Machine learning the abiotic stressor impacts on nitrogen availability and photo energy use in dryland forage systems under different tillage and green manuring practices

Authors: SARKAR, RESHMI - Texas A&M Agrilife; Northup, Brian; LONG, CHARLES - Texas A&M Agrilife; SINGH, VIJAY - Texas A&M University

Submitted to: Discover Soil
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 1/2/2025
Publication Date: 1/28/2025
Citation: Sarkar, R., Northup, B.K., Long, C.R., Singh, V.P. 2025. Machine learning the abiotic stressor impacts on nitrogen availability and photo energy use in dryland forage systems under different tillage and green manuring practices. Discover Soil. 2. Article 5. https://doi.org/10.1007/s44378-025-00029-4.
DOI: https://doi.org/10.1007/s44378-025-00029-4

Interpretive Summary: The U.S. Southern Great Plains (SGP) includes parts of Kansas, Oklahoma, Texas, Colorado, and New Mexico, and contains a wide assortment of land types, soils, climates, and agricultural operations. Key features of the climate that affects rainfed agriculture in the region is the regular occurrence of droughts and high temperatures, and irregular amounts and timing of precipitation during years. These features of the climate affect plant growth directly, and indirectly by their influences on the how efficiently crops use solar radiation, and by causing nitrogen (N) stress in crops. One important crop used to produce both forage for cattle and as a bioenergy crop is sorghum sudangrass (SSG). It is a drought tolerant hybrid grass, but the stresses caused by dry conditions, high temperatures, and other stressors are challenges to efficient production. While SSG can be an effective forage, the important stress factors that negatively affect production must be identified to help develop suitable management strategies for crop production. We applied different machine learning (ML) techniques to data collected in central Oklahoma during 2005 to 2015 to help define which stress factors were most important in causing N stress and reducing efficiencies in use of solar radiation, and whether management techniques could reduce stresses on growth. We used a long-term model simulation to test the influences of four N treatments (spring-fallowed combined with inorganic N fertilizer, and three spring-grown green manures) under no-till (NT) and conventional (CT) tillage systems. We found that differences in forage production by sorghum-sudangrass in response to the treatments fell within a narrow (500 lb/acre) range. Some treatments allowed higher efficiencies in using solar radiation or generated lower N stresses, while others resulted in low efficiencies in using solar radiation and high levels of N stress. The ML models highlighted significant impacts related to average daily temperatures on amounts of water (78% of responses) and N (79% of responses) stresses. Amount of soil water, timing of rainfall, and solar radiation had lesser effects. Results also identified different features of the climate that had large effects on N stress during wet (tillage system), normal (water stress), and drought (temperature) years. While environmental factors defined the majority of influences on N stress, the applied management systems showed lesser effects. Different combinations of treatments were effective at reducing N stress during specific types of growing seasons, with the no-till x field pea combination the most consistent across all types of growing seasons. Oats were also effective under both tillage systems during dry growing seasons. These results show that producers in the SGP, depending on conditions, have multiple options for using green manures to reduce N stress in supporting forage production by sorghum-sudangrass.

Technical Abstract: Sorghum sudangrass (SSG), is a drought tolerant C4 hybrid grass grown as a forage and bioenergy crop in the southern Great Plains (SGP). However, it faces challenges from abiotic stressors, particularly during droughts. While growing SSG can be an effective biomass crop, stress factors must be identified to describe suitable strategies for crop production. This study used machine-learning (ML) to examine relationships at the soil-plant-atmosphere interface and define important abiotic stressors and agri-environmental variables that influence nitrogen (N) stress. A long-term simulation involving four N treatments (spring-fallowed combined with inorganic N fertilizer, and one of three spring-grown green manures) under no-till (NT) and conventional (CT) tillage was applied. Machine learning was used to describe how abiotic stressors interacted with oats (OG), field pea (FP) and grass pea (GP) green manures x tillage system combinations. Mean amounts of biomass produced by the treatments fell in a narrow range (500 kg ha-1). Higher photosynthetically active radiation use efficiency (PARUE) and lower N stress occurred under NT+FP compared to low PARUE and high N-stress under CT+inorganic N. Regression models highlighted significant impacts for average daily temperatures on water (R2 = 0.78), and N (R2 = 0.79) stresses. Multiple regressions identified influential environmental variables during wet (tillage system), normal (water stress), and dry (temperature) seasons on N stress. Management showed lesser effects on N stress, and different combinations were effective during specific types of growing seasons, particularly wet and dry growing seasons. Results show SGP producers, depending on conditions, have multiple options for using green manures to reduce N stress in support of forage production by SSG.