ASAP Hotspots

Monthly identification of agricultural production hotspot countries and summary narratives for agriculture and food security analysts. The historical hotspot time series is available in .csv format. To visualize the information in a GIS environment, use the field asap0_id and join with the spatial layer gaul0_asap (see section Administrative Boundaries). The latest ASAP hotspot layer can also be downloaded as shapefile directly from the ASAP homepage and is available as WFS/WMS

hotspots_ts.zip  
Historical data set in .csv format with the time series of all ASAP hotspot assessments since October 2016 (to date)

ASAP Warnings

Ten-daily automatic warnings about low or delayed vegetation performance at the sub-national level, both for crop and rangelands. The historical time series is available in .csv format. To visualize the information in a GIS environment, use the field asap1_id and join with the spatial layer gaul1_asap (see section Administrative Boundaries). The latest ASAP warning layer can also be downloaded as a shapefile directly from the ASAP Warning Explorer page and is available as WFS/WMS

warnings_ts.zip  
Historical data set in .csv format with the complete time series of ASAP warnings (crop and rangeland) at sub-national level since March 2003 (to date)

HOW TO CITE: Rembold, Felix (2018): Ten-daily automatic warnings about agricultural vegetation anomalies. European Commission, Joint Research Centre (JRC) [Dataset] PID: http://data.europa.eu/89h/jrc-10112-10001

WFS and WMS

The latest ASAP warnings and hotspots can be visualized directly in a GIS desktop environment (e.g. QGIS or even, with some limitations for WFS, ESRI ArcGIS) as WFS and WMS, using the following URL:

https://agricultural-production-hotspots.ec.europa.eu/public/ows?

wfs_hotspots_notes.txt  
Description of the fields in the WFS layer HOTSPOTS

wfs_warnings_crop_notes.txt  
Description of the fields in the WFS layer WARNING_CROP

wfs_warnings_rangeland_notes.txt  
Description of the fields in the WFS layer WARNING_RANGELAND

Historical ASAP warning frequency

The 2 maps show the historical frequency of ASAP warnings of level 2 or higher at the sub-national level and can be interpreted as sensitivity to "agricultural drought conditions". The color intensity is proportional to the frequency of ASAP warnings, i.e. the percentage of dekads with a warning, out of the total number of active dekads in the period 2003-2022.

frequency_crop_v06.png  
Frequency of crop warnings

frequency_rangeland_v06.png  
Frequency of rangeland warnings

Administrative Boundaries

ASAP information is provided at two spatial levels: national (Gaul0) and sub-national (Gaul1). In particular, hotspots are related to national units while warnings to sub-national units. An additional more detailed sub-national level (Gaul2) is available for a set of selected countries of interest, for which we provide aggregated statistics (see "Indicator Statistics" sub section). The reference spatial layers (gaul0_asap, gaul1_asap, gaul2_asap) are derived from the The Global Administrative Unit Layers (GAUL) dataset, implemented by FAO within the CountrySTAT and Agricultural Market Information System (AMIS) projects. The original data sets were adapted to match the specific ASAP requirements (e.g. in some countries, the sub-national units were aggregated, in others they were split in smaller units) and the geometry was simplified (e.g. the number of vertex reduced and small islands removed). The geographic and thematic information contained does not imply a policy position of the European Commission. Here you can download the three layers as shapefiles. These layers can be used to visualize the historical time series of Hotspots, Warnings and Indicator statistics (see related sections below). The key fields are asap0_id, asap1_id and asap2_id. Please note that these layers are intended for visualization at a broad scale and due to the simplification of the geometries, the original topology is not always preserved.

gaul0_asap_v04.zip  
ASAP national units (shapefile)

gaul1_asap_v04.zip  
ASAP sub-national units - level 1 (shapefile)

gaul2_asap_v04.zip  
ASAP sub-national units - level 2 (shapefile)

Land Cover Masks

Global crop and rangeland masks at 0.0089286 deg resolution (about 1 square kilometer). Each pixel represents the area fraction of the specific cover (i.e. percentage of the pixel with crops/rangeland). Data are scaled between 1 and 200 (50 = 25%, 100 = 50%, 150 = 75%, 200 = 100%): image values V = 0-200, scaling 0.5 - > physical value 0-100%. These layers were generated for ASAP, combining existing data sets (Perez-Hoyos, A., Rembold, F., Kerdiles, H., Gallego, J., 2017b. Comparison of global land cover datasets for cropland monitoring. Remote Sens. 9. doi:10.3390/rs9111118). The data sets were last updated on 12-03-2019 (version 03).

asap_mask_crop_v03.tif  
ASAP crop mask (area fraction)

asap_mask_rangeland_v03.tif  
ASAP rangeland mask (area fraction)

asap_mask_crop_v04.tif  
New ASAP 500m crop mask (area fraction), still not operational

asap_mask_rangeland_v04.tif  
New ASAP 500m rangeland mask (area fraction), still not operational

Phenology

The mean growing season period is defined by a satellite-derived phenology computed on the long-term average of 10-day MODIS NDVI data. The MODIS data preprocessing was done by BOKU University (Klisch and Atzberger, 2016) starting from MOD13A2 and MYD13A2 V006 16-day Global data at 1 km resolution. Phenology was extracted using the SPIRITS software (Eerens et al., 2014; Rembold et al., 2015) applied to the historical average of the smoothed NDVI over the period 2003-2016.
The following key parameters are retrieved for each pixel: number of growing seasons per year (i.e. one or two); start of season (SOS, occurring at the time at which NDVI reaches 25% of the ascending amplitude of the seasonal profile); time of maximum NDVI (TOM); start of senescence period (SEN, when NDVI drops below 75% of the descending amplitude); and end of the season (EOS, when NDVI drops below 35%).
Phenology timings: the value V in the image is expressed over a 3 years period, from dekads (i.e. ten-day periods) 1 to 108. This is done in order to treat the circular variable "dekad" as an ordinal variable (the mean of SOS cam be computed as the usual average). Example for a season crossing the calendar year:
phenos1 = 30, last dekad of October, phenoe1 = 46, first dekad of April -> T = 46 - 36 = 10
Therefore, to get the timing T in the 1-36 dekad units (calendar year), one can use:
IF V <= 36 THEN T = V
IF 36 < V <= 72 THEN T = V -36
IF V > 72 THEN T = V - 72

phenos1_v03.tif  
Phenology: start of the season, first season

phenos2_v03.tif  
Phenology: start of the season, second season

phenom1_v03.tif  
Phenology: max of the season, first season

phenom2_v03.tif  
Phenology: max of the season, second season

phenoe1_v03.tif  
Phenology: end of the season, first season

phenoe2_v03.tif  
Phenology: end of the season, second season

phenosen1_v03.tif  
Phenology: start of senescence, first season

phenosen2_v03.tif  
Phenology: start of senescence, second season

phenonseasons_v03.tif  
Phenology: number of seasons (1 or 2)

Crop Calendars

The national and sub-national crop calendars are based on FAO data and include only crops for which the original calendars match with the ASAP remote sensing based phenology. The file includes the start and end dekad (over the 36 dekads in a year) of the planting period (sos_s and sos_e), the start and end dekad of the growth period (sos_e and eos_s), and the start and end dekad of the harvest period (eos_s and eos_e). The rest of the year is considered out of season. To visualize the information in a GIS environment, use the field asap1_id and join with the spatial layer gaul0_asap (see section Administrative Boundaries).

crop_calendar_gaul0.zip  
Crop calendar at national level

crop_calendar_gaul1.zip  
Crop calendar at national level

Pre-operational Country-level Yield Forecasting with Machine Learning Algorithms and Agrometeorological Data

Country-level yield forecasts are computed for 77 food insecure countries using machine learning algorithms and agrometeorological data. The system is currently in testing phase and utilizes over 20 years of FAOSTAT yield data and ASAP agrometeorological data to provide in-season prediction of crop yields. Remote sensing and meteorological data are used to extract important information about crop development, including maximum NDVI, average temperature, and the sum of radiation and rainfall. Similar countries are grouped together based on their crop development characteristics to improve the accuracy of the models. Finally, only models that achieved Rcv2 > 0.3 were considered suitable for operational forecasts and retained. If no valid model is retained for a given crop-country, no forecast is available.

Forecast model outputs are provided as pdf and csv files and they report: The forecasted yield for the current season, average yield based on the last five years and the percentage difference between the forecasted and the 5 years average yield, input feature combination (RSTR or TR) and the best pooling approach (All, Main or Individual), root mean square error (RMSE) and the coefficient of determination (R2) in prediction from the hindcasting experiments. PDF files also contain plots of observed and predicted yield values (since 2006) as well as the benchmark model (average of last 5 years). Additional information on the methodology can be found in the Methodology summary.