Data about radar rainfall nowcasts of 80 high-intensity rainfall in 5 cities in the Netherlands from 2008 to 2021

This repository contains the radar rainfall nowcasts for 80 high-intensity rainfall events spread over 5 urban areas (Amsterdam, the Hague, Groningen, Maastricht, and Eindhoven) in the Netherlands. <br>Rainfall events chosen criteria: <br>The high-intensity events were selected for each municipality by using RadarTools (http://radartools.nl). RadarTools can list the rainfall events based on the highest climatological rainfall accumulation at any 1-km2 grid cell within each municipality in the Netherlands since 2008. RadarTools sets different thresholds to select rainfall events of different periods. For example, the selection criteria for 1-h period is that the precipitation sum in any grid cell is above 15 mm. For 24-h period, the threshold is 30 mm. Rainfall is not necessarily continuous during the period. We selected the eight highest 1-h and eight highest 24-h events in each of the 5 municipalities from 2008 to 2021. When selecting the 24-h events, the events that covered the period of the previously selected 1-h events were excluded per urban area to ensure independence of the events. This selection procedure led to 5 (cities) × 2 (durations) × 8 (events) = 80 events. <br>Nowcasts: <br>For every five-minute time step in each of the events, a probabilistic nowcast was made by pySTEPS (v0.2 used) libraries (Ayzel et al., 2019; Pulkkinen et al., 2019). We used the same Pysteps setup as in Imhoff et al. (2020, https://doi.org/10.1029/2019WR026723), which consisted of the STEPS (short-term ensemble prediction system) nowcasting method using a semi-Lagrangian advection method and the Lucas‐Kanade optical flow method (using the QPE from time t-3 to t). An autoregressive model of order 2 and 8 cascade levels was used. <br>All nowcasts have a time step of 5 min and a forecast horizon of 4 hours. Nowcasts are first run with real-time QPE from Royal Netherlands Meteorological Institute (https://dataplatform.knmi.nl/dataset/nl-rdr-data-rtcor-5m-1-0). The 1-h events before 2020 are also run with climatological (https://dataplatform.knmi.nl/dataset/rad-nl25-rac-mfbs-em-5min-netcdf4-2-0) and CARROTS-adjusted QPE (https://doi.org/10.4121/13573814). Hence, in total, there are 9,060 separate norecasts (each with 4-hour forecast horizon and 20 ensemble members). The nowcasts are saved as netCDF files in folders with the following structure: event duration (1 or 24 hrs) and QPE products used (realtime, CARROTS, or climatological) &gt; city name &gt; event folder &gt; individual nowcasts. The filenames of the event folder indicate the onsets of the high-intensity rainfall events in UTC. The filenames of nowcasts indicate the issue time of the nowcasts in UTC.

本仓库包含荷兰5个城市区域（阿姆斯特丹、海牙、格罗宁根、马斯特里赫特及埃因霍温）内80场强降雨事件的雷达降雨临近预报数据。

降雨事件筛选标准：
本次研究通过RadarTools（http://radartools.nl）筛选各市政辖区的强降雨事件，该工具可基于2008年以来荷兰各市政辖区内任意1平方公里网格单元的最高气候学降雨累积量，列出符合条件的降雨事件。RadarTools会针对不同时长的降雨事件设置差异化阈值：例如，1小时时长事件的筛选标准为任意网格单元的降水总量超过15毫米；24小时时长事件的阈值则为30毫米。降雨过程在上述时段内未必连续。我们从2008年至2021年间的5个市政辖区中，各选出8场1小时强度最高及8场24小时强度最高的事件。在筛选24小时事件时，会排除覆盖此前已选1小时事件时段的事件，以确保各事件相互独立。本次筛选流程最终得到5（城市数）×2（时长）×8（事件数）=80场事件。

临近预报：
针对每场事件的每5分钟时间步长，使用pySTEPS（v0.2版本，Ayzel等，2019；Pulkkinen等，2019）库生成概率临近预报。我们采用与Imhoff等（2020，https://doi.org/10.1029/2019WR026723）一致的PySTEPS配置，该配置采用基于半拉格朗日平流方法和卢卡斯-卡纳德光流法（使用t-3至t时刻的定量降水估计（Quantitative Precipitation Estimation, QPE））的STEPS（短期集合预报系统，short-term ensemble prediction system）临近预报方法，同时使用了2阶自回归模型及8级级联结构。
所有临近预报的时间步长为5分钟，预报时效为4小时。临近预报首先采用荷兰皇家气象研究所（Royal Netherlands Meteorological Institute, KNMI）的实时定量降水估计QPE数据运行（https://dataplatform.knmi.nl/dataset/nl-rdr-data-rtcor-5m-1-0）。2020年之前的1小时事件还分别使用了气候学QPE数据（https://dataplatform.knmi.nl/dataset/rad-nl25-rac-mfbs-em-5min-netcdf4-2-0）及CARROTS校准后的QPE数据（https://doi.org/10.4121/13573814）运行。综上，总计生成9060份独立的临近预报（每份均具备4小时预报时效，包含20个集合成员）。
临近预报以netCDF格式文件存储，文件夹结构如下：事件时长（1小时或24小时）与所用QPE产品（实时、CARROTS或气候学）>城市名称>事件文件夹>单个临近预报文件。事件文件夹的文件名以协调世界时（UTC）标注强降雨事件的起始时刻，临近预报文件的文件名则以UTC标注临近预报的发布时刻。