|Paper title||Global to local risk assessment for agricultural commodities by integrating different sources of climate, weather and satellite data into an operational web service|
|Form of presentation||Poster|
With the rising awareness and visibility of impacts caused by climate change and linked extreme weather events, the need for rapid dissemination of and access to information is becoming a progressively pressing matter in many different anthropogenic, social and economic sectors. To meet these needs, the existing wealth of free, open and globally available analysis-ready weather and climate data serves as a valuable source.
However, the lack of understanding how to access, handle and combine data sets from different sources often prevents end-users in the previously mentioned sectors from making use of the data. Furthermore, domain-specific knowledge to extract additional information out of climate and weather data is hardly existing.
Because of the global interconnection of supply-and-demand chains, the food commodity sector is one of the most vulnerable economic sectors to the effects of extreme weather events. The timely identification of abnormal weather and weather risks is a key point in guaranteeing stable supplies by pointing out geographic areas under risk. This risk assessment is directly supporting the accomplishment of United Nation’s sustainable development goal (SDG) 2, particularily by backing the achievement of food security. Realizing the latter has been the main focus during the development of our web application, through supporting stakeholders in the food commodity trading sector in planning advance purchases of supplies. An early planning of purchasing volumes is necessary to prevent disruptions in supply chains and sudden price increases for consumers of final goods.
Over the course of the past years, green spin has been developing, in close exchange with users in the food commodity industry, a web-based application in which data from the Copernicus Climate Change Service (C3S), Copernicus Land Monitoring Service (CLMS), the German Weather Service (“Deutscher Wetterdienst”, DWD), National Oceanic and Atmospheric Administration (NOAA), Global Inventory Monitoring and Modeling System (GIMMS), MODIS and SMAP satellites have been combined to not only provide access to the data but also to extract information to support decision making processes.
Based on the extracted user needs, the systemic knowledge of crop cultivation cycles (mainly wheat, corn and rice) and constant evaluations during the development phase, it has been found that the following parameters contain useful information:
1) Parameters with daily temporal resolution: precipitation (DWD), temperature (DWD), soil water index (CLMS), leaf area index (MODIS), vegetation health index (NOAA), NDVI anomalies (GIMMS), snow cover extent (MODIS) and snow mass (SMAP)
2) Parameters with monthly temporal resolution: temperature 3 months forecasts (C3S), precipitation 3 months forecasts (C3S)
The whole processing pipeline is fully automated and includes downloading, conversion, cleaning of data errors and data extraction. Prepared data are then stored in data bases, checked for integrity and completeness and can be accessed via APIs. So far, data since the year 2000 have been integrated (with exception of soil water index which is only available since 2007). All daily input data are aggregated on administrative levels, ranging from district level (corresponding to “Kreise” in Germany) up to country level, and integrated as interactive maps into the web application. Parameters with a monthly resolution are displayed as continuous vector maps, since they are mostly used as approximation for a quick global assessment of potential medium-range climate developments. This extensive framework has been operational for two years and is continuously evaluated regarding the inclusion of new data.
In addition to data visualization via interactive maps, the data was further used as input for a weather-based risk indicator and the modelling of production, yield and area of specific crops. It was necessary to integrate those additional parameters in order to make the transition from a mere data visualization application to an actively used application in which EO climate and weather data and derived products truly serve as a basis for decision-making for stakeholders in non-EO disciplines.
Other existing solutions like GADAS or geoGLAM provide very good overviews on a number of different parameters and data sets, but are oftentimes difficult to use and interpret due to a high level of complexity. For example, the better part of such portals doesn’t provide analysis tools with which the plethora of displayed parameters and indices can be inter-compared in time as well as in space by the end-user.
Therefore, we developed our web application as a “collaboration framework” with the goal of enabling non-EO users to access EO data and derived information through a solution-oriented approach. On the one hand, our approach does leave out the raw raster data which can be seen as information loss. But on the other hand, the resulting gain in simplicity of interpretation enabling the simplified derivation of information caused by the means of data condensation (spatial aggregation) largely outweighs the perceived loss of information. This approach is based on constant exchange with users, leading to constant adjustments in the application.
One example of a simplified derivation of information is the above-mentioned weather-based risk indicator. It is used to spot “risk areas” on a sub-national scale (first level below country level). The aim of these risk areas is to identify regions in the world where crop areas are under risk due to extreme weather events. Therefore, different existing algorithms have been analyzed to find a representative index which not only provides more information for the risk assessment task but also can be interpreted and understood correctly by non-expert end-users.
The calculation of the risk indicator is based on the computation of the Standard Precipitation Index (SPI) from the National Drought Mitigation Center. Instead of using only precipitation data, additionally temperature and soil water index data were used. Thereby, an index was created which is especially adapted to and targeted on plant growth. Crop production and area statistics are used in order to grade the severeness of the detected risk in an area. The more crop production proportionally present in an area, the higher the severity.
Compared to other existing data portals and applications, the following novel features have been introduced:
• Information is made available for countries and lower administrative units for which data is typically not available (e.g., provincial data for Russia and China)
• Possibility to not only compare time series of different parameters among each other but also to put them directly in relation to crop harvest quantities, enabling the merge of the subjective knowledge of end-users with objective data
• Detection and monitoring of extreme weather events, including their impact on the development of crop production
In conclusion, with this web application we show:
• how EO data can be made accessible for a range of applications (EO- and non-EO related)
• how new parameters from already used or new data providers can be flexibly integrated into the operational data processing pipeline and visualization
• a “best practice” approach how to condense data into useful information with a focus on facilitating comprehension for decision makers from non-EO fields
As an outlook, we are testing the expansion of the presented risk indicator with the integration of temperature and precipitation forecasts as well as population data as an additional measure for assessing severity. These modifications are intended to make the risk indicator more by taking into account the differences between the global (supply chain management) and the local ("direct-to-food" production) context of application.