1. ABOUT THE DATASET ------------ Title: Skillful spatial scales and representativity of seasonal rainfall forecasts over Africa Creator(s): Matthew Young [1,2], Viola Heinrich [3], Emily Black [1,2] 1. Department of Meteorology, University of Reading, UK. 2. National Centre for Atmospheric Science, UK. 3. School of Geographical Sciences, University of Bristol, UK. Rights-holder(s): University of Reading, Viola Heinrich Publication Year: 2020 Description: This dataset contains gridded estimates of the skillful spatial scales of seasonal rainfall forecasts (tercile probabilities) over Africa and a measure of how representative these skillful spatial scales are for anticipating local rainfall conditions (as described by terciles). The skillful spatial scales presented apply to tercile probability hindcasts (re-forecasts) of seasonal total rainfall derived from the European Centre for Medium-Range Weather Forecasts (ECMWF) seasonal forecasting System 5 (SEAS5). The SEAS5 hindcasts analysed here are initialised at a lead time of 1 month ahead of the start of a given season. The representativity of the SEAS5 skillful spatial scales is computed using observed seasonal rainfall terciles from the Climate Hazards Group InfraRed Precipitation with Station data (CHIRPS) version 2.0 dataset. The skillful spatial scales and representativity data are combined to delineate regions where the SEAS5 seasonal hindcasts are both skillful and representative for anticipating local seasonal rainfall conditions. Note that the these data are specific to tercile probabilities derived from the SEAS5 hindcasts and different skillful scales and representativity results are likely to be found when applying this methodology to other dynamical weather forecasting systems and event/quantile categories. A subset of this dataset is presented in Figure 6 of Young, M., Heinrich, V., Black, E., and Asfaw, D., 2020: Optimal spatial scales for seasonal forecasts over Africa, Environmental Research Letters. In press https://doi.org/10.1088/1748-9326/ab94e9. Cite as: Young, Matthew, Heinrich, Viola, and Black, Emily (2020): Skillful spatial scales and representativity of seasonal rainfall forecasts over Africa. University of Reading. Dataset. http://dx.doi.org/10.17864/1947.252. Related publication: Young, M., Heinrich, V., Black, E., and Asfaw, D., 2020: Optimal spatial scales for seasonal forecasts over Africa, Environmental Research Letters. In press https://doi.org/10.1088/1748-9326/ab94e9. 2. TERMS OF USE ----------------- This dataset is licensed under a Creative Commons Attribution 4.0 International Licence: https://creativecommons.org/licenses/by/4.0/ 3. PROJECT AND FUNDING INFORMATION ------------ Title: Atmospheric hazard in developing Countries: Risk assessment and Early Warning (ACREW) Dates: March 2017 - March 2021 Funding organisation: Global Challenges Research Fund, Natural Environment Research Council Grant no.: NE/R000034/1 The research leading to this dataset has been supported by the climate division of the National Centre for Atmospheric Science and the Global Challenges Research Fund, via Atmospheric hazard in developing Countries: Risk assessment and Early Warning (ACREW) (NE/R000034/1). 4. CONTENTS ------------ File listing 12 data files are stored in one zipped folder. There is one file containing gridded (0.05 degree) skillful spatial scale and representativity data for each season and tercile category, with the following filename convention: skill_representativity_ecmwf-seas5_lead1month_1994_2016_tercile_MMM_XX.nc Where, - 'MMM' are the first letters of each month in a given 3 month meteorological season. This can be one of DJF (December-January-February), MAM (March-April-May), JJA (June-July-August) or SON (September-October-November). - 'XX' corresponds to the rainfall tercile category. This can can be one of 'BN' (Below Normal; lower tercile), 'NN' (Near-Normal; middle tercile) and 'AN' (Above-Normal; upper tercile). The format of the data is netCDF4 (file extension '.nc'). List of variables: Each file contains 9 variables, as described below: 'latitude': List of latitudes with units: degrees North. 'longitude': List of longitudes with units: degrees East. 'time': singleton time dimension. 'pod_threshold': Probability of detection (POD) thresholds corresponding to dimension 2 for the 'skillful_representative' variable. No units. Values range from 0.5 to 0.95 every 0.05. 'dry_mask' (dimensions: time, latitude, longitude): Dry area mask for seasonal total rainfall computed from CHIRPSv2.0 daily precipitation estimates. No units. Values equal to 1 delineate grid boxes where the climatological mean seasonal rainfall observed by CHIRPSv2.0 is greater than or equal to 100 mm. Values equal to 0 delineate grid boxes where the climatological mean seasonal rainfall is less than 100 mm. 'skillful_scale' (dimensions: time, latitude, longitude): Skillful spatial scale of SEAS5 hindcast probabilities for a given tercile category. Units are in degrees. Values range from 1-10. 'skillful_scale_pod' (dimensions: time, latitude, longitude): Observed (CHIRPSv2.0) probability of detecting (POD) the local (0.05 degree) tercile using the tercile observed by CHIRPSv2.0 at the skillful spatial scale ('skillful_scale') of the SEAS5 hindcasts. 'skillful_scale_roc_area' (dimensions: time, latitude, longitude): The Relative Operating Characteristic (ROC) Area value at the skillful spatial scale. No units. Values range from 0 to 1. 'skillful_representative' (dimensions: time, pod_threshold, latitude, longitude): Skillful & representative. Delineates regions where SEAS5 is both skillful and representative for various representativity (POD) thresholds applied to the 'skillful_scale_pod' field. No units. Values equal to 1 delineate skillful and representative grid-boxes for a given POD threshold, whereas values equal to 0 delineate regions which are skillful but not-representative of local (0.05 degree) scales. Missing data / fillvalue = NaN Data are assigned the missing value over regions where the hindcasts are not skillful at any spatial scale, according to criteria defined in the methodology. 5. METHODS -------------------------- The dataset has been produced using: 1) Hindcasts (re-forecasts) of seasonal total rainfall over Africa from the European Centre for Medium-Range Weather Forecasts (ECMWF) seasonal forecasting System 5 (SEAS5). The hindcasts used are initialised at a lead time of 1 month ahead of the start of the season, cover the period 1993-2016 and have a horizontal resolution of 1 degree. The hindcasts have been obtained from the Copernicus Climate Change Service Climate Data Store at https://cds.climate.copernicus.eu/. 2) Observations of seasonal total rainfall from the Climate Hazards Group InfraRed Precipitation with Station data (CHIRPS) version 2.0 dataset available from https://www.chc.ucsb.edu/data. The CHIRPS data has a horizontal resolution of 0.05 degree and CHIRPS estimates between 1993-2016 are used here to measure representativity. The methodology used to produce this dataset using 1) and 2) is described in detail in Young, M., Heinrich, V., Black, E. and Asfaw, D. T. (2020) Optimal spatial scales for seasonal forecasts over Africa. Environmental Research Letters. In Press https://doi.org/10.1088/1748-9326/ab94e9.