A hydrologist's view of climate model downscaling for assessments of the hydrology and water resources implications of climate change

Most assessment of hydrologic and water resources implications of climate change require information at spatial scales smaller than those of global, and regional climate models. Furthermore, climate models essentially all have biases that are large enough to dominate the climate change signal (e.g., difference between simulated future and past climate), hence a bias correction step is usually performed along with downscaling. From a scientific standpoint, dynamic downscaling, in which a regional climate model is run with boundary conditions from a global model, is preferred on the basis of physical consistency. In practice though, most applications studies have used statistical downscaling, mostly because it is much less computationally intensive, and allows a greater range of future conditions to be evaluated. Furthermore, some downscaling is usually required even of regional climate model output, and bias correction is almost always required We review a number of recent studies that have used dynamic and/or statistical downscaling for hydrologic and/or water resources climate impact assessments, and attempt to identify conditions under which dynamic downscaling is essential.