Refine
Has Fulltext
- yes (8)
Document Type
- Article (7)
- Doctoral Thesis (1)
Language
- English (8)
Keywords
An important aspect of rainfall estimation is to accurately capture extreme events. Commercial microwave links (CMLs) can complement weather radar and rain gauge data by estimating path-averaged rainfall intensities near ground. Our aim with this paper was to investigate attenuation induced complete loss of signal (blackout) in the CML data. This effect can occur during heavy rain events and leads to missing extreme values. We analyzed 3 years of attenuation data from 4,000 CMLs in Germany and compared it to a weather radar derived attenuation climatology covering 20 years. We observed that the average CML experiences 8.5 times more blackouts than we would have expected from the radar derived climatology. Blackouts did occur more often for longer CMLs (e.g., >10 km) despite their increased dynamic range. Therefore, both the hydrometeorological community and network providers can consider our analysis to develop mitigation measures.
The most reliable areal precipitation estimation is usually generated via combinations of different measurements. Path-averaged rainfall rates can be derived from commercial microwave links (CMLs), where attenuation of the emitted radiation is strongly related to rainfall rate. CMLs can be combined with data from other rainfall measurements or can be used individually. They are available almost worldwide and often represent the only opportunity for ground-based measurement in data-scarce regions. However, deriving rainfall estimates from CML data requires extensive data processing. The separation of the attenuation time series into rainy and dry periods (rain event detection) is the most important step in this processing and has a high impact on the resulting rainfall estimates. In this study, we investigate the suitability of Meteosat Second Generation Spinning Enhanced Visible and InfraRed Imager (MSG SEVIRI) satellite data as an auxiliary-data-based (ADB) rain event detection method. We compare this method with two time-series-based (TSB) rain event detection methods. We used data from 3748 CMLs in Germany for 4 months in the summer of 2021 and data from the two SEVIRI-derived products PC and PC-Ph. We analyzed all rain event detection methods for different rainfall intensities, differences between day and night, and their influence on the performance of rainfall estimates from individual CMLs. The radar product RADKLIM-YW was used for validation. The results showed that both SEVIRI products are promising candidates for ADB rainfall detection, yielding only slightly worse results than the TSB methods, with the main advantage that the ADB method does not rely on extensive validation for different CML datasets. The main uncertainty of all methods was found for light rain. Slightly better results were obtained during the day than at night due to the reduced availability of SEVIRI channels at night. In general, the ADB methods led to improvements for CMLs performing comparatively weakly using TSB methods. Based on these results, combinations of ADB and TSB methods were developed by emphasizing their specific advantages. Compared to basic and advanced TSB methods, these combinations improved the Matthews correlation coefficient of the rain event detection from 0.49 (or 0.51) to 0.59 during the day and from 0.41 (or 0.50) to 0.55 during the night. Additionally, these combinations increased the number of true-positive classifications, especially for light rainfall compared to the TSB methods, and reduced the number of false negatives while only leading to a slight increase in false-positive classifications. Our results show that utilizing MSG SEVIRI data in CML data processing significantly increases the quality of the rain event detection step, in particular for CMLs which are challenging to process with TSB methods. While the improvement is useful even for applications in Germany, we see the main potential of using ADB methods in data-scarce regions like West Africa where extensive validation is not possible.
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.
We present high-resolution rainfall maps from commercial microwave link (CML) data in the city of Ouagadougou, Burkina Faso. Rainfall was quantified based on data from 100 CMLs along unique paths and interpolated to achieve rainfall maps with a 5-min temporal and 0.55-km spatial resolution for the monsoon season of 2020. Established processing methods were combined with newly developed filtering methods, minimizing the loss of data availability. The rainfall maps were analyzed qualitatively both at a 5-min and aggregated daily scales. We observed high spatiotemporal variability on the 5-min scale that cannot be captured with any existing measurement infrastructure in West Africa. For the quantitative evaluation, only one rain gauge with a daily resolution was available. Comparing the gauge data with the corresponding CML rainfall map pixel showed a high agreement, with a Pearson correlation coefficient > 0.95 and an underestimation of the CML rainfall maps of ∼10%. Because the CMLs closest to the gauge have the largest influence on the map pixel at the gauge location, we thinned out the CML network around the rain gauge synthetically in several steps and repeated the interpolation. The performance of these rainfall maps dropped only when a radius of 5 km was reached and approximately one-half of all CMLs were removed. We further compared ERA5 and GPM IMERG data with the rain gauge and found that they had much lower correlation than data from the CML rainfall maps. This clearly highlights the large benefit that CML data can provide in the data-scarce but densely populated African cities.
Climate observations are crucial for societies around the globe to adapt to natural hazards in a changing climate. However, large parts of the world, especially developing countries, do not have sufficient access to climate information. Rainfall is the major driver of hydrological processes that cause flooding or droughts which are responsible for the majority of natural disasters. Precipitation is especially hard to estimate and forecast due to its high spatial and temporal variability. There is also a large heterogeneity in the abundance of conventional rainfall sensors such as rain gauges and weather radars and their setup is cost and maintenance-intensive.
To close the observational gap for rainfall sensors commercial microwave links (CMLs) to measure path-averaged rainfall are a promising alternative since more than 90% of the global population lives in areas where they are deployed. However, due to their opportunistic nature and indirect measurement, they are prone to systematic and random errors that require quantification, attribution to causes, and correction in order to provide high-quality quantitative precipitation estimates (QPE). The same holds for systematic errors in weather radar QPE. The objective of this thesis is the improvement of CML and weather radar QPE by mitigating systematic errors. The main innovation is the application of deep learning techniques which have proven to provide high-performing solutions to model atmospheric processes. Convolutional neural networks (CNNs) are applied to improve the detection of rain events in commercial microwave link data in order to reduce the impact of attenuation falsely attributed to rainfall by more than 50%. Another application is the simultaneous increase of the temporal resolution, ground-adjustment, and advection-correction of radar QPE to reduce biases by 20% and mitigate a sampling error. Additional studies to investigate and disentangle the complex error structure of commercial microwave links have been conducted: First, the performance of state-of-the-art CML processing techniques and the resulting CML QPE were compared using one year of country-wide rainfall observations identifying processing steps with the highest impact on QPE quality. Second, missing rainfall extremes due to a complete loss of signal in heavy rain (blackouts) have been investigated showing that they occur more frequently than radar-derived climatology suggests. Third, signal fluctuations that are not due to rainfall (anomalies) have been detected using manual data flagging and their impact on CML QPE has been investigated. While there was ambiguity in the flagging, removing anomalies significantly improved the quality of rainfall estimates. In summary, the presented results show that systematic errors in CML and weather radar QPE can be quantified and corrected using a data-driven approach, but attribution to causes remains difficult. Trained artificial neural networks prove to be a robust tool to provide high-quality QPE that can be easily transferred to new locations and future time periods within the same climatic region. CML QPE is shown to have a remarkably high quality when compared to gauge-adjusted weather radar QPE and the results presented will foster a successful deployment of CMLs to close the observational gap in climate science.
Global spatio-temporal ERA5 precipitation downscaling to km and sub-hourly scale using generative AI
(2025)
The spatial and temporal distribution of precipitation significantly impacts human lives. While reanalysis datasets provide consistent long-term global precipitation information that allows investigations of rainfall-driven hazards like larger-scale flooding, they lack the resolution to capture the high spatio-temporal variability of precipitation and miss intense local rainfall events. Here, we introduce spateGAN-ERA5, the first deep learning-based spatio-temporal downscaling of precipitation data on a global scale. SpateGAN-ERA5 enhances ERA5 precipitation data from 24 km and 1 h to 2 km and 10 min, delivering high-resolution rainfall fields with realistic spatio-temporal patterns and accurate rain rate distribution, including extremes. Its computational efficiency enables the generation of a large ensemble of solutions, addressing uncertainties inherent to downscaling challenges and supports practical applicability for generating high-resolution precipitation data for arbitrary ERA5 time periods and regions on demand. Trained solely on data from Germany and validated in the US and Australia, considering diverse climates, including tropical rainfall regimes, spateGAN-ERA5 demonstrates strong generalization, indicating robust global applicability. It fulfills critical needs for high-resolution precipitation data in hydrological and meteorological research.
