Development of a countrywide spatially predictive hydrological model for
Panama using the Soil and Water Assessment Tool
Abstract
Water availability and extremes in river discharge associated with
floods and droughts are critical determinants of human welfare and
ecological function. Modeling the effects of climate scenarios and other
social and environmental changes on waterways is thus a key component of
effective planning and risk mitigation. Yet, the calibration of
multiple-basin models, such as for a national planning framework, can be
difficult due to limitations on quality and spatial coverage of
available hydrological observations. In this manuscript, we build a
process-based whole-country hydrological model for Panama using the Soil
and Water Assessment Tool (SWAT). We also extend SWAT by deriving a
precipitation interpolation model that incorporates regional climatic
variability and spatial autocorrelation of precipitation, and we
validate the model using data from 35 hydrological stations. Without
calibration, the default application of SWAT reasonably predicted
spatiotemporal variability in mean monthly discharge (NSE=0.70), but
largely failed to predict variability (NSE=0.26) and maxima (NSE=0.22).
However, with our relatively simply precipitation interpolation
sub-model, we were able to strengthen predictions of discharge
(NSE=0.87), but also able to more than double predictive ability for
variance (NSE=0.62) and maxima (NSE=0.53). This moderate modification
may allow process-based hydrological models such as SWAT to be much more
broadly applied; crucially, even across regions with scarce hydrological
data. The resulting precipitation and hydrology layers provide important
baseline information for Panama.