A combined downscaling-disaggregation weather generator for stochastic generation of multisite hourly weather variables over complex terrain: Development and multi-scale validation for the Upper Rhone River basin

This paper presents a combined downscaling and disaggregation weather generator developed for multisite generation of hourly precipitation and temperature time series over complex terrain. Daily regional weather variables are first generated from Generalized Linear Models based on daily atmospheric circulation indices from NCEP reanalysis. They are then disaggregated to the required spatial and temporal scales using a K-nearest neighbour approach. The weather generator is applied to the Upper Rhone River basin in the Swiss Alps. It successfully reproduces standard statistics for temperature as well as total and liquid precipitation at the temporal and spatial resolutions required for hydrological modelling of the system (3 h, 100 km2) and at lower resolutions down to those relevant at the river basin scale (3 days, ∼5500 km2). In addition, it reproduces the monthly distributions of the 1 °C isotherm elevation and of maximum precipitation amounts while preserving the spatial heterogeneity of weather variables and their spatial and temporal correlations. The weather generator can also be used to produce weather scenarios for the studied basin in a future climate.