Refine
Has Fulltext
- yes (2)
Is part of the Bibliography
- yes (2)
Year of publication
- 2023 (2) (remove)
Document Type
- Journal article (2)
Language
- English (2)
Keywords
- CORDEX Africa (1)
- Deep learning (1)
- Klima (1)
- Modell (1)
- Neuronales Netz (1)
- Zambia (1)
- climate (1)
- model output statistics (1)
- neural networks (1)
- renewable energy (1)
Institute
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.
Wind energy is a key option in global dialogues about climate change mitigation. Here, we combined observations from surface wind stations, reanalysis datasets, and state‐of‐the‐art regional climate models from the Coordinated Regional Climate Downscaling Experiment (CORDEX Africa) to study the current and future wind energy potential in Zambia. We found that winds are dominated by southeasterlies and are rarely strong with an average speed of 2.8 m·s\(^{−1}\). When we converted the observed surface wind speed to a turbine hub height of 100 m, we found a ~38% increase in mean wind speed for the period 1981–2000. Further, both simulated and observed wind speed data show statistically significant increments across much of the country. The only areas that divert from this upward trend of wind speeds are the low land terrains of the Eastern Province bordering Malawi. Examining projections of wind power density (WPD), we found that although wind speed is increasing, it is still generally too weak to support large‐scale wind power generation. We found a meagre projected annual average WPD of 46.6 W·m\(^{−2}\). The highest WPDs of ~80 W·m\(^{−2}\) are projected in the northern and central parts of the country while the lowest are to be expected along the Luangwa valley in agreement with wind speed simulations. On average, Zambia is expected to experience minor WPD increments of 0.004 W·m\(^{−2}\) per year from 2031 to 2050. We conclude that small‐scale wind turbines that accommodate cut‐in wind speeds of 3.8 m·s\(^{−1}\) are the most suitable for power generation in Zambia. Further, given the limitations of small wind turbines, they are best suited for rural and suburban areas of the country where obstructions are few, thus making them ideal for complementing the government of the Republic of Zambia's rural electrification efforts.