.. _technical_description: Technical Description ======================= HydroMet combines the methods of DWD's KOSTRA and CatRaRE and integrates them in the Climate DT workflow. Here, the most important concepts are summarised. For more detailed information, the reader is referred to the following publications. Essential literature for KOSTRA is - Koutsoyiannis et al. (1998) :cite:`koutsoyiannis_mathematical_1998` - Shehu et al. (2023) :cite:`shehu_regionalisation_2023` and for CatRaRE - Lengfeld et al. (2021) :cite:`lengfeld_catrare_2021` KOSTRA (Kostradamus) ---------------------- Kostradamus is the implementation of the KOSTRA methods in HydroMet. It is designed to facilitate dynamic threshold computation, which is essential for extreme precipitation analysis. Kostradamus computes statistical parameters independently for each grid cell or pixel based on the principles outlined in Koutsoyiannis et al. (1998). In the method, only the most extreme rainfalls of each year are used to estimate a Generalized Extreme Value (GEV) distribution, which is heavy-tailed towards extreme events. Therefore, the approach does not take into account all the observed rainfall of a year, but the heavily aggregated Intensity Annual Maximum Series (IAMS). IAMS is the maximum recorded precipitation for different duration periods in a given year. It is calculated with the one-pass algorithms of the Climate DT workflow. The estimated GEV distribution characterising the strength of extreme precipitation is described by four parameters (Xi, Alpha, Theta, Eta), which Kostradamus estimates systematically. The parameters are estimated using bounded minimization algorithms, maximum likelihood estimators, and L-moment ratio computations. After determining the GEV, Kostradamus derives the Depth Duration Frequencies (DDF) for a range of predefined return periods (1, 2, 3, 5, 10, 20, 30, 50, 100 years) by accessing quantiles of the distribution. The results of a Kostradamus run are stored in two NetCDF files: one containing the parameters and the IAMS (KO_params.nc), and the other containing the derived DDF (DDF.nc), encompassing computations for the entire data series. CatRaRE (WetCat) ------------------ The methods of CatRaRE are implemented in the WetCat routines of HydroMet, consisting of two parts: The detection of extreme event objects (a continuous area of heavy rainfall) and the subsequent filtering to determine independent events from the objects. Through iterative processing of precipitation time-series data, WetCat automatically summarises the data using predefined duration windows. For every duration, it is checked whether precipitation exceeds a return period greater than 5 years, leveraging the Depth Duration Frequency (DDF) provided by Kostradamus. Continuous areas that exceed the threshold are extreme precipitation objects that are saved for later analysis. The following event detection checks the detected objects for independence to determine a final catalogue of extreme events. There are two criteria for determining whether two or more objects are independent: temporal and spatial independence. Between two objects, there must be a minimum break of four hours or the minimum duration of the shorter of two succeeding events, if each of them lasts longer than four hours. For example, two independent 1-hour objects need a 4-hour break to be independent, but a 6-hour and a 12-hour object need at least a 6-hour break. Spatial independence is assessed by checking whether any grid cells overlap between two objects. Lastly, if multiple objects are not independent of each other, only the most representative object for the event is listed in the catalogue. There are different approaches to determining the most representative object. HydroMet uses the Extremity (Müller and Kaspar, 2014 :cite:`muller_event-adjusted_2014`) concept as a measure, which links the object's return period to the affected area, to indicate the object's strength. Therefore, with multiple objects, the one with the highest Extremity is listed in the catalogue. In the dataset, Extremity is represented by the Parameter “Eta_KOSTRA”. Integration in the workflow (HydroMet) ---------------------------------------- :numref:`hydromet_workflow` depicts the integration of HydroMet into the Climate DT workflow. Before HydroMet is run, a few contributing processes must be run. First, data from the climate models is extracted using the GSV interface, which regrids it onto a regular lat/lon grid. This data is used in two ways. The first path continuously filters the data using the One-pass layer to extract the Intensity Annual Maximum Series (IAMS) for each calendar year. These IAMS are used to determine the extreme rain statistics at the end of the model experiment by the Kostradamus package. The other path feeds raw precipitation in the HydroMet application. Since updated statistics are only available after a model run, archived Kostradamus computations are used to determine extreme precipitation objects using the WetCat package. In summary, this procedure allows HydroMet to use Kostradamus statistics for extreme event identification while Climate DT simulations are running, enabling the creation of an extreme event catalogue on par with the running Climate simulations. New statistics for the current run are available at the end and can be compared with previous ones, or new statistics can be calculated from data across multiple simulations. Lastly, the outputs from Kostradamus and WetCat are uploaded to the DestinE Data Lake. .. figure:: ../../figures/hydromet_workflow.PNG :name: hydromet_workflow Scheme of the integration of HydroMet with its packages Kostradamus and WetCat into the Climate DT workflow.