Wetland Hydrologic Modeling through Physically-based and Data-driven Approaches

Abstract

Hydrology influences vegetation composition, species richness, primary productivity, accumulation of organic material, and nutrient availability and cycling in wetlands. Any alteration to wetland hydrology can change the biogeochemistry of wetlands and the dominant nutrient removal pathways. A better understanding of wetland hydrologic processes offers valuable insights into the study and modeling of wetland nutrient removal and cycling. This study, performed in four stages, was undertaken to advance the current knowledge of wetland hydrologic modeling by introducing various models. In Chapter 2, two artificial neural network (ANN)-based models were developed and validated for predicting hourly water levels (WLs) in wetlands characterized by water tables at or near the surface that respond rapidly to precipitation. The first method makes use of hourly precipitation data and WL data from nearby sites. The second method is a combination of ANN, recursive digital filter, and recession curve method and does not require any nearby site. The proposed methods were tested at two headwater wetlands in coastal Alabama. Site 17 had two nearby sites whose WLs were highly correlated with Site 17’s. The root-mean-square error and Nash–Sutcliffe efficiencies were 2.9 cm and 0.98, respectively, when the first method was applied to Site 17. The second method was tested at Site 32. A combination of ANN and base-flow separation methods proved to be very efficient for WL prediction at this site, especially when the duration of quick-response components of individual events was less than 6 h. The proposed methodologies, therefore, proved useful in predicting WLs in wetlands dominated by both surface water and groundwater. This chapter has been published in Journal of Hydrologic Engineering. In Chapter 3, we built on the previous stage of the study by eliminating the need for WL data from nearby wetlands as inputs to the model. A hybrid modeling approach was developed for improved WL predictions in wetlands, by coupling the watershed model SWAT with ANNs. To demonstrate the utility of this approach, the developed model was used to assess the potential impacts of climate change on WL fluctuations at the headwater wetland site 17. Model results forecast potential increase in medium (20th-80th percentile) WLs and a decreasing trend for low (0th-20th percentile) and high (80th-100th percentile) WLs. Water levels predicted with this hybrid model were also used to explore possible teleconnections between El Niño Southern Oscillation (ENSO) and WLs in the study wetland. Results showed that both precipitation and the variations in WLs were partially affected by ENSO. The findings suggested wetter conditions in winter during El Nino in Coastal Alabama. However, WL reduction in spring during El Nino is expected. Hence, understanding the hydrologic processes in wetlands going through wetting and drying cycles and the biochemical and ecological implications of those cycles is a critical task. A manuscript written from this chapter is currently under review in Hydrological Sciences. In wetlands going through wetting/drying cycles, simulation of nutrient processes and biogeochemical reactions in both ponded and unsaturated wetland zones are needed for an improved understanding of wetland functioning for water quality improvement. Sharifi et al. (2017) extended the ponded version of WetQual model (Hantush et al. 2013, Kalin et al. 2013, Sharifi et al. 2013) by adding a soil moisture accounting module and biogeochemical relationships for improved N and C cycles in variably saturated zones of wetlands. He used Richards’ Equation (RE) to calculate soil moisture dynamics. This resulted in an unnecessarily complex model because not only RE equation is notoriously complex and numerically difficult to deal with, but also WetQual only needs average moisture content (it is a lumped model). In Chapter 4, a depth-averaged solution to the RE (called DARE) for one-dimensional vertical unsaturated flow is presented to predict the temporal variation of the average moisture content of the root zone and the layer below it in unsaturated parts of wetlands. This will make Sharifi et al. (2017) version of WetQual more practical and computationally efficient. The DARE model was verified versus Hydrus-1D which utilizes the full RE, using field data from the Hupselse Beek watershed in the Netherlands. Gravity drainage version of DARE works well with a comparison to Hydrus-1D, under all the assigned atmospheric boundary conditions of varying fluxes for all examined soil types including sandy loam, loam, sandy clay loam, and sand. A full-term version of DARE offers reasonable accuracy dominantly in the root zone. The focus of Chapter 5 was adding a flow routing module to ponded version of WetQual and creating a graphical user interface (GUI) that brings the hydrologic and water quality modeling under one umbrella. Earlier versions of WetQual required users take care of wetland hydrology independently. The developed GUI also provides opportunities for processing and visualization of input/outputs and helps the users identify any source of error in inputs or model simulations. In addition, the WetQual GUI was equipped with powerful post-processing and sensitivity analysis modules. The GUI performs Generalized Likelihood Uncertainty Estimation (GLUE) and Bayesian Monte Carlo simulation and Maximum Likelihood estimation (BMCML) analyses. The utility of the WetQual GUI was demonstrated through a case study in a small restored wetland called Barnstable wetland, located in Kent Island, Maryland. This GUI can be used as a learning tool for hydrology and water quality processes in wetlands.