Radar for Improving Precipitation Estimates and Optimization of Hydropower Energy Production

RadPro has improved the production chain for numerical weather prediction with a strong link to the end user (eg. hydro power producers, flood forecasters and Yr.no). Important outcomes of the project are high resolution (1x1 km) hourly analyses and forecasts of key weather variables like temperature, precipitation, wind, air pressure at sea level, cloud area fraction, relative humidity, shortwave radiation by combining multiple data sources and numerical weather predictions models. The project improved the weather forecasts by developing a four-dimensional variational data assimilation technique (4D-Var), which allows the use of more continuous observations like radar. To further improve the forecast, the project also developed new statistical methods for post-processing of weather parameters. In addition to weather data from the Norwegian Meteorological Institute, the amateur Netatmo temperature and precipitation observations are now used operationally in post-processing. The project developed new hourly analyses (i.e., best estimates at observation time) on a 1km-by-1km grid. The analyses of precipitation and temperature combine model outputs with observations (in-situ and remote sensing), while for the other variables a statistical downscaling scheme has been applied. A reliable inflow forecast needs both weather analyses and forecasts, and a key to the success of this project was to establish a consistency between the analysis and forecast products. In the end, the analysis and forecasting products were used to produce both simulations and forecasts of water flowing into hydropower reservoirs. The results show that inflow simulations are, in many cases, improved by using the new weather analysis products from RadPro whereas the inflow forecasts benefits from the improved consistency between weather analysis and forecasts. One post-processed forecast re-run for each day will be produced and made available to all users. Note that the post-processed forecast re-run does not imply a re-run of the numerical model. The 4D-Var approach has been tested in parallel run next to the operational system in Autumn 2019. Most of the partners succeed to implement and evaluate the impact of the disseminated (historic and real-time) RadPrO analyses and post-processed forecasts. Calibration of the hydrological models using the historical results was indispensable. Most of the partners found the new RadPrO (field) products better compared to the point based data available through the YR-API. The project focused on the user's activities, preferences and weather sensitivities. The project allowed partners to get familiar with the services, data management and expertise at MET Norway, in line with user needs. In other words, the usefulness of the meteorological data increases when production chain continues into the users' systems. To summarise, Radpro contributed in producing better inflow forecasts and in optimising the planning of hydropower generation. The post-processed products are made available through MET Norway's broadcasting infrastructure (thredds.met.no) from 19th of March 2018. Besides, the post-processed products are also distributed via Yr.no from 19th March 2018. The RadPro development products are now available from September 2013 on thredds.met.no for evaluation to all partners. It improved the access to observations and their usage with more advanced methods for data assimilation. Freely available software was developed to quality control observations and to post-process weather model data. A dedicated website has been created at https://github.com/metno/NWPdocs (see also the wiki pages) to host the documentation and keep track of the dataset issues.

MET Nordic (Analysis & Forecast): product that contains weather data (temperature, precipitation, wind, air pressure, cloud area fraction, relative humidity, shortwave downwelling radiation) on a 1km x 1km grid for a given hour from 2013 over Norway, Denmark, Sweden, Finland, part of the Baltic countries. New statistical methods for spatial analysis (Lussana et al., 2019) and post-processing (Nipen et al., 201). Publicly available programs/software tools; e.g. TITAN for automatic data quality control and GRIDPP for post-processing gridded weather forecasts. Partners found the new RadPrO (field) products gave better results when calibrating hydrological models compared to the point based data available through the YR-API. Improved simulations and forecasts of water flow into hydropower reservoirs in most study cases when using RadPro data. MET Nordic datasets will be regularly updated by MET Norway.

With lower electricity prices and reduced profitability, the optimization of energy production becomes essential for the hydropower industry. Hydropower is also under pressure from increased competition from solar and wind power. In Norway, where hydropower represents 95% of the electricity production, reducing production costs and increasing efficiency are thus of paramount importance. Improved forecasts of inflow to hydropower reservoirs will be essential to achieve these aims. The RadPro project will exploit the full potential of radar measurements in producing precipitation products for hydrological applications. RadPro will deliver improved high-resolution best estimates (analyses) and short-range forecasts of precipitation, temperature, and wind suitable for hydrological applications. It will improve the interpretation of radar measurements and optimise their combination with in-situ observations and numerical weather prediction outputs. As a final product, RadPro will provide: i) improved high-resolution (1hr accumulated precipitation on a 1km grid) precipitation analyses especially over areas where radar coverage is poor and where in-situ observations are sparse; ii) a newly developed flow-dependent data assimilation scheme, capable of using all available radar observations around the assimilation time to improve the timing and the spatial distribution of precipitation; iii) improved high-resolution (1 km grid) forecasts of accumulated precipitation, by exploiting relationships between historical radar measurements and historical forecast values; and iv) a new archive of historical times series of precipitation estimates useful for planning and calibration purposes.