Generating synthetic rainfall with geostatistical simulations

Abstract

Rainfall is an important driver of many Earth surface and subsurface processes such as floods, groundwater recharge, or plants growth. Models are used to investigate the physical response of different environmental aspects to a range of possible rainfall events. To provide meaningful outputs, such models require realistic inputs. However, a major challenge in these models is the representation of the chaotic behavior of rainfall as well as its high temporal and spatial variability. The primary sources of information about rainfall are rainfall measurements, numerical weather models and climate models. Because these sources of information are incomplete and uncertain, stochastic models have been developed to augment available data using statistical methods. Applications to rainfall modeling include interpolation of rainfall measurements, downscaling of numerical model outputs, or weather generators. In this study, we focus on geostatistical stochastic rainfall generation models, which aim at characterizing and reproducing the spatial structure of rainfall. To this end, the different steps of the geostatistical rainfall modeling are reviewed. The spatial structure of the rain is first characterized by variogram analysis of rainfall data. Then, geostatistical models are discussed that match the observed rain structure. Finally, geostatistical simulations are applied to the inferred models for the generation of synthetic rain fields. WIREs Water 2017, 4:e1199. doi: 10.1002/wat2.1199 This article is categorized under: Science of Water > Hydrological Processes Science of Water > Methods

When Should We Use Multiple-Point Geostatistics?

“…Rainfall is an inherently intermittent and highly spatially variable process (Benoit and Mariethoz 2017). Moreover, in some cases rain gauge data can be of poor quality, and it is not uncommon to only have binary wet/dry information (as opposed to rainfall accumulation).…”

Dealing with non-stationarity in sub-daily stochastic rainfall models

“…This model involves 11 parameters and aims at modelling both the marginal distribution of observed rain intensities and the space-time dependencies that exist within rain fields. It is briefly introduced hereafter; for more details the reader is referred to Benoit et al (2018). In this model, the marginal distribution of rain rates is accounted for by considering that rain measurements (R) originate from the censoring and power transform (involving parameters a 0 , a 1 , a 2 ) of a standardized multivariate Gaussian random field (Z) tainted by an additive measurement noise ( ∼ N (0, σ )) (Eq.…”

“…Stochastic rainfall models are statistical models that aim at simulating realistic random rains. For this purpose, they generate rainfall simulations which reproduce, in a distributional sense, a set of key rainfall statistics derived from an observation dataset (Benoit and Mariethoz, 2017). The practical interest of stochastic rainfall models is notably to complement numerical weather models for the simulation of rainfall heterogeneity at fine scales, and to quantify the uncertainty associated with rainfall reconstructions.…”

“…Recently, considerable attention has been paid to increasing the resolution of stochastic rainfall models so that they can mimic rainfall at sub-daily timescales. Currently, several high-resolution stochastic rainfall models are able to deal with precipitation data at typical resolutions of 1 min to 1 h in time and of 100×100 m 2 to 1×1 km 2 in space (see for example Leblois and Creutin, 2013;Paschalis et al, 2013;Benoit et al, 2018). At such scales, not only the marginal distribution of observed rain intensity matters but the space-time dependencies within rain fields are also important features of the rain process (Emmanuel et al, 2012;Marra and Morin, 2018).…”

“…In the case of stochastic rainfall modelling, the identification of stationary datasets or sub-datasets relies on some phenomenological guesses about rainfall, which serve as fuzzy guidelines to delineate stationary domains. Depending on the application and modelling choices, the parametrization of sub-daily stochastic rainfall models is considered as changing at scales ranging from seasons (Paschalis et al, 2013;Bárdossy and Pegram, 2016;Peleg et al, 2017) to single rainstorms (Caseri et al, 2016;Benoit et al, 2018).…”