Stochastic downscaling of precipitation to high-resolution scenarios in orographically complex regions: 1. Model evaluation

Abstract

[1] The simulation of space-time precipitation has been studied since the late 1980s. However, there are still many open issues concerning the most appropriate approach to simulate it, specially in highly heterogeneous areas, such as in mountain environments. For this reason, we present here a comprehensive investigation of the Space-Time NeymanScott Rectangular Pulses model, with the purpose of analyzing its performance in a challenging Alpine environment of Switzerland and identifying weaknesses that can drive future improvements. The results point at the suitability of the model in reproducing not only the basic statistics at different temporal aggregations, but also the more challenging distributional and scaling properties. The intrinsic stationarity of the model in space, induced by the parameter estimation procedure, poses occasional limitations with regard to the accurate simulation of the variability of the observed climate characteristics, which are strongly influenced by local microclimates. However, the model is able, even in the complex Alpine environment, to preserve the spatial patterns observed in the actual precipitation process. The study allowed (i) to conclude about the robustness of the model and its suitability for multisite downscaling of precipitation estimated from climate model simulations, as reported in the companion paper, and (ii) to put in evidence some limitations that require further consideration to improve space-time rainfall generation.Citation: Bordoy, R., and P. Burlando (2014), Stochastic downscaling of precipitation to high-resolution scenarios in orographically complex regions: 1. Model evaluation, Water Resour. Res., 50, 540-561,

An Analysis of the Occurrence Probabilities of Wet and Dry Periods through a Stochastic Monthly Rainfall Model

“…In particular, mathematical modeling of hydrological data as a stochastic process is of interest to a variety of hydrological areas such as flood forecasting, reservoir operations and agricultural planning [61][62][63]. Current state-of-the-art stochastic precipitation models at a single site or for multiple sites can reproduce a wide range of statistics from hourly scale extremes to larger aggregation periods [64][65][66][67][68][69][70][71][72][73][74][75]. In particular, numerous approaches for the stochastic modeling of daily rainfall data are available in the hydrological and climatological literature [76][77][78][79][80][81][82][83][84].…”

Evaluation of a spatial rainfall generator for generating high resolution precipitation projections over orographically complex terrain

“…Rainfall generators exist for both single site and spatial applications and three main categories of rainfall generators can be recognized (Bordoy and Burlando 2014). The first category is the one of the Markovian models, in which rainfall occurrence and rainfall amount are modelled separately in a two-step approach (Wilks 1998;Mehrotra et al 2006;Kim et al 2008;Wilks 2009).…”

“…Given their formulation, the models are not expected to exhibit scaling behaviour (e.g., Marani 2003). However, Olsson and Burlando (2002), Bordoy and Burlando (2014) empirically demonstrated that a NSRP model can well reproduce rainfall properties at different timescales, applying both its single site (20 min-1 week) and spatial-temporal configuration (1 h-45 days). Another advantage of this type of models is the possibility to include third-order moments in the simulation algorithm to better model extremes (Burton et al 2008).…”

“…The choice was driven by the advantages related to the elegance of the model (limited number of parameters and fitting through observed time series), the inclusion of third order moments to model extremes, and the wide use of this type of models in the literature, which has demonstrated their robustness and adaptability to work in different climates and environments (e.g., Van Vilet et al 2012;Borgomeo et al 2014;Forsythe et al 2014). The only study, which has made a thorough evaluation of a spatiotemporal NSRP model over a mountainous area, is the one of Bordoy and Burlando (2014). These authors tested the model in an alpine catchment in Switzerland (5244 km 2 ) for 10 reference rain gauges.…”

“…The evaluation considers basic rainfall statistical properties, rainfall spatial and temporal (lag 1) intermittency, and indices of extremes. While our study area size is almost the same as that of Bordoy and Burlando (2014), the analysis of 145 gauges, as compared to their 10 stations, allows a much better analysis of the capacity of the spatiotemporal model to simulate spatial differences as well as a quantification of the loss of model capacity under specific geographical conditions. We also conduct a direct comparison between the output of the spatiotemporal model and the output of a single site generator for the 145 gauges.…”