Theses and Dissertations
Issuing Body
Mississippi State University
Advisor
Paz, Joel O.
Committee Member
Mercer, Andrew E.
Committee Member
Pote, Jonathan W.
Committee Member
Tagert, Mary Love M.
Date of Degree
8-12-2016
Document Type
Dissertation - Open Access
Major
Biological Engineering
Degree Name
Doctor of Philosophy
College
James Worth Bagley College of Engineering
Department
Department of Agricultural and Biological Engineering
Abstract
The continuous decline of groundwater levels caused by variations in climatic conditions and crop water demands is an increased concern for the agricultural community. It is necessary to understand the factors that control these changes in groundwater levels so that we can better address declines and develop improved conservation practices that will lead to a more sustainable use of water. In this study, two machine learning techniques namely support vector regression (SVR) and the nonlinear autoregressive with exogenous inputs (NARX) neural network were implemented to predict daily groundwater levels in a well located in the Mississippi Delta Region (MDR). Results of the NARX model indicate that a Bayesian regularization algorithm with two hidden nodes and 100 time delays was the best architecture to forecast groundwater levels. In another study, the SVR and the NARX model were compared for the prediction of groundwater withdrawal and recharge periods separately. Results from this study showed that input data classified by seasons lead to incremental improvements in the model accuracy, and that the SVR was the most efficient machine learning model with a Mean Squared Error (MSE) of 0.00123 m for the withdrawal season. Analysis of input variables such as previous daily groundwater levels (Gw), precipitation (Pr), and evapotranspiration (ET) showed that the combination of Gw+Pr provides the optimal set for groundwater prediction and that ET degraded the modeling performance, especially during recharge seasons. Finally, the CROPGRO-Soybean crop model was used to simulate the impacts of different volumes of irrigation on the crop height and yield, and to generate the daily irrigation requirements for soybean crops in the MDR. Four irrigation threshold scenarios (20%, 40%, 50% and 60%) were obtained from the CROGRO-Soybean model and used as inputs in the SVR to evaluate the predicted response of daily groundwater levels to different irrigation demands. This study demonstrated that conservative irrigation management, by selecting a low irrigation threshold, can provide good yields comparable to what is produced by a high volume irrigation management practice. Thus, lower irrigation volumes can have a big impact on decreasing the amount of groundwater withdrawals, while still maintaining comparable yields.
URI
https://hdl.handle.net/11668/19988
Recommended Citation
Gutierrez, Sandra Milena Guzman, "Methodology to Predict Daily Groundwater Levels by the Implementation of Machine Learning and Crop Models" (2016). Theses and Dissertations. 3078.
https://scholarsjunction.msstate.edu/td/3078
Comments
water management||machine learning||Irrigation||Mississippi Delta||Groundwater