Soil-water storage dynamics link soils, vegetation and climate with consequent impacts on water and biogeochemical cycles and on ecosystem functions. Projected hydro-climatic changes include increased frequency of extreme events (droughts; floods) which have impacts on most major ecosystem processes, including changes in phenology, GPP, soil respiration and NEE. We explored landscape-scale spatial-temporal dynamics of water storage in the critical zone, using a spatially explicit numerical simulation model (NASA-LIS) over the Midwest US region (~2 million km2). Hydro-climatic forcing (precipitation; potential evapotranspiration) was derived based on downscaled, coarse-resolution (4x4 km) reanalysis products (NLDAS-2). We generated LIS model outputs over three decades and used long-term, time-series daily data from 10 in-situ observation sites (Ameriflux and SCAN networks) to compare: (1) trends in spatiotemporal precipitation and storage patterns; (2) empirical and analytical probability density functions (pdfs) of soil-water storage; and (3) empirical and analytical pdfs for soil-water threshold crossing times as a risk assessment framework for drought persistence. We hypothesized that the analytical pdf models would be comparable with numerical LIS output, which would justify analytical stochastic modeling approaches for certain long-term applications. Additionally, we hypothesized that changes in rainfall patterns would manifest in a more variable soil-water regime, including increasing drought duration.
Total mean daily rainfall over the past 30 years has remained stable in the Midwest US, but with decreased rainfall frequency and increased rainfall depths (p < 0.05), translating to a more extreme hydro-climatic regime. These patterns, and changes in PET, are manifested in time-series of soil-water storage across the region. However, we find no corresponding changes in soil-water storage or drought frequency and persistence. Agreement between analytical and empirical pdfs for soil-water storage depended upon: (1) factors controlling modulation of the impacts of seasonal hydro-climatic forcing (e.g., root-zone depth); (2) differences in spatial scales (e.g., site-specific data vs grid-scale averages; and (3) differences in conceptual representations of key processes in the models. We also found differences in agreement depending on plant functional type, with largest disagreement in forested ecosystems. The analytical pdfs serve as surrogates for LIS simulations for estimation of exceedance probability of soil-water thresholds set as “drought” conditions, and to evaluate persistence/recurrence of such “droughts” based on crossing time pdfs. The direct link between rainfall depth and frequency and the crossing time pdfs provides a probabilistic framework for drought risk assessment under a changing hydro-climate, and could be connected to changes in ecosystem processes.