Skip to main content
ARS Home » Research » Publications at this Location » Publication #299085

Title: Global sensitivity analysis for UNSATCHEM simulations of crop production with degraded waters

Author
item Skaggs, Todd
item Suarez, Donald
item Corwin, Dennis

Submitted to: Vadose Zone Journal
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 2/22/2014
Publication Date: 6/13/2014
Publication URL: http://handle.nal.usda.gov/10113/62827
Citation: Skaggs, T.H., Suarez, D.L., Corwin, D.L. 2014. Global sensitivity analysis for UNSATCHEM simulations of crop production with degraded waters. Vadose Zone Journal. DOI: 10.2136/vzj2013.09.0171.

Interpretive Summary: Irrigated agriculture produces nearly 40 % of the global food harvest and is critical for meeting global food demand. However, irrigated agriculture is increasingly faced with diminished water and land availability. One strategy for maintaining or increasing productivity in the face of resource scarcity is to make greater use of marginal quality lands and waters. In implementing such a strategy, a key factor for sustainability is soil salinity. Irrigation waters, especially recycled or otherwise marginal quality waters, contain salts that can accumulate in soils over time and reduce yields. In arid and semi-arid regions where rainfall is not sufficient to flush the salts from the root zone, it is necessary to apply excess irrigation water to leach the soil. To avoid wasting water, and to lessen impacts on groundwater quality, it is desirable that soil leaching be minimized to the extent possible. In this work, we investigated an advanced simulation model and decision support tool called UNSATCHEM that can be used to design water reuse systems and optimally manage soil salinity. Model sensitivity analyses were performed to identify the model inputs or parameters having the greatest impact on model outputs. We found that model parameters specifying crop salt tolerance were among the most significant for managing recycled irrigation waters. The results will guide future research and development of modeling tools for the design and management of irrigation systems using marginal quality lands and waters.

Technical Abstract: One strategy for maintaining irrigated agricultural productivity in the face of diminishing resource availability is to make greater use of marginal quality waters and lands. A key to sustaining systems using degraded irrigation waters is salinity management. Advanced simulation models and decision support tools can aid in the design and management of water reuse systems, but at present model predictions and related management recommendations contain significant uncertainty. Sensitivity analyses can help characterize and reduce uncertainties by revealing which parameter variations or uncertainties have the greatest impact on model outputs. In this work, the elementary effects method was used to obtain global sensitivity analyses of UNSATCHEM seasonal simulations of forage corn (Zea mays L.) production with differing irrigation rates and water compositions. Sensitivities were determined with respect to four model outcomes: crop yield, average root zone salinity, water leaching fraction, and salt leaching fraction. For a multiple-season, quasi-steady scenario, the sensitivity analysis found that overall the most important model parameters were the plant salt tolerance parameters, followed by the solute dispersivity. For a single-season scenario with irrigation scheduling based on soil water deficit, soil hydraulic parameters were the most important; the computed salt leaching fraction was also strongly affected by the initial ionic composition of the exchange phase because of its impact on mineral precipitation. In general, parameter sensitivities depend of the specifics of a given modeling scenario, and procedures for routine use of models for site-specific degraded irrigation water management should include site-specific uncertainty and sensitivity analyses. The elementary effects method used in this work is a useful approach for obtaining parameter sensitivity information at relatively low computational cost.