Date of Award

5-2026

Document Type

Thesis

Degree Name

Master of Science

Department

Natural Resources and Environmental Sciences

Abstract

Climate change and increasing drought frequency pose significant challenges to agricultural sustainability in Texas. Cotton production is especially sensitive to temperature variability and water availability, making it vulnerable to drought stress, reduced soil moisture, and changing growing-season conditions. Despite numerous studies on climate impacts on agriculture, limited research has integrated climate projections with drought indices, machine learning, and crop simulation models to assess future cotton yield vulnerability across Texas climate divisions. Therefore, this study evaluated historical climate trends, compared predictive modeling approaches, and projected future cotton yields under climate change scenarios. Historical climate variables, including maximum temperature (Tmax), minimum temperature (Tmin), precipitation, potential evapotranspiration, soil water storage, and drought indices, were analyzed with cotton yield data from 1968 to 2024. Multiple Linear Regression and Random Forest models were developed and evaluated using correlation coefficient (r) and Root Mean Square Error. Future climate projections from CMIP6 LOCA datasets under SSP245 and SSP585 scenarios were used to estimate cotton yield for 2030–2050. The FAO AquaCrop model was also applied to simulate crop response to water availability and support model comparison.

Results indicated that temperature and PET were the most influential predictors of cotton yield across Texas climate divisions. Random Forest showed improved predictive performance compared to Multiple Linear Regression, with correlation coefficient (r) values ranging from approximately 0.24 to 0.72, while MLR values ranged from approximately 0.06 to 0.63 across different climate divisions and production systems. While AquaCrop showed low to moderate agreement with observed yields (r ≈ 0.08–0.48). Future projections show irrigated yields ranging from approximately 851–1173 lb/ac (SSP245) and 839–1115 lb/ac (SSP585) using AquaCrop, compared to 597–916 lb/ac (SSP245) and 601–899 lb/ac (SSP585) from RF. Overall yields ranged from about 457–917 lb/ac (AQ) and 338–551 lb/ac under SSP245, while non-irrigated yields ranged from approximately 322–706 lb/ac (AQ) and 318–498 lb/ac (RF). Irrigated cotton consistently showed greater resilience than non-irrigated cotton. Overall, this study demonstrated that integrating machine learning and process-based crop models improves climate impact assessment and provides useful insights for climate-resilient cotton production and agricultural water management in Texas.

Index Terms— AquaCrop model, cotton yield, machine learning (ML), random forest (RF), standardized precipitation index (SPI), standardized precipitation evapotranspiration index (SPEI), Texas.

Committee Chair/Advisor

Ali Fares

Committee Co-Chair

Ripendra Awal

Committee Member

Rabi Mohtar

Committee Member

Ashok Mishra

Committee Member

Anoop Veettil

Committee Member

Atikur Rahman

Publisher

Prairie View A&M University

Rights

© 2021 Prairie View A & M University

Creative Commons License
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.

Date of Digitization

6/17/2026

Contributing Institution

John B Coleman Library

City of Publication

Prairie View

MIME Type

Application/PDF

Download


Share

COinS