Repeatable high-resolution statistical downscaling through deep learning

Abstract

Abstract. One of the major obstacles for designing solutions against the imminent climate crisis is the scarcity of high spatio-temporal resolution model projections for variables such as precipitation. This kind of information is crucial for impact studies in fields like hydrology, agronomy, ecology, and risk management. The currently highest spatial resolution datasets on a daily scale for projected conditions fail to represent complex local variability. We used deep-learning-based statistical downscaling methods to obtain daily 1 km resolution gridded data for precipitation in the Eastern Ore Mountains in Saxony, Germany. We built upon the well-established climate4R framework, while adding modifications to its base-code, and introducing skip connections-based deep learning architectures, such as U-Net and U-Net++. We also aimed to address the known general reproducibility issues by creating a containerized environment with multi-GPU (graphic processing unit) and TensorFlow's deterministic operations support. The perfect prognosis approach was applied using the ERA5 reanalysis and the ReKIS (Regional Climate Information System for Saxony, Saxony-Anhalt, and Thuringia) dataset. The results were validated with the robust VALUE framework. The introduced architectures show a clear performance improvement when compared to previous statistical downscaling benchmarks. The best performing architecture had a small increase in total number of parameters, in contrast with the benchmark, and a training time of less than 6 min with one NVIDIA A-100 GPU. Characteristics of the deep learning models configurations that promote their suitability for this specific task were identified, tested, and argued. Full model repeatability was achieved employing the same physical GPU, which is key to build trust in deep learning applications. The EURO-CORDEX dataset is meant to be coupled with the trained models to generate a high-resolution ensemble, which can serve as input to multi-purpose impact models.

Downscaling CORDEX Through Deep Learning to Daily 1 km Multivariate Ensemble in Complex Terrain

“…This region is an extension of the one employed in Quesada‐Chacón et al. (2022) since we sought to test the scalability of the methodology and to cover a larger and more relevant region for impact modelers. The subregions shown in Figure 1a were selected as representative, that is, Dresden exemplifies the climate of the northern flatlands, Fichtelberg (elevation 1,215 m) depicts the climate of the highest elevations with complex topography, while Vogtland exhibits intermediate climatic conditions.…”

“…The present study region (see Figure 1a) includes the Ore Mountains/Vogtland Nature Park (the longest nature park in Germany), the Saxon Switzerland National Park and a large portion of the flatlands toward the north of Saxony, including its capital, Dresden. This region is an extension of the one employed in Quesada-Chacón et al (2022) since we sought to test the scalability of the methodology and to cover a larger and more relevant region for impact modelers. The subregions shown in Figure 1a were selected as representative, that is, Dresden exemplifies the climate of the northern flatlands, Fichtelberg (elevation 1,215 m) depicts the climate of the highest elevations with complex topography, while Vogtland exhibits intermediate climatic conditions.…”

“…There, the theoretical, methodological, and computational bases were established and tested only for precipitation during past conditions, bearing scalability in mind, that is, also suitable for different: temporal and spatial resolutions, and regions. This paper builds upon the methods and containerized software environment previously developed in Quesada-Chacón et al (2022), while introducing several novel aspects. Specifically, our work encompasses (a) extending the workflow to include additional near-surface variables, (b) expanding the methods into the projection domain, and (c) assessing the quality of the produced data set by scrutinizing the fulfillment of the three major PP assumptions and conducting a comparative analysis to evaluate the climate change signal in relation to coarser resolutions.…”

Using Explainability to Inform Statistical Downscaling Based on Deep Learning Beyond Standard Validation Approaches

“…The UNET model is a popular CNN out-of-the-self configuration which has been widely used in image recognition problems (Ronneberger et al, 2015). In the context of climate downscaling, UNETs have been used in different studies in Europe (Doury et al, 2022;Quesada-Chacón et al, 2022). This model is composed of two different blocks: encoder and decoder (see Figure 2).…”

pyESDv1.0.1: An open-source Python framework for empirical-statistical downscaling of climate information

“…The grid-based analysis would contribute to the design of spatial downscaling models (e.g., Chen et al, 2012;Jia et al, 2011). Moreover, we intend to expand the selection of machine learning techniques by 835 including deep learning models that have been proven useful in downscaling (e.g., Baño-Medina et al, 2020;Quesada-Chacón et al, 2022). Finally, we intend to build a graphical, web-based interface to make the package more accessible and easy to use for researchers, students and people outside the scientific community.…”