STORM v.2: A simple, stochastic decision-support tool for exploring the impacts of climate, and climate change at, and near the land surface in gauged watersheds.

Climate change is expected to have major impacts on land surface and subsurface processes through its expression on the hydrological cycle, but the impacts to any particular basin or region are highly uncertain. Non-stationarities in the frequency, magnitude, duration, and timing of rainfall events have important implications for human societies, water resources, and ecosystems. The conventional approach for assessing the impacts of climate change is to downscale global climate model output and use it to drive regional and local models that express the climate within hydrology near the land surface. While this approach may be useful for linking global general circulation models to regional hydrological cycle, it is limited for examining the details of hydrological response to climate forcing for a specific location over timescales relevant to decision makers. For example, management of flood or drought hazard requires detailed information that includes uncertainty based on variability in storm characteristics, rather than on differences between models within an ensemble. To fill this gap, we present the second version of our STOchastic Rainfall Model (STORM), an open-source, parsimonious and user-friendly modeling framework for simulating climatic expression as rainfall fields over a basin. This work showcases the use of STORM in simulating ensembles of realistic sequences, and spatial patterns of rainstorms for current climate conditions, and bespoke climate change scenarios that are likely to affect the water balance near the Earth's surface. We outline, and detail STORM's new approaches: one copula for linking marginal distributions of storm intensity and duration; orographic stratification of rainfall using the copula approach; a radial decay-rate for rainfall intensity which takes into consideration potential, but unrecorded, maximum storm intensities; an optional component to simulate storm start date-times via circular/directional statistics; and a simple implementation for modelling future climate scenarios. We also introduce a new pre-processing module that facilitates the generation of model input in the form of probability density functions (PDFs) from historical data for subsequent stochastic sampling. Independent validation showed that the average performance of STORM falls within a 5.5% of the historical seasonal total rainfall in the Walnut Gulch Experimental Watershed (Arizona, USA) that occurred in the current century. [ABSTRACT FROM AUTHOR]