Soil system dynamics for learning about complex, feedback-driven agricultural resource problems: model development, evaluation, and sensitivity analysis of biophysical feedbacks

Soils form the foundation of terrestrial ecosystems that regulate the processes and functions driving ecosystem goods and services provisions that humans rely on, including agriculture. Pressing agricultural resource challenges persist, including those involving irrigation, fertilization, and salinization, due to the complex, coupled, and feedback-driven connectivity of various soil processes that are difficult to manage. Soil moisture dynamics cross-cut these processes and is a logical integration point for generating understanding and new insights for improved agroecosystem management. This paper presents an integrated soil-water-nutrient-plant interaction model (built within a system dynamics framework) with the purpose of replicating soil moisture evolution for a set of unique soils and climates, examining model performance given common irrigation (e.g., frequency and application rates) and crop management considerations (e.g., fertilization, tillage, cover cropping), and evaluating via sensitivity analysis model robustness and quantifying influential management parameters effect on core bio-physical feedbacks at the soil-level. The model has four main state variables (soil moisture, soil nitrogen, soil sodium, and plant canopy cover) that interact dynamically via feedback processes (formulated as coupled partial differential equations) between them. Exogenous variables included precipitation time-series data and required climatic parameters to determine reference (potential) evapotranspiration. The time-unit used from simulation was 1 day (time-step = 0.25) with a simulation horizon of 365 days. The model was calibrated using a variety of sources in the literature and with comparison to observed soil moisture data from four sites in Texas, USA, and evaluated statistically for accuracy (mean bias), precision, (coefficient of determination), and overall fit (Theil inequality statistics). Sensitivity analyses were conducted for a variety of hydroclimate forcing and irrigation, fertilization, and crop management decisions to examine the impacts to soil moisture evolution, soil salinity, and cropping profitability, among other variables. Calibration results showed high degrees of agreement between observed and predicted values (mean r2 = 0.67, mean bias = 0.008%). Sensitivity results demonstrated that precipitation frequency was more influential than precipitation depth in regulating soil moisture, that irrigation threshold (i.e., the soil moisture level inducing irrigation) was the variable most influential to crop profitability (which was maximized at the lowest irrigation threshold value), and that several conservation management strategies (i.e., no-till with residue management or cover cropping) improved soil moisture and crop profitability, contrary to common management perceptions. Several other tests for alternative fertilization, irrigation, and tile drain installation strategies produced results that corroborate common observations of agroecosystem management (i.e., despite environmental risks, crop profitability was enhanced). Future model extensions include expansion of the irrigation, fertilization, and crop management decision making factors to better capture how decisions that respond to economic and policy signals influence resource use and soil system dynamics. Modeling these complex, feedback driven agroecosystems processes remains an arena for future modeling innovations that will support important resource management improvements.