration but runo responses increased noticeably. In a two month period, six of the nine runo events used for calibration were obtained. One possible explanation for this increase 72 in stream ow recorded at WS-AGC is the rise in the local groundwater elevation. Figure 4.14 in the following subsection provides evidence of a water table level increasing from November to December to approximately stream bed elevation. Then during the following seven months groundwater elevations continued to connect and disconnect with the stream bed elevation. Veri cation was performed with rainfall runo data from 02/14/2013 to 08/14/2013. As anticipated runo over the converted broad crested weir continued to produce frequent ows during the wet season. However, runo continued strong into the dry season as ten additional runo events were recorded. Total rainfall from June to July 2012 was 189 mm and rainfall from the same period in 2013 was 443 mm. This large increase in rainfall from the prior year has provided an abnormal dry season. The rainfall runo events in Figure 4.7 were used during this veri cation e ort. Results of veri cation are discussed in the following section. Figure 4.7: Rainfall runo events at broad-crested weir used during veri cation e orts 73 Runo recorded at the upstream rectangular weir began in September 2012. Data from this weir is shown in Figure 4.8. The rectangular weir was not used for any modeling or watershed analysis in this study. However it was used to verify runo consistency from the Cipolletti to modi ed broad crested weir. Figure 4.8: Rainfall runo events at rectangular weir Similar trends in peaks and recession periods are seen between the upstream rectangular weir and downstream Cipolletti/Broad-crested Weir. Figures 4.9 and 4.10 show recorded runo comparisons between these two locations prior to the construction performed. In Figure 4.9 an almost identically shaped hydrograph is seen at both locations with di erences in peak ow rates and time to peaks. The average lag time between the recorded peaks at the upstream rectangular weir and downstream weir is around 0.5 hours. 74 Figure 4.9: Comparision between runo at Cipolletti weir and rectangular weir (Pre- Construction February 8th-14th, 2013) The rainfall runo events in Figure 4.10 represent post construction of Cipolletti to broad crested weir at downstream location records. The runo recorded at these two location still show similar hydrograph trends and the lag time between weirs is near the same value of 0.5 hours. As mentioned prior rainfall runo data from this post-construction period was used as veri cation of the SWMM model. Figure 4.10: Comparision between runo at broad-crested weir and rectangular weir (Post- Construction February 23rd-26th, 2013) 75 Presented in Table 4.8 are descriptive statistics for all the rainfall runo events recorded at the downstream weir. Over the course of the study 31 events matching the requirements of minimum peak ows of 0.1 m3s and inter-event periods of at least six hours. Data is presented as a whole over the dry and wet seasons 2012 through 2013. Total ow volumes were normalized by watershed area in order to obtain units of mm allowing for the computation of runo rainfall ration (R/P). Also rainfall totals presented were determined using any rainfall events within 24 hours prior and during the runo event. This was arbitrarily selected to distinguish between rainfall events which did not directly contribute to runo . Table 4.8 also presents the maximum, minimum and average values for rainfall runo statistics. The longest runo event recorded at the downstream weir was 218.3 hours or approximately 9 days long while the shortest was only 13.4 hours. Interestingly the longest duration event has not provided the highest ow rate recorded over this weir. This has occurred July 23rd, 2013 with an event lasting 34.9 hours peaking at 10.2 m3s and a rainfall total of 32 mm. While the longest event July 3rd, 2013 of 218.3 hours produced a peak ow of 0.85 m3s and rainfall total of 90 mm. Table 4.8: Descriptive rainfall runo statistics 76 Runo rainfall ratios have shown dramatic uctuations through the study period. Rang- ing from 0.02 to 0.87 and having an average value of 0.31. One intriguing event during this record was event 12, which started at 03=02=13. During this event no rain was recorded on any of the rain gauges but the downstream weir experienced a runo event lasting 25 hours and producing a peak ow of 0.28 m3s . This event seems to only be explained by a very concentrated rainfall event which had to fall in-between or upstream of recording rain gauges. One other explanation could be water released from the small wetland storage area just upstream of the rectangular weir. However, since WS-AGC is at the headwaters this scenario lacks any de nitive proof to its origins. Data presented Tables 4.9 to 4.11 separates the observed rainfall runo events into wet and dry season events for each year of record. As de ned in Chapter 2, wet and dry seasons were speci ed as (December-April) and (May-November) respectively. One obvious remark from examining these charts is the skewness between events recorded in 2012 to 2013. With this feature acknowledged, there are still some interesting comparisons to be made from this data. Table 4.9: Descriptive rainfall runo statistics and R/P ratios for wet season events 2012- 2013 Dry season R/P values how shown a much larger uctuation from year to year. In 2012 the maximum ratio experienced was 0.06 where as in 2013 a ratio of 0.87 was recorded. These values show the large variability in rainfall seen at the site in 2012 compared to 2013. The number of runo events recorded has increased by approximately 6 times between the 77 months of June and July 2012 to 2013. Over the entire dry seasons the average R/P values has also increased by 10.3 times from 2012 to 2013. Table 4.10: Rainfall runo statistics and R/P ratios for dry season events 2012 Table 4.11: Rainfall runo statistics and R/P ratios for dry season events 2013 Comparing Table 4.8 R/P statistics with R/P statistics formulated from a 13 year study by La Torre Torres et al. (2011), similar trends were experienced at WS-AGC. The study site of La Torre Torres et al. (2011) shows very similar physical characteristics of WS-AGC with the major di erence being the watersheds area of 72.6 km2 compared to the 2.9 km2 of WS-AGC. Interestingly, even with the signi cant catchment area di erence between the responses of each watersheds hydrology are very similar. R/P ratios from La Torre Torres et al. (2011) show ranges during the entire study from (0.01 to 0.80). Similarly WS-AGC has experienced R/P ratios of (0.02 to 0.87) over the course of study. Now when breaking R/P ratios down into wet and dry season tables, similar trends are seen. La Torre Torres et al. (2011) produced R/P ratios from (0.17 to 0.53) and (0.01 to 0.80) during wet and dry seasons respectively. This again is seen when 78 WS-AGC?s R/P ratios are distinguished by wet and dry seasons. Table 4.9 shows R/P values during the wet season from 2012 to 2013, R/P ratios range from (0.02 to 0.51). Now combining Tables 4.10 and 4.11 R/P values range from (0.03 to 0.87), corresponding with results seen from the study presented by La Torre Torres et al. (2011). 4.1.5 Groundwater Recording the shallow groundwater table at WS-AGC began November 1st, 2012 with two shallow wells. One located outside of the stream and another centered in the stream bed. Data from both of these locations are displayed in Figures 4.12-4.15. Water table elevations are represented as meters from the surface (red lines), where both wells are referenced to the out of stream well ground elevation. Also included within these gure rainfall intensities and cumulative totals, (blue lines). During late December 2012 a major storm hit the area and caused damage to both wells. This is re ected in the missing data between December 2012 and January 2013 in these gures. In both wells groundwater elevation values were approximately two meters below the surface prior to this damaging large event. However, after the event water table elevations had risen to approximately 0.6 meters below the surface. Then as the record progresses, groundwater elevation begins to uctuate above and below the grade elevation. This has then resulted in a dynamic groundwater table which became at certain times directly connected with the stream ows. The largest groundwater table elevations recorded during this study period were approximately one meter above surface elevations. These results indicated that the hydrological processes occurring at WS-AGC are typical of losing/gaining of intermittent streams. Figure 4.11 provides a visual representation of the dynamic water table variation which is likely to be representative of WS-AGC. As the seasons change groundwater elevation vary according to rainfall, evaporation and in ltration processes. As wet season occurs water table levels begin to rise due to larger rainfall events, lower temperatures and decreased PET. This 79 increase in rainfall and decrease in ET creates larger head potential between groundwater levels and stream surface levels. This is represented in Figure 4.11 (a). On the other hand, during the dry season lack of constant rainfall and higher temperatures cause larger rates of water loss in the watershed system. Represented in Figure 4.11 (b) part or all of the stream ow which may occur during a large intensity or duration rainfall event in ltrates through the channel interface into the aquifers below. Figure 4.11: Representation of the dynamic groundwater processes of a gaining/losing stream such as the one at WS-AGC, (Winter, 2007) 80 Figure 4.12: Instream ground water level vs rainfall intensity Figure 4.13: Instream ground water level vs cumulative rainfall 81 Figure 4.14: Out of stream ground water level vs rainfall intensity Figure 4.15: Out of stream ground water level vs cumulative rainfall 82 These observations were signi cant with regards to the development of the SWMM watershed model. Physical evidence that groundwater elevations were gaining to above channel bottom grade levels showed the importance of including the aquifer component in modeling e orts. With the larger number of rainfall events occurring during the 2013 dry season, groundwater elevations have maintained comparatively higher levels. Comparing the levels at the end of the dry season 2012 to 2013 dry season levels, a di erence of approximately two meters has been recorded. As stated prior, the higher groundwater elevations in 2013 have been a major factor in the increase of observed runo events. During the dry season 2012 several storms hit WS-AGC and produced no runo , see Figure 4.6. Compared with the data of 2013 many other smaller storms with intensities below 30 mmhr have produced runo , see Figure 4.7. For example, shown in Figures 4.16 and 4.17 are two similar rainfall events during the dry seasons of 2012 and 2013 respectively. This data provides more evidence that higher groundwater levels change the dynamics of the rainfall runo relationships at WS-AGC. Figure 4.16: Rainfall event without runo event during August 6th, 2012 at Cipolletti/Broad- Crested weir 83 Figure 4.17: Smaller rainfall event yielding runo during July 20th, 2013 at Cipolletti/Broad- Crested weir Another interesting phenomenon recorded at WS-AGC was the evidence of diurnal ET. Several studies have recorded this process in the Southeastern region of the USA including Czikowsky and Fitzjarrald (2004), Gribovszki et al. (2008) and Gribovszki et al. (2010) . These studies have recorded shallow groundwater elevations with similar patterns of diurnal uctuations seen at WS-AGC. Due to the large density of vegetation in the riparian zone, ET has caused groundwater levels to drop diurnally at WS-AGC. For example, Figure 4.18 shows groundwater uctu- ations on the magnitude of 2.8 cm over a 12 hour period. The main decreases in water table elevation occur during afternoon hours and recharging periods occur through night time hours. This process also shows the largest rates of loss during long hot days of the dry period. It?s during this time that vegetation in WS-AGC is in full bloom and natural evaporation/transpiration processes are occurring. While in Winter when vegetation has gone mainly dormant for the year, signs of ET tend to decrease to smaller amplitudes. 84 Figure 4.18: Signs of ET from the out of stream ground water well 4.2 SWMM Simulation Comparisons Presented in this section are results from three SWMM models comprised of multiple aquifer, single aquifer and no aquifer con gurations developed to simulate watershed be- havior at WS-AGC. Comparisons of each models ability to predict various key hydrological behaviors such peak discharges, mean discharge, total volumes and ow duration exceedance are analyzed. Using graphical techniques as well as statistical approaches suggested from Moriasi et al. (2007) each con guration is examined and assessed. 4.2.1 Hydrographs Comparison Hydrographs allow the modeler to visually inspect simulation result to observed values and help to identify model bias toward timing as well as recession curves (Moriasi et al., 2007). The following subsection provides graphical representation of key rainfall runo events modeled from each con guration. Each con guration with three separate events ranging from dry to wet periods are displayed in Figures 4.19-4.21. Output hydrographs across both dry and wet periods from the multiple aquifer veri ca- tion results have shown responses matching peaks and duration from observed hydrographs. The top chart from Figure 4.19 shows the models ability to handle the initial ow events by matching peak ows and duration. It however is not able to handle the large ow event that 85 follows with a peak of approximately 2.4 m3s . Proceeding to the second chart from the top in Figure 4.19 a longer duration rainfall runo event starting July 3rd, 2013 is presented. Recession curves during this event last for several days. This con guration simulates the rst portion lasting till July 5th well with respect to peaks and duration. As another rainfall event occurs on WS-AGC during the late afternoon of July 5th a spike in water level occurs. The simulation once again is able to capture this but then begins to recede quickly. This causes a misrepresentation of the slow recession observed in eld measurements. Finally the last chart of Figure 4.19 displays three separate rainfall runo events commencing on July 14th, 2013. Veri cation simulations here produced results which t observed data the best from the two latter charts. Hydrograph results from the single aquifer veri cation simulation have produced results which have generally under predicted runo from all three events shown in Figure 4.20. The top chart of Figure 4.20 displays similar results as the multiple aquifer con guration. Early events beginning February 22nd, 2013 are simulated well with peaks ows and duration corresponding to one another. Then just as seen in the multiple aquifer con guration, simulated runo during February 26th does not agree with observed data. In the second chart from the top of Figure 4.20 simulation runo agrees well with the initial portion of the July 3rd, 2013. Then is unable to model the second halves of the observed events peak and recession. Following similar trends the simulation results displayed in the bottom chart of Figure 4.20 have under predicted peak ows during these three runo events. However, this con guration has reproduced recession curve behavior fairly well. Finally when examining the no aquifer con guration simulation results displayed in Figure 4.21 dramatic di erences are noticed. Runo output from each separate event fell far below observed hydrographs. This was expected since the same outcomes occurred during the calibration period of this analysis. Results were carried through to veri cation for comparison purposes. They have also provided visual evidence reinforcing the importance of using aquifer components in WS-AGC?s model. 86 Multiple Aquifer Simulation Figure 4.19: Output hydrographs produced with multiple aquifer con guration 87 Single Aquifer Simulation Figure 4.20: Output hydrographs produced with single aquifer con guration 88 No Aquifer Simulation Figure 4.21: Output hydrographs produced with no aquifer con guration 89 4.2.2 Flow Duration Exceedance curves Presented in this subsection are ow duration exceedance curves produced from simu- lated and observed runo data during the veri cation period. These gures provide insight into the models ability to capture the range of ow rates recorded in the eld. In Figures 4.22-4.24 blue lines represent observed runo at the downstream weir and red lines represent computed runo . Results from Figure 4.22 show the multiple aquifer con gurations ability to reproduce runo recorded at WS-AGC. Flows in the mid range between 0.7 and 10 % of total duration were modeled at a high level of accuracy, whereas the larger peak ow rates, which occurred from 0.09 to 0.7 % of total duration, were not computed by this con guration. The computed ow line also begins to drop at a faster rate as ows begin to drop lower then 0.04 m3s . This is due to under prediction of the recession curves duration at low ow rates during the simulations. This may suggest that parameters in the aquifer component may need further adjustment to allow for a longer recession curve during a rainfall runo event. Figure 4.23 shows similar results from the single aquifer con guration to the prior dis- cussion. However, results during the period between 0.7 and 10 % of total duration show a larger di erence between computed and observed values. This means that not only were peak discharges somewhat misrepresented but mid range and low ow conditions were also below observed data. The single aquifer con guration was able to produce a curve that followed the observed curves slopes and patterns well. On the contrary this con guration was unable to produce any ows that matched observed data set directly. The no aquifer con guration produced the least tting of all the ow duration ex- ceedance curves. Computed ows from highest to lowest were all well under predicted. This result showed the signi cance of adding in the aquifer component, especially where mid range and low ows occurred. The lack of contribution from a slower releasing groundwater component was one of the major issues in capturing the large amount of low ow events occurring during this veri cation period. 90 Multiple Aquifer Simulation Figure 4.22: Veri cation Flow duration exceedance curves produced with multiple aquifer con guration Single Aquifer Simulation Figure 4.23: Veri cation ow duration exceedance curves produced with single aquifer con- guration 91 No Aquifer Simulation Figure 4.24: Veri cation ow duration exceedance curves produced with no aquifer con gu- ration 4.2.3 Simulation Error Analysis The following subsection aims to provide visual displays of observed versus computed data points for three rainfall runo parameters. Each con gurations performance was ana- lyzed with respect to its ability to simulate peak discharges, mean ows and total volumes observed in the eld. Liner regression plots were determined with each simulated runo event plotted as a black dot with respect to observed values. Percent envelopes were also included to display a 10% and 30% range of values from the 1 : 1 linear regression line. Any data points that fall within 30% of the regression line were considered to be satisfactory. Following this subsection are statistical summaries from both the calibration and simulation periods. 92 Peak Discharge Multiple Aquifer Simulation Peak discharge comparisons from the multiple aquifer con guration are presented in Figure 4.25. Each data point represents a peak discharge value produced from a simulated rainfall runo event during the veri cation period. Simulated peak ows agree well with observed values for events up to approximately 1.25 m3s . Any peak ow above this threshold was under predicted by the model. This produced a few outliers especially with a 10 m3s ow experienced at WS-AGC. Figure 4.25: Veri cation error analysis for max ow during multiple aquifer simulation Single Aquifer Simulation The single aquifer con guration results presented in Figure 4.26 shows similar trends to those seen in the multiple con guration. Peak ows however were not predicted as well as the prior con guration. Events where maximum ow rate values exceeded 0.8 m3s simulation results is unable to achieve values within 30 % of observed values. Also outliers begin to show past 2 m3s with the largest deviation occurring with the 10 m3s observed event. 93 Figure 4.26: Veri cation error analysis for max ow during single aquifer simulation No Aquifer Simulation As expected the no aquifer simulation was unable to predict any peak ows through the entire veri cation period. Results in Figure 4.27 clearly show this as all events fell well below the 1 : 1 linear regression line. Figure 4.27: Veri cation error analysis for max ow during no aquifer simulation 94 Mean Discharge Multiple Aquifer Simulation Mean discharges simulated during the multiple aquifer con guration are shown in Figure 4.28. Each event is highly dependent on the characteristics of the rainfall runo event of interest. Rain events with high intensities may have short duration but large ows thus causing increases in the mean discharge value. It can be seen that computed mean ows tended towards over prediction through the lower portion of ows. Very few events fell within the 30 % envelope from the regression line especially for small mean values. Then during larger computed mean values a generally under prediction trend is seen through the veri cation events. Figure 4.28: Veri cation error analysis for mean ow during multiple aquifer simulation Single Aquifer Simulation The single aquifer con guration has shown computed results with better agreement to observed values for low mean ows. Figure 4.29 shows that once mean ow values reached about 0.2 m3s computed values began to under predicted observed values. After this threshold the model has under predicted mean ows by as much as 46 %. 95 Figure 4.29: Veri cation error analysis for mean ow during single aquifer simulation No Aquifer Simulation Mean ow values were simulated very poorly with the no aquifer con guration. Figure 4.30 displays this clearly without any data points nearing the regression line. Figure 4.30: Veri cation error analysis for mean ow during no aquifer simulation 96 Total Volumes Multiple Aquifer Simulation Total volume analysis for the multiple aquifer con guration is shown in Figure 4.31. Computed total volumes have been predicted moderately well during this simulation. With volumes less then 8000 m3 data points show good agreement between computed and ob- served with some over-estimation tendency. Larger volumes also showed good agreement. The largest observed total volume event producing approximately 20000 m3 was only un- der predicted by approximately 30 %. Which for this study was considered an acceptable predictability range. Figure 4.31: Veri cation error analysis for total ow during multiple aquifer simulation Single Aquifer Simulation The single aquifer con guration has shown similar signs to the multiple aquifer con g- uration during total volume analysis. Some signi cant di erences occurred after the 4000 m3 mark of observed volumes. At this point simulation output was under predicting the observed eld data by more then 30 %. The large event of approximately 20000 m3 has been under predicted by much less then the acceptable 30 % envelope. 97 Figure 4.32: Veri cation error analysis for total ow during single aquifer simulation No Aquifer Simulation As expected from this no aquifer con guration model outputs have produced dramati- cally lower total ow volumes. Figure 4.33 displays this lack of modeling capability clearly. All data points were close to x-axis indicating poor modeling results. Figure 4.33: Veri cation error analysis for total ow during no aquifer simulation 98 4.2.4 Statistical Summary Following recommendations provided by Moriasi et al. (2007), Krause et al. (2005) and Legates and McCabe (1999), not only should graphical methods be used to evaluate a models performance but also a compilation of statistical parameters. More speci cally Legates and McCabe (1999) suggest that at least one dimensionless statistics and one absolute error index statistic should be used in the analysis. For this study the Nash-Sutcli e e ciency (NSE) and root mean square error (RMSE) provided the latter respectively and also included is the coe cient of determination (R2). Each of the previous were determined between the predicted (P) and observed runo values (O) over the entire calibration and veri cation periods. Methods used by PCSWMM to determine each of these statistics are seen in Equations 4.1-4.3. (4.1) (4.2) (4.3) Performance ratings, Table 4.12, were also assigned to NSE values based upon a col- laboration of studies organized by Moriasi et al. (2007). These rating categories were based upon models run at the monthly time scale but are still assumed to be valid in this sub-hourly simulation. 99 Table 4.12: Performance ratings for NSE, (Moriasi et al., 2007) Examining Table 4.13 for calibration statistics it can be seen that out of the three con gurations created the no aquifer con guration performed the poorest with respect to R2 values. During maximum and mean ow results, R2 values have fallen below 0.5 or the minimum value that is considered a satisfactory result, (Moriasi et al., 2007). However, during the total volume the no aquifer con guration did produce a R2 value of 0.9. This is misleading as it would suggest a good agreement between modeled and observed runo which was not the case. Now examining the NSE values of the no aquifer con guration, prior observations of a poor modeling e ort based on R2 values is reinforced by NSE values. With ranges between 1and 1.0, where 1 is a optimal value, 0 is a neutral value and 1is the worst case, this con guration hasn?t produced values above 0.07 in all categories. TheseNSE values indicate a unsatisfactory simulation result even where the R2 value of total volume indicated a good match. Finally, when examining RMSE values for this con guration only the mean ow has resulted in a value close to 0. According with literature presented by Moriasi et al. (2007), values 0 indicate a perfect t between modeled and computed data sets. Based on the prior statistics described and present in Table 4.13 the no aquifer con guration has produced unacceptable results during the calibration period. 100 Table 4.13: Calibration error analysis for all simulations performed Continuing to analyze Table 4.13 it can be seen that the multiple aquifer con gura- tion has exceeded over the single con guration values of R2 in two of the three categories. Though both con gurations produced R2 above the acceptable 0.5 provided from Moriasi et al. (2007), resulting in each having satisfactory error variance. To distinguish these two con gurations further the NSE values must be inspected. NSE values for each of these con gurations are relatively close in each category. However, the multiple aquifer model has produced satisfactory results in the mean ow and total volume categories. This fact pro- vides evidence toward the multiple con guration performing the best for all three categories with the exception of an unsatisfactory NSE value during the maximum ow comparison. Veri cation errors are presented in Table 4.14 for each model con guration and calcu- lated over the entire duration of this period. Once again the no aquifer con guration has proved to be the worst in each statistic over every category. It has not produced R2 values above 0.05 and NSE values followed similar trends as not one value was above 0. This outcome was expected as these results were the main reason for the addition of a aquifer compartment of SWMM. Disregarding the no aquifer con guration, details on multiple and single con gurations now come into light. During maximum ow comparisons both con gurations performed poorly. Each did not achieve an R2 value greater then 0.5 and NSE values were close to zero or below. Mean ow comparisons provided better results as the multiple aquifer con guration displayed a R2 value just above 0.5, a NSE value of 0.45 and a RMSE of 0.45. The single aquifer con guration however under-performed in this category producing results in each statistic below recommendations and the multiple aquifer con guration. Lastly, looking into 101 the total volume category it is seen that both the single and multiple aquifer con gurations produced R2 values above 0.5. The multiple aquifer con guration has produced the higher value once again in this category. Using NSE values to distinguish these two further it is seen that the multiple aquifer con guration has produced an NSE of 0.68 which has fallen into the "Good" performance rating. Table 4.14: Veri cation error analysis for all simulations performed After examining each con gurations outcomes with respect to the statistics described above some conclusions can be made about the best simulation results. During calibration e ort the single and multiple aquifer con gurations provided the best results in the maximum ow, mean ow and total volume categories which were very comparable. Both con gurations provided statistics in the satisfactory to good range with respect to NSE values. R2 values followed the same path as each con guration had values above 0.5. However, the multiple aquifer con guration was able to simulate all three categories with the best performance during calibration e orts. Continuing to the veri cation period results from the single and multiple aquifer con gurations both performed poorly during the maximum ow category. The mean ow produced similarly poor results for the single aquifer con guration as it was unable to meet minimum R2 and NSE values. On the other hand the multiple aquifer con guration met minimum R2 requirements but fell just short of a satisfactory NSE value while exceeding the single aquifer con guration performance. Finally, investigating the total volume category of Table 4.14 the multiple aquifer con guration surpasses in each statistic once again. Combining all of the prior examinations the multiple aquifer con guration has been most consistent in modeling WS-AGC. In summary this SWMM con guration models peak ows poorly but is able to simulate total volumes fairly well. 102 Chapter 5 Summary of Findings and Suggestions for Future Studies 5.1 Summary The investigations involved the development of eld and numerical studies to represent hydrological processes at WS-AGC. During this study the rst objective of developing a eld monitoring program was met and exceeded. Now within WS-AGC there are three rain gauges, two weirs, four groundwater wells and a Kestrel 4500 weather meter. As construction e orts began each piece of equipment was set online at di erent time frames due to delays in weather and construction time. The rst and longest recording devices installed at WS-AGC were three rain gauges. They have recorded local rainfall since early February 2012 producing a short, year and 6 months, duration of data. Even though data from the eld monitoring program has only been collected for this short duration, insights into many dynamic aspects of the local hydrological cycle have been observed. Early eld data collected presented many interesting and signi cant ndings related to the hydrology of WS-AGC. Rainfall runo events have shown evidences of an alternat- ing transition between losing/gaining stream ow regimes. As wet and dry periods occur over the year this phenomena was seen in recorded hydrographs. Wet period runo events showed large increases in recession times while dry period runo events typically displayed short recession times. When comparing similar size rainfall events from dry to wet periods, dramatic di erences were seen from one season to the other. In forested shallow soil watersheds, the vertical hydraulic conductivity of the soil is often high, resulting in rapid conduction of in ltrated water from the near surface of the sub-catchment through the soil matrix to the subsurface boundary below (Axworthy and Karney, 1999). Fast rates of in ltration along with a low permeable subsurface layer are 103 thought to have caused increased groundwater elevations. These increases in groundwater elevations were mainly seen during the wet period and are hypothesized to have supported extended contributions of ow. On the contrary, dry period events seen during 2012 displayed large peak ows but without the extended recession curve. The groundwater levels were unfortunately not monitored during this time frame but the current hypothesis is that low or none existing levels allowed for excessive in ltration and minimal runo . Despite the lack of groundwater monitoring, it is also speculated that during this time there existed a disconnection between groundwater and channel water elevations. Another interesting phenomena seen in the shallow groundwater wells was evapotranspi- ration (ET). During the largest di erences in water table elevations observed, an amplitude of 2.8 cm was measured over a 12 hour period. Declines and ascents throughout the record followed speci c diurnal patterns. Lowest levels were seen during mid-day to mid-afternoon hours and recovering water table elevations occurred during late-afternoon to late-night hours. Similar observations were seen in undeveloped forested watersheds during studies performed by (Czikowsky and Fitzjarrald, 2004; Gribovszki et al., 2008, 2010). As discov- ered from Czikowsky and Fitzjarrald (2004), Gribovszki et al. (2008) and Gribovszki et al. (2010) , ET is responsible for large amounts of losses experienced in the natural watershed. This component of the hydrology must continue to be analyzed in order to better understand its role in the WS-AGC water budget. Field data including rainfall, runo , atmospheric pressure and temperature recorded from the 15 month study period were incorporated in a SWMM watershed model. Focus during model development was on creating a modeling con guration that represented the local physical features as well as model parameters with reasonable and defensible values. This was achieved with the use of GIS, eld surveys, literature studies and engineering judgment. Through a sound approach during modeling development, this study added a realistic assessment of the model?s replication abilities. During calibration e orts the multiple aquifer, 104 single aquifer and no aquifer con gurations were analyzed and compared using graphical and statistical methods. Out of the three con gurations the multiple aquifer proved to be the leader in modeling mean ow and total volumes. Statistical values for the multiple aquifer con guration were (R2=0.94 and NSE=0.53) and (R2=0.98 and NSE=0.69) for both mean ow and total volumes respectively. Maximum ows were not simulated as well with statistical values of (R2=0.67 and NSE=0.29) produced from the multiple aquifer con guration. During the veri cation stage of the model, results from the multiple aquifer con gura- tion showed the best correlation with observed eld data. Maximum ow conditions were poorly simulated with R2 and NSE values well below minimum requirements. When sim- ulating mean ows the multiple aquifer con guration met minimum R2 requirements but fell just short of a satisfactory NSE value, (R2=0.51 and NSE=0.34). However, the multi- ple aquifer con guration exceeded in total volumes simulated. This con guration met and exceed minimum requirements for both statistical values, (R2=0.75 and NSE=0.69). Overall, SWMM performed well with regards to mean and total ows produced from the multiple and single aquifer con gurations. It however was unable to capture peak dis- charge events observed in the eld with any reliable accuracy. Depending on what type of ow category that the user is interested in, SWMM could either constitute as a great tool or a misleading one for pre-development watershed studies. Since most pre-development studies do not have the opportunity to perform collection of localized hydrological data, accuracy of outputs from these constructed model may never truly be known. If peak discharges are of concern during any undeveloped study which does not have observed data, then SWMM is likely to be a ill-advised alternative. However, during this study total ow volumes were of key interest and such model outputs were deemed satisfactory. The model created for WS- AGC will ultimately be applied to simulate the feasibility of a pond to be sustained. In ow volumes for the reservoir, representative of a small pond, will be critical in determining its capabilities of supporting water. 105 5.2 Suggested Future Studies Work completed has provided an excellent foundation to continue studying WS-AGC?s hydrology. One of most important items to be continued is eld data collection. Data should be continually collected and installed equipment maintained in order to uphold measurement integrity. This includes maintenance and collection from weirs, rain gauges, groundwater wells and meteorological station. Long term hydrological data will provided more con dence in judgments about the interactions between intrinsic processes at WS-AGC. SWMM development must also continue as new insights have been produced from soil studies and shallow groundwater well data. Using the four monitoring wells now existing at WS-AGC the groundwater dynamics must be examined in more detail. During soil boring operations, material below the litterfall layer has shown very low permeability of approxi- mately 2:83x10 8 cms . These developments could assist in a new approach for implementing the aquifer compartment in SWMM. During this study data was insu cient from the rectangular weir to attempt calibration and veri cation processes. New model developments should also focus on data from both rectangular and broad crested weirs. This will ensure that upper portions and lower portions of WS-AGC are simulated accurately through numerical models such as SWMM. 106 Bibliography Akan, O. A. (1993). Urban Stormwater Hydrology: A Guide to Engineering Calculations. Lancaster: Technomic Publishing Co., Inc. Amatya, D., Harrison, C., and Trettin, C. (2007). Water quality of two rst order forested watersheds in coastal south carolina. Notes. Axworthy, D. H. and Karney, B. W. (1999). Modeling surface and subsurface runo in a forested watershed. Journal of Hydrologic Engineering, 4(2):165{173. Borah, D. K., Weist, J. H., Wall, J. D., Powell, D. N., et al. (2009). Watershed models for storm water management: Comparing hydrologic and hydraulic procedures. ASCE. Bormann, H. (2011). Sensitivity analysis of 18 di erent potential evapotranspiration models to observed climatic change at german climate stations. Climatic change, 104(3-4):729{ 753. Brater, E. F., King, H. W., and james E. Lindell (1996). Handbook of hydraulics. McGraw- Hill New York. Czikowsky, M. J. and Fitzjarrald, D. R. (2004). Evidence of seasonal changes in evapotran- spiration in eastern us hydrological records. Journal of Hydrometeorology, 5(5):974{988. Davis, J., Rohrer, C., and Roesner, L. (2007). Calibration of rural watershed models in the north carolina piedmont ecoregion. In Proceedings of the 2007 World Environmental and Water Resources Congress: Restoring Our Natural Habitat. Reston: ASCE, pages 1{10. Dodge, R. (2001). Water measurement manual: A guide to e ective water measurement practices for better water management. Government Printing O ce. 107 Federer, C. A. and Lash, D. (1978). Brook: A hydrologic simulation model for eastern forests. Gribovszki, Z., Kalicz, P., Szil agyi, J., and Kucsara, M. (2008). Riparian zone evapotran- spiration estimation from diurnal groundwater level uctuations. Journal of Hydrology, 349(1):6{17. Gribovszki, Z., Szil agyi, J., and Kalicz, P. (2010). Diurnal uctuations in shallow ground- water levels and stream ow rates and their interpretation{a review. Journal of Hydrology, 385(1):371{383. Hamon, W. R. (1963). Computation of direct runo amounts from storm rainfall c1). Harder, S. V., Amatya, D. M., Callahan, T. J., Trettin, C. C., and Hakkila, J. (2007). Hy- drology and water budget for a forested atlantic coastal plain watershed, south carolina1. JAWRA Journal of the American Water Resources Association, 43(3):563{575. Hargreaves, G. H. and Allen, R. G. (2003). History and evaluation of hargreaves evapotran- spiration equation. Journal of Irrigation and Drainage Engineering, 129(1):53{63. James, W. R. C., Wan, B., and James, W. (2002). Implementation in pcswmm using genetic algorithms for auto calibration and design-optimization. In Proc. 9th Int. Conf. Urban Drainage{Global Solutions for Urban Drainage. Jang, S., Cho, M., Yoon, J., Yoon, Y., Kim, S., Kim, G., Kim, L., and Aksoy, H. (2007). Using swmm as a tool for hydrologic impact assessment. Desalination, 212(1):344{356. Krause, P., Boyle, D., and B ase, F. (2005). Comparison of di erent e ciency criteria for hydrological model assessment. Advances in Geosciences, 5(5):89{97. La Torre Torres, I. B., Amatya, D. M., Sun, G., and Callahan, T. J. (2011). Seasonal rainfall{ runo relationships in a lowland forested watershed in the southeastern usa. Hydrological Processes, 25(13):2032{2045. 108 Legates, D. R. and McCabe, G. J. (1999). Evaluating the use of goodness-of- t measures in hydrologic and hydroclimatic model validation. Water Resources Research, 35(1):233{241. Lu, J., Sun, G., McNulty, S. G., and Amatya, D. M. (2005). A comparison of six potential evapotranspiration methods for regional use in the southeastern united states1. JAWRA Journal of the American Water Resources Association, 41(3):621{633. Makkink, G. (1957). Testing the penman formula by means of lysimeters. J. Inst. Water Eng, 11(3):277{288. McCuen, R. H., Johnson, P. A., and Ragan, R. M. ((1996)). Highway Hydrology Hydraulic Design Series, No.2. Rep. no. FHWA-SA-067. Moriasi, D., Arnold, J., Van Liew, M., Bingner, R., Harmel, R., and Veith, T. (2007). Model evaluation guidelines for systematic quanti cation of accuracy in watershed simulations. Transactions of the ASABE, 50(3):885{900. Onset (2011). Data Logging Rain Gauge RG3 and RG3-M User?s Manual. Onset Computer Corportaion. Priestley, C. and Taylor, R. (1972). On the assessment of surface heat ux and evaporation using large-scale parameters. Monthly weather review, 100(2):81{92. Rossman, L. A. and Supply, W. (2005). Storm water management model user?s manual, version 5.0. National Risk Management Research Laboratory, O ce of Research and Development, US Environmental Protection Agency. Sun, G., McNulty, S., Amatya, D., Skaggs, R., Swift Jr, L., Shepard, J., and Riekerk, H. (2002). A comparison of the watershed hydrology of coastal forested wetlands and the mountainous uplands in the southern us. Journal of Hydrology, 263(1):92{104. 109 Sun, G., Noormets, A., Gavazzi, M., McNulty, S., Chen, J., Domec, J.-C., King, J., Amatya, D., and Skaggs, R. (2010). Energy and water balance of two contrasting loblolly pine plan- tations on the lower coastal plain of north carolina, usa. Forest Ecology and Management, 259(7):1299{1310. Thornthwaite, C. and Mather, J. (1955). The water balance. Climatology, 8:1{104. Thornthwaite, C. and Mather, J. (1957). Instructions and tables for computing potential evapotranspiration and the water balance. Climatology, 10. Thornthwaite, C. W. (1948). An approach toward a rational classi cation of climate. Geo- graphical review, 38(1):55{94. Trajkovic, S. (2005). Temperature-based approaches for estimating reference evapotranspi- ration. Journal of irrigation and drainage engineering, 131(4):316{323. Turc, L. (1961). Evaluation des besoins en eau dirrigation, evapotranspiration potentielle. Ann. agron, 12(1):13{49. Viessman, W., Lewis, G. L., Knapp, J. W., and Harbaugh, T. E. (2003). Introduction to hydrology. Prentice Hall NJ. V or osmarty, C. J., Federer, C. A., and Schloss, A. L. (1998). Potential evaporation func- tions compared on us watersheds: Possible implications for global-scale water balance and terrestrial ecosystem modeling. Journal of Hydrology, 207(3):147{169. Wear, D. N. and Greis, J. G. (2002). Southern forest resource assessment: summary of ndings. Journal of Forestry, 100(7):6{14. Winter, T. C. (2007). The role of ground water in generating stream ow in headwater areas and in maintaining base ow1. JAWRA Journal of the American Water Resources Association, 43(1):15{25. 110 Appendices 111 Appendix A Soil Analysis 112 Terracon Consultants, Inc. 110 12 TH Street North Birmingham, Alabama 35203 P [205] 942-1289 F [205] 443- 5302 terracon.com September 4, 2013 Alabama AGC 5000 Grantswood Road Irondale, Alabama 35210 Attn: Mr. Jeff Rogers E: jeffr@alagc.org Re: Geotechnical Engineering Report AGC - Auburn University Dam Site Russell County, Alabama Terracon Project No. E1135088 Dear Mr. Rogers: Terracon has completed the geotechnical engineering services for the above referenced project. This study was performed in general accordance with our Proposal PE1130324, dated May 15, 2013. Subsurface conditions encountered at each boring location are indicated on the accompanying individual boring logs. The approximate location of each boring is indicated on the accompanying Figure A-2, Boring Location Plan. Selected samples were tested in our laboratories to determine physical engineering characteristics of the onsite soils. These tests included: Atterberg limits, grain-size analyses, moisture contents, standard Proctors, Triaxial Shear and permeability tests. The results of the laboratory analysis are included in Appendix B. We appreciate the opportunity to be of service to you on this project. If you have any questions concerning this report, or if we may be of further service, please contact us. Sincerely, Terracon Consultants, Inc. Charlie L. Bragg Jerome A. Smith, P.E Field Project Manager Manager, Geotechnical Services Alabama P.E. No. 20478 APPENDIX A ? FIELD EXPLORATION Exhibit A-1 Site Location Map Exhibit A-2 Boring Location Plan Exhibit A-3 Field Exploration Description Exhibits A-4 to A-9 Boring Logs, Borings B-1 through B-6 APPENDIX B ? LABORATORY TESTING Exhibit B-1 Laboratory Testing APPENDIX C ? SUPPORTING DOCUMENTS Exhibit C-1 General Notes Exhibit C-2 Unified Soil Classification System APPENDIX A FIELD EXPLORATION 110 12th Street North Birmingham, Alabama 35203 PH. (205) 942-1289 FAX. (205) 443-5302 CLB JAS JAS A-1 Exhibit SITE LOCATION MAP Project Manager: Drawn By: Checked By: Approved By: NTS CLB Proposal No. Scale: File Name: Date: E1135088 E1135088.1.pdf 07-10-2013 AGC ? AUBURN UNIVERSITY DAM SITE RUSSELL COUNTY, ALABAMA PROJECT SITE 110 12th Street North Birmingham, Alabama 35203 PH. (205) 942-1289 FAX. (205) 443-5302 CLB JAS JAS A-2 Exhibit BORING LOCATION PLAN Project Manager: Drawn By: Checked By: Approved By: As Shown CLB Proposal No. Scale: File Name: Date: E1135088 E1135088.2.pdf 07-10-2013 AGC ? AUBURN UNIVERSITY DAM SITE RUSSELL COUNTY, ALABAMA BORING LOCATION DIAGRAM IS FOR GENERAL LOCATION ONLY, AND IS NOT INTENDED FOR CONSTRUCTION PURPOSES DIAGRAM PROVIDED BY KYLE MOYNIHAN Geotechnical Engineering Report AGC- Auburn University Dam Site ? Russell County, Alabama September 4, 2013 ? Terracon Project No. E1135088 Exhibit A-3 Field Exploration Description A total of six (6) test borings were performed across the site. These borings were extended to a depth of approximately 20 feet below the existing surface grade. The boring locations were marked in the field by measuring from the dam abutment stakes set by Auburn University representatives. The location of each boring was then recorded by Auburn University representatives utilizing a GPS receiver and plotted on the provided Boring Location Plan included in Appendix A. The borings were drilled with an ATV-mounted CME-550 rotary drill rig with an automatic hammer using hollow-stem augers and rock coring equipment to advance the borehole. Samples of the soil encountered in the boring were obtained using the split-barrel sampling procedures (general accordance with ASTM D1586. In the split-barrel sampling procedure, the number of blows required to advance a standard 2-inch O.D. split- barrel sampler the last 12 inches of the typical total 18-inch penetration by means of a 140-pound hammer with a free fall of 30 inches, is the Standard Penetration Test N-value (SPT-N). This value is used to estimate the in situ relative density of cohesionless soils and consistency of cohesive soils. Following the completion of the SPT borings, boring B-2 was offset minimally and the drill rigs were utilized to push a total of three (3) thin-walled Shelby tubes. The Shelby tubes recovered relatively undisturbed samples of the in-situ soils. Additionally, two (2) bulk samples were collected at boring location B-1, at depths of 0 -5 feet and 5-10 feet. The soil samples were placed in containers to reduce moisture loss, tagged for identification, and taken to our laboratory (general accordance with ASTM D4220) for further examination, testing, and classification. Information provided on the boring logs attached to this report includes soil descriptions, consistency evaluations, boring depths, sampling intervals, and groundwater conditions. A field log of the boring was prepared by the Terracon engineer. The log included visual classifications (general accordance with ASTM D5434) of the materials encountered during drilling as well as the engineer?s interpretation of the subsurface conditions between samples. Final boring log included with this report represent the engineer's interpretation of the field log and include modifications based on laboratory observation and tests of the samples. 5.0 18.520.0 SANDY CLAY (CL), brown and yellowish red mottled, softbecomes medium stiff FAT CLAY (CH), trace fine sand, light gray and brownish yellow mottled, stiff,micaceous SANDY CLAY (CL), dark gray, hard, micaceousBoring Terminated at 20 Fet 67882043 32 1-2N=33-3 N=62-4-6N=10 3-6-7N=13 4-5-8N=13 7-13-19N=32 45-25-2059-30-29 See Exhibit A-2GRAPHIC LOG Stratification lines are approximate. In-situ, the transition may be gradual. LOCATIONDEPTH THIS BORING LOG IS NOT VA LID IF SEPARATED FROM ORIGINAL RE PORT. Aubrn Uiversity , AlabmaSITE: No fre water observed during boringWATER LEVL OBSERVATIONS PROJECT: AGC Aubrn Uiversity DamSite Page 1 of 1 Advancement Method:Holow-stem augerAbandonment Method:Borigs backfiled with soil cuttings upon completion. 110 12th Street NorthBirmingham, Alabama Notes: Project No.: E1135088Dril Rig: CME-550Boring Started: 7/10/2013 BORING LOG NO. B-1Alabma AGCCLIENT:Irondle, labma Driler: B.C.Boring Completed: 7/10/2013A-4Exhibit: See Exhibit A-3 for description of fieldprocedures.See Appendix B for description of laboratoryprocedures ad additional data (if any).See Appendix C for explanation of symbols andabbreviations. PERCENT FINESWATER CONTENT (%)FIELD TEST RESULTSDEPTH (Ft.) 5 10 15 20 SAMPLE TYPEWATER LEVEL OBSERVATIONS ATERBEGLIMITSLL-PL-PI 4.0 8.0 13.5 20.0 SANDY CLAY (CL), brown and yellowish red mottled, medium stiff FAT CLAY (CH), trace fine sand, light gray and brownish yellow mottled, soft SILTY SAND (SM), brown and gray, loose SANDY CLAY (CL), dark gray, very stiff, micaceous becomes hardBoring Terminated at 20 Fet 3637 29 2-3N=52-3 N=51-3N=4 1-3-5N=8 4-10-14N=24 6-13-22N=35 See Exhibit A-2GRAPHIC LOG Stratification lines are approximate. In-situ, the transition may be gradual. LOCATIONDEPTH THIS BORING LOG IS NOT VA LID IF SEPARATED FROM ORIGINAL RE PORT. Aubrn Uiversity , AlabmaSITE: Water observed at 9 fet during boringATER LEVL OBSERVATIONS PROJECT: AGC Aubrn Uiversity DamSite Page 1 of 1 Advancement Method:Holow-stem augerAbandonment Method:Borigs backfiled with soil cuttings upon completion. 110 12th Street NorthBirmingham, Alabama Notes: Project No.: E1135088Dril Rig: CME-550Boring Started: 7/10/2013 BORING LOG NO. B-2Alabma AGCCLIENT:Irondle, labma Driler: B.C.Boring Completed: 7/10/2013A-5Exhibit: See Exhibit A-3 for description of fieldprocedures.See Appendix B for description of laboratoryprocedures ad additional data (if any).See Appendix C for explanation of symbols andabbreviations. PERCENT FINESWATER CONTENT (%)FIELD TEST RESULTSDEPTH (Ft.) 5 10 15 20 SAMPLE TYPEWATER LEVEL OBSERVATIONS ATERBEGLIMITSLL-PL-PI 0.5 5.0 8.5 13.5 20.0 6 inches TOPSILSANDY CLAY (CL), brown and yellowish red mottled, stiff SANDY CLAY (CL), light gray and brownish yellow mottled, stiff SANDY SILT (ML), dark gray, hard, micaceous SANDY CLAY (CL), dark gray, very stiff, micaceous becomes hardBoring Terminated at 20 Fet 22 25 25 3-5N=82-3-5 N=83-6-5N=11 17N=50/3" 4-11-15N=26 24-19-27N=46 See Exhibit A-2GRAPHIC LOG Stratification lines are approximate. In-situ, the transition may be gradual. LOCATIONDEPTH THIS BORING LOG IS NOT VA LID IF SEPARATED FROM ORIGINAL RE PORT. Aubrn Uiversity , AlabmaSITE: Water observed at 9 fet during boringATER LEVL OBSERVATIONS PROJECT: AGC Aubrn Uiversity DamSite Page 1 of 1 Advancement Method:Holow-stem augerAbandonment Method:Borigs backfiled with soil cuttings upon completion. 110 12th Street NorthBirmingham, Alabama Notes: Project No.: E1135088Dril Rig: CME-550Boring Started: 7/10/2013 BORING LOG NO. B-3Alabma AGCCLIENT:Irondle, labma Driler: B.C.Boring Completed: 7/10/2013A-6Exhibit: See Exhibit A-3 for description of fieldprocedures.See Appendix B for description of laboratoryprocedures ad additional data (if any).See Appendix C for explanation of symbols andabbreviations. PERCENT FINESWATER CONTENT (%)FIELD TEST RESULTSDEPTH (Ft.) 5 10 15 20 SAMPLE TYPEWATER LEVEL OBSERVATIONS ATERBEGLIMITSLL-PL-PI 0.32.5 8.5 13.5 20.0 4 inches TOPSILSANDY CLAY (CL), yellowish red, very softSANDY CLAY (CL), light gray and yellowish red mottled, medium stiff SILTY SAND (SM), light gray and light brown, medium dense SANDY SILT (ML), dark gray and brown, hard, micaceous Boring Terminated at 20 Fet 21 31 22 WOH3-4N=7 3-4-6N=10 3-5-8N=13 8-13-24N=37 12-17-20N=37 See Exhibit A-2GRAPHIC LOG Stratification lines are approximate. In-situ, the transition may be gradual. LOCATIONDEPTH THIS BORING LOG IS NOT VA LID IF SEPARATED FROM ORIGINAL RE PORT. Aubrn Uiversity , AlabmaSITE: No fre water observed during boringWATER LEVL OBSERVATIONS PROJECT: AGC Aubrn Uiversity DamSite Page 1 of 1 Advancement Method:Holow-stem augerAbandonment Method:Borigs backfiled with soil cuttings upon completion. 110 12th Street NorthBirmingham, Alabama Notes: Project No.: E1135088Dril Rig: CME-550Boring Started: 7/10/2013 BORING LOG NO. B-4Alabma AGCCLIENT:Irondle, labma Driler: B.C.Boring Completed: 7/10/2013A-7Exhibit: See Exhibit A-3 for description of fieldprocedures.See Appendix B for description of laboratoryprocedures ad additional data (if any).See Appendix C for explanation of symbols andabbreviations. PERCENT FINESWATER CONTENT (%)FIELD TEST RESULTSDEPTH (Ft.) 5 10 15 20 SAMPLE TYPEWATER LEVEL OBSERVATIONS ATERBEGLIMITSLL-PL-PI 2.5 5.0 8.5 13.5 20.0 SANDY CLAY (CL), brown and yellowish red, mottled, stiffCLAY (CL), with fine sand, gray, brown and yellowish red mottled, medium stiff SANDY CLAY (CL), brown, stiff SILTY SAND (SM), brown and gray, loose SANDY CLAY (CL), dark gray, very stiff, micaceous Boring Terminated at 20 Fet 4223 31 3-4-5N=92-3-3 N=63-5-5N=10 2-3-3N=6 5-10-14N=24 6-11-15N=26 See Exhibit A-2GRAPHIC LOG Stratification lines are approximate. In-situ, the transition may be gradual. LOCATIONDEPTH THIS BORING LOG IS NOT VA LID IF SEPARATED FROM ORIGINAL RE PORT. Aubrn Uiversity , AlabmaSITE: Water observed at 8 fet during boringATER LEVL OBSERVATIONS PROJECT: AGC Aubrn Uiversity DamSite Page 1 of 1 Advancement Method:Holow-stem augerAbandonment Method:Borigs backfiled with soil cuttings upon completion. 110 12th Street NorthBirmingham, Alabama Notes: Project No.: E1135088Dril Rig: CME-550Boring Started: 7/10/2013 BORING LOG NO. B-5Alabma AGCCLIENT:Irondle, labma Driler: B.C.Boring Completed: 7/10/2013A-8Exhibit: See Exhibit A-3 for description of fieldprocedures.See Appendix B for description of laboratoryprocedures ad additional data (if any).See Appendix C for explanation of symbols andabbreviations. PERCENT FINESWATER CONTENT (%)FIELD TEST RESULTSDEPTH (Ft.) 5 10 15 20 SAMPLE TYPEWATER LEVEL OBSERVATIONS ATERBEGLIMITSLL-PL-PI 0.3 5.0 8.5 13.5 18.520.0 TOPSILSANDY CLAY (CL), brown and yellowish red mottled, medium stiff CLAYE SAND (SC), brown, loose SILTY SAND (SM), gray, very loose SANDY CLAY (CL), dark gray, hard SANDY SILT (ML), dary gray, hardBoring Terminated at 20 Fet 38 21 21 2-3-3N=63-4 N=73-4-3N=7 0-3N= 15-16-20N=36 N=50+ See Exhibit A-2GRAPHIC LOG Stratification lines are approximate. In-situ, the transition may be gradual. LOCATIONDEPTH THIS BORING LOG IS NOT VA LID IF SEPARATED FROM ORIGINAL RE PORT. Aubrn Uiversity , AlabmaSITE: Water observed at 10 fet during boringATER LEVL OBSERVATIONS PROJECT: AGC Aubrn Uiversity DamSite Page 1 of 1 Advancement Method:Holow-stem augerAbandonment Method:Borigs backfiled with soil cuttings upon completion. 110 12th Street NorthBirmingham, Alabama Notes: Project No.: E1135088Dril Rig: CME-550Boring Started: 7/10/2013 BORING LOG NO. B-6Alabma AGCCLIENT:Irondle, labma Driler: B.C.Boring Completed: 7/10/2013A-9Exhibit: See Exhibit A-3 for description of fieldprocedures.See Appendix B for description of laboratoryprocedures ad additional data (if any).See Appendix C for explanation of symbols andabbreviations. PERCENT FINESWATER CONTENT (%)FIELD TEST RESULTSDEPTH (Ft.) 5 10 15 20 SAMPLE TYPEWATER LEVEL OBSERVATIONS ATERBEGLIMITSLL-PL-PI APPENDIX B LABORATORY TESTING Geotechnical Engineering Report AGC- Auburn University Dam Site ? Russell County, Alabama September 4, 2013 ? Terracon Project No. E1135088 Exhibit B-1 Laboratory Testing Selected soil samples were tested for properties such as Atterberg limits, grain-size analyses, moisture contents, standard Proctors, Triaxial Shear and permeability tests. The results of the laboratory analysis are included in Appendix B and/or on the boring logs included in Appendix A. Descriptive classifications of the soils indicated on the boring logs are in accordance with the enclosed General Notes and the Unified Soil Classification System. Also shown are estimated Unified Soil Classification Symbols. A brief description of this classification system is attached to this report. All classification was by visual manual procedures. 0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 100 0.0010.010.1110100 HYDROMETERU.S. SIEVE OPENING IN INCHES U.S. SIEVE NUMBERS 44 10063 2 fine coarse SOIL DESCRIPTION CU BORING ID 10 14 506 2001.5 8 % FINES % CLAY USCS B2 0.0 0.0 72.3 DEPTH 0.078 0.132 8 - 10 GRAIN SIZE Brownish Gray Silty Sand with Clay 16 20 100 90 80 70 60 50 40 30 20 10 0 REMARKS SILT OR CLAYCOBBLES GRAVEL SANDmedium 27.7 GRAIN SIZE IN MILLIMETERS PERCENT FINER 3/4 1/2 3/8 SIEVE (size) D60 30 403 601 140 coarse fine COEFFICIENTS % COBBLES % GRAVEL % SAND D30 D10 CC PE RC EN T F IN ER B Y W EI GH T PE RC EN T C OA RS ER B Y W EI GH T % SILT 1 1/2" 1" 3/4" 1/2" 3/8" #4 #10 #20 #40 #60 #100 #200 100.0 99.96 99.87 98.92 93.58 67.06 27.66 GRAIN SIZE DISTRIBUTION ASTM D422 51 Lost Mound Drive, Suite 135 Chattanooga, Tennessee PROJECT NUMBER: E1135088PROJECT: AGC Dam Site SITE: AGC Dam CLIENT: LA BO RA TO RY TE ST S AR E NO T V AL ID IF S EP AR AT ED FR OM O RI GI NA L R EP OR T. G RA IN S IZE : U SC S 1 E1 13 50 88 A GC D AM S ITE .G PJ T ER RA CO N2 01 2.G DT 8 /21 /13 0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 100 0.0010.010.1110100 HYDROMETERU.S. SIEVE OPENING IN INCHES U.S. SIEVE NUMBERS 44 10063 2 fine coarse SOIL DESCRIPTION CU BORING ID 10 14 506 2001.5 8 % FINES % CLAY USCS B2 0.0 0.0 37.0 DEPTH 13 - 15 GRAIN SIZE Greenish Gray Silty Sandy Clay 16 20 100 90 80 70 60 50 40 30 20 10 0 REMARKS SILT OR CLAYCOBBLES GRAVEL SANDmedium 63.0 GRAIN SIZE IN MILLIMETERS PERCENT FINER 3/4 1/2 3/8 SIEVE (size) D60 30 403 601 140 coarse fine COEFFICIENTS % COBBLES % GRAVEL % SAND D30 D10 CC PE RC EN T F IN ER B Y W EI GH T PE RC EN T C OA RS ER B Y W EI GH T % SILT 1 1/2" 1" 3/4" 1/2" 3/8" #4 #10 #20 #40 #60 #100 #200 100.0 99.97 99.89 99.72 99.51 98.58 92.6 62.99 GRAIN SIZE DISTRIBUTION ASTM D422 51 Lost Mound Drive, Suite 135 Chattanooga, Tennessee PROJECT NUMBER: E1135088PROJECT: AGC Dam Site SITE: AGC Dam CLIENT: LA BO RA TO RY TE ST S AR E NO T V AL ID IF S EP AR AT ED FR OM O RI GI NA L R EP OR T. G RA IN S IZE : U SC S 1 E1 13 50 88 A GC D AM S ITE .G PJ T ER RA CO N2 01 2.G DT 8 /21 /13 Project : Date: P-1 Project No. : Boring No.: ap = 0.031416 cm2 Equilibrium 1.6 cm3 Sample: aa = 0.767120 cm2 Pipet Rp 16.8 cm3 Depth (ft): M1 = 0.030180 C = 0.0017912 Annulus Ra 1.0 cm3 Other Location: M2 = 1.040953 T = 0.0658646 Material Description : SAMPLE DATA Wet Wt. sample + ring or tare : 124.70 g Tare or ring Wt. : 0.0 g Before Test After Test Wet Wt: of Sample : 124.70 g Tare No.: 348 Tare No.: 123 Diameter : 1.37 in 3.48 cm2 Wet Wt.+tare: 127.67 Wet Wt.+tare: 132.69 Length : 2.80 in 7.11 cm Dry Wt.+tare: 100.31 Dry Wt.+tare: 106.08 Area: 1.47 in^2 9.51 cm2 Tare Wt: 21.00 Tare Wt: 31.26 Volume : 4.13 in^3 67.64 cm3 Dry Wt.: 79.31 Dry Wt.: 74.82 Unit Wt.(wet): 115.04 pcf 1.84 g/cm^3 Water Wt.: 27.36 Water Wt.: 26.61 Unit Wt.(dry): 85.54 pcf 1.37 g/cm^3 % moist.: 34.5 % moist.: 35.6 2.70 OMC = % of max = +/- OMC = Calculated % saturation: 98.93 Void ratio (e) = 0.97 Porosity (n)= 0.49 55.00 50.00 5.00 psi TEST READINGS 15.8 cm 28.00 Date elapsed t Z 'Zp temp D k k (seconds) (pipet @ t) (cm ) (deg C) (temp corr) (cm/sec) (ft./day) Reset = * 8/20/2013 300 16.4 0.382666 21 0.977 1.49E-07 4.22E-04 8/20/2013 600 16 0.782666 21 0.977 1.54E-07 4.37E-04 8/20/2013 900 15.6 1.182666 21 0.977 1.58E-07 4.47E-04 8/20/2013 1200 15.2 1.582666 21 0.977 1.61E-07 4.55E-04 SUMMARY ka = 1.55E-07 cm/sec Acceptance criteria = 50% ki Vm k1 = 1.49E-07 cm/sec 4.2 % Vm = | ka-ki | x 100 k2 = 1.54E-07 cm/sec 0.6 % ka k3 = 1.58E-07 cm/sec 1.5 % k4 = 1.61E-07 cm/sec 3.3 % Hydraulic conductivity k = 1.55E-07 cm/sec 4.40E-04 ft/day Void Ratio e = 0.97 Porosity n = 0.49 Bulk Density J = 1.84 g/cm3 115.0 pcf Water Content W = 0.47 cm3/cm3 ( at 20 deg C) Intrinsic Permeability kint = 1.59E-12 cm2 ( at 20 deg C) Hydraulic Gradient = 8.0-10.0 Tube Brownish Gray Silty Sand with Clay Assumed Specific Gravity: 8/21/2013 E1135088 Back Pressure (psi) = Confining Pressure = HYDRAULIC CONDUCTIVITY DETERMINATION Panel Number : Permometer Data FLEXIBLE WALL PERMEAMETER - CONSTANT VOLUME (Mercury Permometer Test) Note: The above value is Effective Confining Pressure Z1(Mercury Height Difference @ t1): AGC Dam Site B2 Max Dry Density(pcf) = Set Mercury to Pipet Rp at beginning Test Pressures During Hydraulic Conductivity Test Cell Pressure (psi) = Project : Date: P-1 Project No. : Boring No.: ap = 0.031416 cm2 Equilibrium 1.6 cm3 Sample: aa = 0.767120 cm2 Pipet Rp 13.7 cm3 Depth (ft): M1 = 0.030180 C = 0.0003415 Annulus Ra 1.1 cm3 Other Location: M2 = 1.040953 T = 0.0826999 Material Description : SAMPLE DATA Wet Wt. sample + ring or tare : 429.70 g Tare or ring Wt. : 0.0 g Before Test After Test Wet Wt: of Sample : 429.70 g Tare No.: 207 Tare No.: 248 Diameter : 2.80 in 7.11 cm2 Wet Wt.+tare: 89.45 Wet Wt.+tare: 112.69 Length : 2.23 in 5.66 cm Dry Wt.+tare: 72.77 Dry Wt.+tare: 90.88 Area: 6.16 in^2 39.73 cm2 Tare Wt: 20.87 Tare Wt: 21.05 Volume : 13.73 in^3 225.02 cm3 Dry Wt.: 51.9 Dry Wt.: 69.83 Unit Wt.(wet): 119.16 pcf 1.91 g/cm^3 Water Wt.: 16.68 Water Wt.: 21.81 Unit Wt.(dry): 90.18 pcf 1.45 g/cm^3 % moist.: 32.1 % moist.: 31.2 2.70 OMC = % of max = +/- OMC = Calculated % saturation: 97.02 Void ratio (e) = 0.87 Porosity (n)= 0.47 55.00 50.00 5.00 psi TEST READINGS 12.6 cm 28.00 Date elapsed t Z 'Zp temp D k k (seconds) (pipet @ t) (cm ) (deg C) (temp corr) (cm/sec) (ft./day) Reset = * 8/20/2013 60 9.8 3.891909 21 0.977 2.16E-06 6.12E-03 8/20/2013 120 7.5 6.191909 21 0.977 1.99E-06 5.65E-03 8/20/2013 180 5.8 7.891909 21 0.977 1.96E-06 5.56E-03 8/20/2013 240 4.5 9.191909 21 0.977 1.98E-06 5.63E-03 SUMMARY ka = 2.02E-06 cm/sec Acceptance criteria = 50% ki Vm k1 = 2.16E-06 cm/sec 6.7 % Vm = | ka-ki | x 100 k2 = 1.99E-06 cm/sec 1.5 % ka k3 = 1.96E-06 cm/sec 3.2 % k4 = 1.98E-06 cm/sec 2.0 % Hydraulic conductivity k = 2.02E-06 cm/sec 5.74E-03 ft/day Void Ratio e = 0.87 Porosity n = 0.47 Bulk Density J = 1.91 g/cm3 119.2 pcf Water Content W = 0.47 cm3/cm3 ( at 20 deg C) Intrinsic Permeability kint = 2.07E-11 cm2 ( at 20 deg C) Hydraulic Gradient = 13.0-15.0 Tube Greenish Gray Silty Sandy Clay Assumed Specific Gravity: 8/21/2013 E1135088 Back Pressure (psi) = Confining Pressure = HYDRAULIC CONDUCTIVITY DETERMINATION Panel Number : Permometer Data FLEXIBLE WALL PERMEAMETER - CONSTANT VOLUME (Mercury Permometer Test) Note: The above value is Effective Confining Pressure Z1(Mercury Height Difference @ t1): AGC Dam Site B2 Max Dry Density(pcf) = Set Mercury to Pipet Rp at beginning Test Pressures During Hydraulic Conductivity Test Cell Pressure (psi) = Project : Date: P-1 Project No. : Boring No.: ap = 0.031416 cm2 Equilibrium 1.6 cm3 Sample: aa = 0.767120 cm2 Pipet Rp 16.8 cm3 Depth (ft): M1 = 0.030180 C = 0.0004288 Annulus Ra 1.0 cm3 Other Location: M2 = 1.040953 T = 0.0658646 Material Description : SAMPLE DATA Wet Wt. sample + ring or tare : 543.40 g Tare or ring Wt. : 0.0 g Before Test After Test Wet Wt: of Sample : 543.40 g Tare No.: 123 Tare No.: 318 Diameter : 2.80 in 7.11 cm2 Wet Wt.+tare: 89.45 Wet Wt.+tare: 118.56 Length : 2.80 in 7.11 cm Dry Wt.+tare: 78.90 Dry Wt.+tare: 99.87 Area: 6.16 in^2 39.73 cm2 Tare Wt: 21.55 Tare Wt: 21.26 Volume : 17.24 in^3 282.53 cm3 Dry Wt.: 57.35 Dry Wt.: 78.61 Unit Wt.(wet): 120.02 pcf 1.92 g/cm^3 Water Wt.: 10.55 Water Wt.: 18.69 Unit Wt.(dry): 101.37 pcf 1.62 g/cm^3 % moist.: 18.4 % moist.: 23.8 2.70 106.7 OMC = 16.4 % of max = 95.0 +/- OMC = 2.00 Calculated % saturation: 96.84 Void ratio (e) = 0.66 Porosity (n)= 0.40 55.00 50.00 5.00 psi TEST READINGS 15.8 cm 28.00 Date elapsed t Z 'Zp temp D k k (seconds) (pipet @ t) (cm ) (deg C) (temp corr) (cm/sec) (ft./day) Reset = * 8/20/2013 600 16.2 0.582666 21 0.977 2.73E-08 7.74E-05 8/20/2013 1200 15.6 1.182666 21 0.977 2.83E-08 8.02E-05 8/20/2013 1800 15 1.782666 21 0.977 2.91E-08 8.24E-05 8/20/2013 2400 14.5 2.282666 21 0.977 2.84E-08 8.06E-05 SUMMARY ka = 2.83E-08 cm/sec Acceptance criteria = 50% ki Vm k1 = 2.73E-08 cm/sec 3.4 % Vm = | ka-ki | x 100 k2 = 2.83E-08 cm/sec 0.1 % ka k3 = 2.91E-08 cm/sec 2.8 % k4 = 2.84E-08 cm/sec 0.5 % Hydraulic conductivity k = 2.83E-08 cm/sec 8.02E-05 ft/day Void Ratio e = 0.66 Porosity n = 0.40 Bulk Density J = 1.92 g/cm3 120.0 pcf Water Content W = 0.30 cm3/cm3 ( at 20 deg C) Intrinsic Permeability kint = 2.90E-13 cm2 ( at 20 deg C) Hydraulic Gradient = 0.0-5.0 Remolded Light Brown Sandy Lean Clay Assumed Specific Gravity: 8/21/2013 E1135088 Back Pressure (psi) = Confining Pressure = HYDRAULIC CONDUCTIVITY DETERMINATION Panel Number : Permometer Data FLEXIBLE WALL PERMEAMETER - CONSTANT VOLUME (Mercury Permometer Test) Note: The above value is Effective Confining Pressure Z1(Mercury Height Difference @ t1): AGC Dam Site Bulk Max Dry Density(pcf) = Set Mercury to Pipet Rp at beginning Test Pressures During Hydraulic Conductivity Test Cell Pressure (psi) = I ' = 26.8 deg c' = 2.4 psi 1 2 3 4 18.4 18.4 18.4 101.4 101.4 101.4 2.80 2.80 2.80 5.60 5.60 5.60 101.9 102.9 105.1 2.81 2.80 2.78 5.62 5.59 5.54 10.0 20.0 40.0 15.60 22.79 37.55 55.4 60.5 72.0 0.00060 0.00060 0.00060 1.2 4.6 10.5 20.19 32.28 55.60 4.59 9.49 18.05 LL: 45.4 PL: 25.3 PI: 20.1 SAMPLE LOCATION: Bulk 0.0-5.0 ft Percent -200: 67 TERRACON EFFECTIVE STRESS PARAMETERS SPECIMEN NO. Moisture Content - % INITIAL REMARKS: Specimens remolded to 95% +2 opt. TEST DESCRIPTION TYPE OF TEST & NO: CU with Pore Pressure SAMPLE TYPE: Remolded DESCRIPTION: Light Brown Sandy Lean Clay (CL) ASSUMED SPECIFIC GRAVITY: 2.7 Final Moisture - % Dry Density - pcf Calculated Diameter (in.) AT TEST Dry Density - pcf Diameter - inches Height - inches Strain Rate - inches/min. Failure Strain - % V1' Failure - psi V3' Failure - psi Height - inches Effect. Cell Pressure - psi Failure Stress - psi Total Pore Pressure - psi PROJECT INFORMATION PROJECT: AGC Dam Site LOCATION: AGC Dam PROJECT NO: E1135088 CLIENT: DATE: 8/21/13 0 10 20 30 40 50 0 10 20 30 40 50 60 70 80 SH EA R ST RE SS - PS I PRINCIPAL STRESS - PSI TRIAXIAL SHEAR TEST REPORT 0.00 10.00 20.00 30.00 40.00 50.00 0.0 5.0 10.0 15.0 20.0 DE VI AT OR S TR ES S -P SI AXIAL STRAIN - % TRIAX_Bulk 95 +2.xls R2 = 1.00 D (deg) = 24.3 a (psi) = 2.1EFFECTIVE STRESS PARAMETERS TYPE OF TEST & NO: CU with Pore Pressure TERRACON PROJECT: AGC Dam Site PROJECT NO: E1135088 DESCRIPTION: Light Brown Sandy Lean Clay (CL) 0 10 20 30 40 50 0 10 20 30 40 50 60 70 80 q - ps i p' - psi p - q DIAGRAM 0 10 20 30 40 50 0 5 10 15 20 SPECIMEN NO. 1 Deviator Stress - psi Excess Pore Pressure - psi 0 10 20 30 40 50 0 5 10 15 20 SPECIMEN NO. 2 Deviator Stress - psi Excess Pore Pressure - psi 0 10 20 30 40 50 0 5 10 15 20 SPECIMEN NO. 3 Deviator Stress - psi Excess Pore Pressure - psi 0 10 20 30 40 50 0 5 10 15 20 SPECIMEN NO. 4 Deviator Stress - psi Excess Pore Pressure - psi TRIAX_Bulk 95 +2.xls I = 15.6 deg c = 3.1 psi 1 2 3 4 18.4 18.4 18.4 101.4 101.4 101.4 2.80 2.80 2.80 5.60 5.60 5.60 101.9 102.9 105.1 2.81 2.80 2.78 5.62 5.59 5.54 10.0 20.0 40.0 15.60 22.79 37.55 55.4 60.5 72.0 0.00060 0.00060 0.00060 1.2 4.6 10.5 25.60 42.79 77.55 10.00 20.00 40.00 LL: 45.4 PL: 25.3 PI: 20.1 TERRACON PROJECT INFORMATION PROJECT: AGC Dam Site LOCATION: AGC Dam PROJECT NO: E1135088 CLIENT: DATE: 8/21/13 Failure Stress - psi Total Pore Pressure - psi Strain Rate - inches/min. Failure Strain - % V1 Failure - psi V3 Failure - psi Final Moisture - % Dry Density - pcf Calculated Diameter (in.) AT TEST Height - inches Effect. Cell Pressure - psi Dry Density - pcf Diameter - inches Height - inches TOTAL STRESS PARAMETERS SPECIMEN NO. Moisture Content - % INITIAL REMARKS: Specimens remolded to 95% +2 opt. TEST DESCRIPTION TYPE OF TEST & NO: CU with Pore Pressure SAMPLE TYPE: Remolded DESCRIPTION: Light Brown Sandy Lean Clay (CL) ASSUMED SPECIFIC GRAVITY: 2.7 SAMPLE LOCATION: Bulk 0.0-5.0 ft Percent -200: 67 0 10 20 30 40 50 0 10 20 30 40 50 60 70 80 SH EA R ST RE SS - PS I PRINCIPAL STRESS - PSI TRIAXIAL SHEAR TEST REPORT 0.00 10.00 20.00 30.00 40.00 50.00 0.0 5.0 10.0 15.0 20.0 DE VI AT OR S TR ES S -P SI AXIAL STRAIN - % TRIAX_Bulk 95 +2.xls APPENDIX C SUPPORTING DOCUMENTS TraceWithModifier Water Levl After Spcified Priod of Time GRAIN SIZE TRMINOLGYRELATIVE PROPRTIONS OF SAND AND GRAVEL TraceWithModifier Standar Pentration orN-ValueBlows/Ft.Descriptive Trm(Consistncy)Lose Very Stiff Exhibit C-1 Standar Pentration orN-ValueBlows/Ft. Ring SamplerBlows/Ft. Ring SamplerBlows/Ft. Medium DenseDenseVery Dense 0 - 1 < 34 - 9 2 - 4 3 - 4Medium-Stiff 5 - 930 - 50 WATER LEVEL AugerShelby Tube Ring SamplerGrab Sample 8 - 15 Split SponMacro Cre Rock Core PLASTICITY DESCRIPTIONTerm< 1515 - 29> 30 Descriptive Trm(s)of othe constituent Water InitiallyEncouterdWater Levl After aSpcified Period of Time Major Componetf SalePercnt ofDy Weight (More than 50% retained on No. 20 sieve.)Density determined by Snard Petration ResistanceInclus gravels, snds and silts. Hard Very Lose0 - 3 0 - 6 Very Soft7 - 18 Soft10 - 29 19 - 58 59 - 8 Stiff less than 5050 to 1,001,00 to 2,00 2,00 to 4,004,00 to 8,00> 9 LOCATION AD ELVATION OTES SAMPLING FIELD TESTS (HP)(T)(b/f) (PID)(OVA) DESCRIPTION F SYMBOLS AND ABREVIATIONS Descriptive Trm(Density) Non-plasticLowMediumHigh BouldersCobblesGravelSandSilt or Clay 10 - 18> 50 15 - 30 19 - 42> 30 > 42_ Hand PentrometrTorvaneStandar Pentration Test (blows p fot)Phot-Ionization DetctorOrganic Vapor AnalyzerWater levels indicated on the soil boringlogs are the levels measured in theborehole at the times indicated.Groundwater level variations will occurover time. In low permeability soils,accurate determination of groundwater levels is not possible with short termwater level observations. CONSISTENCY OF FINE-GRAINED SOILS(50% or more passing the No. 20 sieve.)Consistency determined by laboratory shear strength testing, fieldvisual-maual proceures r standard pnetratio resistance DESCRIPTIVE SOIL CASIFICATION > 8,00 Unless otherwise noted, Latitude and Longitude are approximately determined using a hand-held GPS device. The accuracyof such devices is variable. Surface elevation data annotated with +/- indicates that no actual topographical survey wasconducted to confirm the surfation. Instead, the surface elevation was approximately determined from topographicmaps of the area. oil classification is based on the Unified Soil Classification System. Coarse Grained Soils have more than 50% of their dryweight retained on a #200 sieve; their principal descriptors are: boulders, cobbles, gravel or sand. Fine Grained Soils haveless than 50% of their dry weight retained on a #200 sieve; they are principally described as clays if they are plastic, andsilts if they are slightly plastic or non-plastic. Major constituents may be added as modifiers and minor constituents may beadded according to the relative proportions based on grain size. In addition to gradation, coarse-grained soils are definedon the basis of their in-place relative density and fine-grained soils on the basis of their consistency. Plasticity Index01 - 1011 - 30> 30RELATIVE PROPRTIONS OF FINESDescriptive Trm(s)of othe constituent Percnt ofDy Weight< 55 - 12> 12 No Recovery RELATIVE DNSITY OF COARSE-GRAINED SOILS Particle SizeOver 12 in. (300 mm)12 in. to 3 in. (300 to 75m)3 in. to #4 sieve (75mm to 4.75 mm)#4 to #200e (4.75mm to 0.075Passing #2 sieve (0.075m) STRENGTH TERMS Unconfined CompresiveStrgth, Qu, f 4 - 8 GERAL NOTES UNIFIED SOIL CLASSIFICATION SYSTEM Exhibit C-2 Criteria for Assigning Group Symbols and Group Names Using Laboratory Tests A Soil Classification Group Symbol Group Name B Coarse Grained Soils: More than 50% retained on No. 200 sieve Gravels: More than 50% of coarse fraction retained on No. 4 sieve Clean Gravels: Less than 5% fines C Cu ? 4 and 1 ? Cc ? 3 E GW Well-graded gravel F Cu ? 4 and/or 1 ? Cc ? 3 E GP Poorly graded gravel F Gravels with Fines: More than 12% fines C Fines classify as ML or MH GM Silty gravel F,G,H Fines classify as CL or CH GC Clayey gravel F,G,H Sands: 50% or more of coarse fraction passes No. 4 sieve Clean Sands: Less than 5% fines D Cu ? 6 and 1 ? Cc ? 3 E SW Well-graded sand I Cu ? 6 and/or 1 ? Cc ? 3 E SP Poorly graded sand I Sands with Fines: More than 12% fines D Fines classify as ML or MH SM Silty sand G,H,I Fines classify as CL or CH SC Clayey sand G,H,I Fine-Grained Soils: 50% or more passes the No. 200 sieve Silts and Clays: Liquid limit less than 50 Inorganic: PI ? 7 and plots on or above ?A? line J CL Lean clay K,L,M PI ? 4 or plots below ?A? line J ML Silt K,L,M Organic: Liquid limit - oven dried ? 0.75 OL Organic clay K,L,M,N Liquid limit - not dried Organic silt K,L,M,O Silts and Clays: Liquid limit 50 or more Inorganic: PI plots on or above ?A? line CH Fat clay K,L,M PI plots below ?A? line MH Elastic Silt K,L,M Organic: Liquid limit - oven dried ? 0.75 OH Organic clay K,L,M,P Liquid limit - not dried Organic silt K,L,M,Q Highly organic soils: Primarily organic matter, dark in color, and organic odor PT Peat A Based on the material passing the 3-inch (75-mm) sieve B If field sample contained cobbles or boulders, or both, add ?with cobbles or boulders, or both? to group name. C Gravels with 5 to 12% fines require dual symbols: GW-GM well-graded gravel with silt, GW-GC well-graded gravel with clay, GP-GM poorly graded gravel with silt, GP-GC poorly graded gravel with clay. D Sands with 5 to 12% fines require dual symbols: SW-SM well-graded sand with silt, SW-SC well-graded sand with clay, SP-SM poorly graded sand with silt, SP-SC poorly graded sand with clay E Cu = D 60 /D 10 Cc = 6010 2 30 DxD )(D F If soil contains ? 15% sand, add ?with sand? to group name. G If fines classify as CL-ML, use dual symbol GC-GM, or SC-SM. H If fines are organic, add ?with organic fines? to group name. I If soil contains ? 15% gravel, add ?with gravel? to group name. J If Atterberg limits plot in shaded area, soil is a CL-ML, silty clay. K If soil contains 15 to 29% plus No. 200, add ?with sand? or ?with gravel,? whichever is predominant. L If soil contains ? 30% plus No. 200 predominantly sand, add ?sandy? to group name. M If soil contains ? 30% plus No. 200, predominantly gravel, add ?gravelly? to group name. N PI ? 4 and plots on or above ?A? line. O PI ? 4 or plots below ?A? line. P PI plots on or above ?A? line. Q PI plots below ?A? line.