Modeling of saturated thickness of the Ogallala aquifer in Texas

Intensive water level declines in the Southern High-Plains (SHP) aquifer have been is byproduct of intensive groundwater production for irrigation purposes. Groundwater modeling to estimate groundwater availability is important because groundwater planning and management is necessary to avoid aquifer depletion. The saturated thickness of the aquifer is an indicator of the water availability in this unconfined aquifer.
Geographically weighted regression (GWR) is an innovative data driven modeling that was used in this study to evaluate local relationships between factors and processes that affects groundwater availability. The GWR was compared to a least squares regression evaluate its performance at a regional scale. GWR models in general out-performed their ordinary least squares regression counterparts. However, both models could not predict the larger saturated thicknesses found in the northern portions of SHP. The GWR model delineates the spatial non-stationarity in relationship between the inputs and the outputs. NDVI (a surrogate for agricultural production) was seen to be an important indicator of saturated thickness. Well depth (a surrogate for aquifer thickness), especially in conjunction with slope (an indicator for regional flow) were also noted to be significant. GWR also reduced the spatial autocorrelation of the residuals and thus was able to explain a greater degree of variance in the data. Nearly 75%-80% of the predictions were within a factor of two for GWR while only 58% – 72% of the predictions satisfied that criteria for the ordinary least squares model. GWR is useful to delineate the spatial variability of the factors influencing the saturated thickness and can be useful to parameterize physically-based regional-scale groundwater flow models.