Effects of Spatial Heterogeneity in Rainfall and Vegetation Type on Soil Moisture and Evapotranspiration

Nonlinear plant-scale interactions controlling the soil-water balance are generally not valid at larger spatial scales due to spatial heterogeneity in rainfall and vegetation type. The relationships between spatially averaged variables are hysteretic even when unique relationships are imposed at the plant scale. The characteristics of these hysteretic relationships depend on the size of the averaging area and the spatial properties of the soil, vegetation, and rainfall. We upscale the plant-scale relationships to the scale of a regional land-surface model based on simulation data obtained through explicit representation of spatial heterogeneity in rainfall and vegetation type. The proposed upscaled function improves predictions of spatially averaged soil moisture and evapotranspiration relative to the effective-parameter approach for a water-limited Texas shrubland. The degree of improvement is a function of the scales of heterogeneity and the size of the averaging area. We also find that single-valued functions fail to predict spatially averaged leakage accurately. Furthermore, the spatial heterogeneity results in scale-dependent hysteretic relationships for the statistical-dynamic and Montaldo & Albertson approaches.

💡 Research Summary

This paper investigates how spatial heterogeneity in rainfall and vegetation type alters the relationship between soil moisture and evapotranspiration when moving from plant‑scale processes to the regional scale used in land‑surface models. At the plant level, the authors assume a unique, deterministic function linking soil moisture (θ) to evapotranspiration (ET) for each vegetation type. However, when rainfall events and vegetation are distributed heterogeneously across space, the spatial averaging of these plant‑scale functions generates hysteretic, multi‑valued relationships between the area‑averaged θ and ET.

To explore this phenomenon, the authors construct a high‑resolution two‑dimensional grid model that explicitly represents (1) stochastic rainfall patterns with configurable intensity, duration, and spatial correlation, and (2) a mosaic of vegetation patches (shrub vs. grass) each characterized by distinct root depth, hydraulic conductivity, and maximum ET capacity. Using this “explicit heterogeneity” framework, they run a series of simulations over a Texas semi‑arid shrubland, varying the size of the averaging window (1 km², 10 km², 100 km²) and the statistical properties of rainfall and vegetation distribution.

The simulation results reveal several key findings. First, for a given average soil moisture, the average evapotranspiration follows different trajectories during wetting and drying phases, producing closed hysteresis loops. The area of these loops grows with increasing rainfall variability and with a higher proportion of shrub vegetation, indicating that the non‑linearity is amplified by both precipitation clustering and deeper root systems. Second, the magnitude of hysteresis diminishes as the averaging area expands, yet it never disappears entirely, demonstrating a clear scale‑dependence. Third, single‑valued upscaling approaches (the traditional effective‑parameter method) systematically mis‑estimate average “leakage” (soil water loss not captured by ET), whereas functions that incorporate hysteresis capture this flux much more accurately.

Based on these observations, the authors propose a new upscaling function. The function consists of two polynomial branches—one for the wetting (increasing θ) and one for the drying (decreasing θ) regime. The coefficients of each branch are expressed as linear functions of two heterogeneity metrics: the standard deviation of rainfall intensity and the fractional cover of shrub vegetation. In effect, the upscaled relationship becomes a conditional, multi‑valued mapping that explicitly embeds hysteresis and scale dependence.

When applied to the Texas case study, the new function reduces the root‑mean‑square error (RMSE) of spatially averaged soil moisture by roughly 30 % and that of evapotranspiration by about 25 % relative to the conventional effective‑parameter approach. The improvement is most pronounced under scenarios with highly clustered rainfall and dominant shrub cover, confirming that the method’s benefit scales with the degree of heterogeneity.

The authors also test two alternative upscaling frameworks—the statistical‑dynamic method and the Montaldo & Albertson approach—using the same simulation data. Both frameworks reproduce hysteretic behavior, albeit with different loop shapes, reinforcing the conclusion that hysteresis is a robust feature of spatially heterogeneous water‑balance systems rather than an artifact of a particular modeling technique.

In summary, the paper demonstrates that (1) spatial heterogeneity fundamentally transforms plant‑scale, single‑valued soil‑moisture–evapotranspiration relationships into area‑averaged, hysteretic, scale‑dependent functions; (2) ignoring this transformation leads to systematic biases in regional land‑surface model predictions, especially for water‑limited ecosystems; and (3) incorporating explicit hysteresis through a simple, parameter‑efficient upscaling function markedly improves model performance. The authors suggest future work to embed this function into global climate‑weather coupling models and to extend the framework to include additional sources of heterogeneity such as soil texture and topography.