@article{SteiningerAbelZiegleretal.2023,
  author    = {Steininger, Michael and Abel, Daniel and Ziegler, Katrin and Krause, Anna and Paeth, Heiko and Hotho, Andreas},
  title     = {ConvMOS: climate model output statistics with deep learning},
  series = {Data Mining and Knowledge Discovery},
  volume    = {37},
  journal   = {Data Mining and Knowledge Discovery},
  number    = {1},
  issn      = {1384-5810},
  doi       = {10.1007/s10618-022-00877-6},
  url       = {http://nbn-resolving.de/urn:nbn:de:bvb:20-opus-324213},
  pages     = {136-166},
  year      = {2023},
  abstract  = {Climate models are the tool of choice for scientists researching climate change. Like all models they suffer from errors, particularly systematic and location-specific representation errors. One way to reduce these errors is model output statistics (MOS) where the model output is fitted to observational data with machine learning. In this work, we assess the use of convolutional Deep Learning climate MOS approaches and present the ConvMOS architecture which is specifically designed based on the observation that there are systematic and location-specific errors in the precipitation estimates of climate models. We apply ConvMOS models to the simulated precipitation of the regional climate model REMO, showing that a combination of per-location model parameters for reducing location-specific errors and global model parameters for reducing systematic errors is indeed beneficial for MOS performance. We find that ConvMOS models can reduce errors considerably and perform significantly better than three commonly used MOS approaches and plain ResNet and U-Net models in most cases. Our results show that non-linear MOS models underestimate the number of extreme precipitation events, which we alleviate by training models specialized towards extreme precipitation events with the imbalanced regression method DenseLoss. While we consider climate MOS, we argue that aspects of ConvMOS may also be beneficial in other domains with geospatial data, such as air pollution modeling or weather forecasts.},
  subject      = {Klima},
  language  = {en}
}
@article{IbebuchiSchoenbeinAdakudluetal.2022,
  author    = {Ibebuchi, Chibuike Chiedozie and Sch{\"o}nbein, Daniel and Adakudlu, Muralidhar and Xoplaki, Elena and Paeth, Heiko},
  title     = {Comparison of three techniques to adjust daily precipitation biases from regional climate models over Germany},
  series = {Water},
  volume    = {14},
  journal   = {Water},
  number    = {4},
  issn      = {2073-4441},
  doi       = {10.3390/w14040600},
  url       = {http://nbn-resolving.de/urn:nbn:de:bvb:20-opus-262064},
  year      = {2022},
  abstract  = {This study compares the performance of three bias correction (BC) techniques in adjusting simulated precipitation estimates over Germany. The BC techniques are the multivariate quantile delta mapping (MQDM) where the grids are used as variables to incorporate the spatial dependency structure of precipitation in the bias correction; empirical quantile mapping (EQM) and, the linear scaling (LS) approach. Several metrics that include first to fourth moments and extremes characterized by the frequency of heavy wet days and return periods during boreal summer were applied to score the performance of the BC techniques. Our results indicate a strong dependency of the relative performances of the BC techniques on the choice of the regional climate model (RCM), the region, the season, and the metrics of interest. Hence, each BC technique has relative strengths and weaknesses. The LS approach performs well in adjusting the first moment but tends to fall short for higher moments and extreme precipitation during boreal summer. Depending on the season, the region and the RCM considered, there is a trade-off between the relative performances of the EQM and the MQDM in adjusting the simulated precipitation biases. However, the MQDM performs well across all considered metrics. Overall, the MQDM outperforms the EQM in improving the higher moments and in capturing the observed return level of extreme summer precipitation, averaged over Germany.},
  language  = {en}
}
@article{AichAkhundzadahKnuerretal.2017,
  author    = {Aich, Valentin and Akhundzadah, Noor Ahmad and Knuerr, Alec and Khoshbeen, Ahmad Jamshed and Hattermann, Fred and Paeth, Heiko and Scanlon, Andrew and Paton, Eva Nora},
  title     = {Climate change in Afghanistan deduced from reanalysis and coordinated regional climate downscaling experiment (CORDEX)—South Asia Simulations},
  series = {Climate},
  volume    = {5},
  journal   = {Climate},
  number    = {2},
  issn      = {2225-1154},
  doi       = {10.3390/cli5020038},
  url       = {http://nbn-resolving.de/urn:nbn:de:bvb:20-opus-198024},
  pages     = {38},
  year      = {2017},
  abstract  = {Past and the projected future climate change in Afghanistan has been analyzed systematically and differentiated with respect to its different climate regions to gain some first quantitative insights into Afghanistan's vulnerability to ongoing and future climate changes. For this purpose, temperature, precipitation and five additional climate indices for extremes and agriculture assessments (heavy precipitation; spring precipitation; growing season length (GSL), the Heat Wave Magnitude Index (HWMI); and the Standardized Precipitation Evapotranspiration Index (SPEI)) from the reanalysis data were examined for their consistency to identify changes in the past (data since 1950). For future changes (up to the year 2100), the same parameters were extracted from an ensemble of 12 downscaled regional climate models (RCM) of the Coordinated Regional Climate Downscaling Experiment (CORDEX)-South Asia simulations for low and high emission scenarios (Representative Concentration Pathways 4.5 and 8.5). In the past, the climatic changes were mainly characterized by a mean temperature increase above global level of 1.8 °C from 1950 to 2010; uncertainty with regard to reanalyzed rainfall data limited a thorough analysis of past changes. Climate models projected the temperature trend to accelerate in the future, depending strongly on the global carbon emissions (2006-2050 Representative Concentration Pathways 4.5/8.5: 1.7/2.3 °C; 2006-2099: 2.7/6.4 °C, respectively). Despite the high uncertainty with regard to precipitation projections, it became apparent that the increasing evapotranspiration is likely to exacerbate Afghanistan's already existing water stress, including a very strong increase of frequency and magnitude of heat waves. Overall, the results show that in addition to the already extensive deficiency in adaptation to current climate conditions, the situation will be aggravated in the future, particularly in regard to water management and agriculture. Thus, the results of this study underline the importance of adequate adaptation to climate change in Afghanistan. This is even truer taking into account that GSL is projected to increase substantially by around 20 days on average until 2050, which might open the opportunity for extended agricultural husbandry or even additional harvests when water resources are properly managed.},
  language  = {en}
}