Search

Climate models face limitations in their ability to accurately represent highly variable atmospheric phenomena. To resolve fine‐scale physical processes, allowing for local impact assessments, downscaling techniques are essential. We propose spateGAN, a novel approach for spatio‐temporal downscaling of precipitation data using conditional generative adversarial networks. Our method is based on a video super‐resolution approach and trained on 10 years of country‐wide radar observations for Germany. It simultaneously increases the spatial and temporal resolution of coarsened precipitation observations from 32 to 2 km and from 1 hr to 10 min. Our experiments indicate that the ensembles of generated temporally consistent rainfall fields are in high agreement with the observational data. Spatial structures with plausible advection were accurately generated. Compared to trilinear interpolation and a classical convolutional neural network, the generative model reconstructs the resolution‐dependent extreme value distribution with high skill. It showed a high fractions skill score of 0.6 (spatio‐temporal scale: 32 km and 1 hr) for rainfall intensities over 15 mm h−1 and a low relative bias of 3.35%. A power spectrum analysis confirmed that the probabilistic downscaling ability of our model further increased its skill. We observed that neural network predictions may be interspersed by recurrent structures not related to rainfall climatology, which should be a known issue for future studies. We were able to mitigate them by using an appropriate model architecture and model selection process. Our findings suggest that spateGAN offers the potential to complement and further advance the development of climate model downscaling techniques, due to its performance and computational efficiency. Plain Language Summary: Natural disasters like floods, hail, or landslides originate from precipitation. Global climate models are an important tool to understand these hazards and derive expected changes in a future climate. However, they operate on spatial and temporal scales that limit the regional ability to reflect their small‐scale characteristics. This has led to the development of dynamical and statistical downscaling methods. Due to their computational efficiency, machine learning algorithms recently got increased attention as a method for improving the spatial resolution of climate data. Here, we describe a new deep learning model that allows to simultaneously increasing both the temporal and spatial resolution of precipitation data. Our presented approach enhances the spatial resolution by a factor of 16 and the temporal resolution by a factor of 6. The generated rain fields are hardly identifiable as artificially generated and exhibit the typical structure, movement, and distribution of observed rain fields. Key Points: High performance simultaneous spatial and temporal precipitation downscaling enabled by 3D convolution approachGeneration of realistic high‐resolution ensembles using probabilistic conditional generative adversarial networksLow computational effort compared to dynamical downscaling approaches [ABSTRACT FROM AUTHOR]

Reliable prediction of heavy precipitation events causing floods in a world of changing climate is crucial for the development of appropriate adaption strategies. Many attempts to provide such predictions have already been conducted but there is still much potential for improvement left. This is particularly true for statistical downscaling of heavy precipitation due to changes present in the corresponding atmospheric drivers. In this study, a circulation pattern (CP) conditional downscaling to the station level is proposed which considers occurring frequency changes of CPs. Following a strict circulation‐to‐environment approach we use atmospheric predictors to derive CPs. Subsequently, precipitation observations are used to derive CP conditional cumulative distribution functions (CDFs) of daily precipitation. Raw precipitation time series are sampled from these CDFs. Bias correction is applied to the sampled time series with quantile mapping (QM) and parametric transfer functions (PTFs) as methods being tested. The added value of this CP conditional downscaling approach is evaluated against the corresponding common non‐CP conditional approach. The performance evaluation is conducted by using Kling–Gupta Efficiency (KGE), root mean squared error (RMSE), and mean absolute error (MAE) metrics. In both cases the applied bias correction is identical. Potential added value can therefore only be attributed to the CP conditioning. It can be shown that the proposed CP conditional downscaling approach is capable of yielding more reliable and accurate downscaled daily precipitation time series in comparison to a non‐CP conditional approach. This can be seen in particular for the extreme parts of the distribution. Above the 95th percentile, an average performance gain of +0.24 and a maximum gain of +0.6 in terms of KGE is observed. These findings support the assumption of conserving and utilizing atmospheric information through CPs can be beneficial for more reliable statistical precipitation downscaling. Due to the availability of these atmospheric predictors in climate model output, the presented method is potentially suitable for downscaling precipitation projections. [ABSTRACT FROM AUTHOR]

Precipitation change is critical for the Three-River Headwaters (TRH) region, which serves downstream communities in East Asia. The spring (March–May) precipitation over the TRH region shows an increasing trend from 1979 to 2018, as revealed by a Chinese gridded precipitation product (CN05.1). However, the physical processes responsible for this precipitation change are still unclear. This study investigated the characteristics of spring precipitation and the water budget over the TRH region using the ERA5 global reanalysis and the Weather Research and Forecast (WRF) Model. The WRF version employed in this study includes online calculations of the atmospheric water budget and an evapotranspiration (ET) tagging procedure to trace evapotranspired water in the atmosphere. Both ERA5 and WRF reproduce the spring precipitation increase. Moreover, WRFD02 (with a 3-km domain) reduces the wet bias by around 60% and 77% compared to WRFD01 (9 km) and ERA5 (30 km). Both ERA5 and WRF demonstrate that the increase of spring precipitation is dominated by moisture convergence, especially the atmospheric water fluxes from the southern boundary. The enhanced moisture inflow is sustained by enhanced mass flux while the enhanced moisture outflow is sustained by increased moisture. The ET-tagging results further demonstrate the weakened precipitation recycling process because of the significant increase of precipitation produced by external moisture. Compared to ERA5, the reduced wet bias with WRF is attributed to a better spatial resolution of orographic barrier effects, which reduce the southerly water fluxes. The results highlight the potential of regional climate downscaling to better represent the atmospheric water budget in complex terrain. [ABSTRACT FROM AUTHOR]