| dc.description.abstract | Precise estimates of irrigation yield effects are essential for farmers when making irrigation investment decisions. The project goal is to improve our knowledge of irrigation yield effects in the soils of the Alabama Blackland Prairie region. The study objectives encompass estimating the irrigation yield effects of corn (Zea mays) and soybeans (Glycine max L.) based on terrain, soil, and climate parameters and determining the variables with the highest effect on crop yield. We evaluated two statistical analyses: Integrated Nested Laplace Approximation (INLA) and machine learning. The first analysis, assess the individual effect of irrigation, soil, and terrain properties on crop yield, and the second analysis evaluated four machine learning algorithms, including Sup port Vector Machine, Elastic Net Regression, Stepwise Regression, and Random Forest, under the following aspects: 1) identification of variables influencing yield predictions, 2) yield predictions based on adjacent fields, 3) yield predictions from irrigated fields, 4) yield prediction for a specific year utilizing data from other years within the same field, and 5) yield prediction of one field for a specific year using other fields with crop production in the same years. On an area of 3,400 ha with 22 pivot-irrigated and rainfed fields in the Alabama Blackland Prairie region, the research utilized 183 yield datasets collected from shrink and swell soils from 2012 to 2021. Spa tial derivatives derived from elevation information from the National Elevation Database, soil data from the POLARIS database, and drought indices calculated using precipitation, evapotranspira tion, and temperature data from the Parameter elevation Regressions on Independent Slopes Model (PRISM) Climate Group were integrated into the analysis. Results from INLA showed that terrain variables had a greater effect in corn than in soybeans yield. Although, these variables had the greatest effect, the yield increase or decrease was minimum in both crops. Results from machine learning showed that the accuracy of corn and soybean yield predictions is lower when relying only on one year of training data, where terrain attributes exhibited more significant influence compared to soil and climate properties. Conversely, incorporating data from multiple fields spanning several years and diverse crop yields into the training dataset led to greater accuracy of predictions, where the impact of climate properties is more notorious. | en_US |
