Geospatial Data and Weather Risk Mapping for Agriculture Risk Management
Dr. Nathan Torbick - Applied Geosolutions
FARMD (March 2012) | Many regions experience shocks of severe weather variability and inter-seasonal volatility that affect cash crops, food security, and consequently livelihoods. Drought, flooding, temperature extremes, changes in disease range, and shifting timing of rain patterns are a few of the risks agricultural producers face across the globe. In the developing world these stressors can often cause significant hardships due to the strong socioeconomic dependency on agricultural production. At the same time threats from weather variability are predicted to become more extreme (higher impact and greater frequency) due to climate change. To address agricultural weather risk we need more targeted and geospatial decision support tools at appropriate scales for risk assessment, adaptation, and mitigation planning. The quality of these tools will largely be influenced by the availability and resolution of geospatial data.
Geospatial data is information about a particular location that is organized to understand how locational attributes -such as temperature, crop type, population density, and distance to city- vary in space. For example, where are rice paddy fields and potential irrigation sources? These spatial data can be compiled over time to enable the assessment of trends. For example, what is the annual rate of change in population density and commodity prices for a set of districts for the past decade? Geospatial data can cover a broad spectrum of locational attributes that include administrative, environmental, economic, transportation, and many other thematic categories, such as weather and agriculture. Data is generated from surveys, local knowledge, mapping software, Global Positioning Systems (GPS), aerial photography and satellite imagery, station instruments, and other sources. Ultimately, geospatial data lets people understand how different factors and attributes relate over space and time.
The growth in geospatial data available for weather risk mapping and agricultural risk management is growing rapidly. Satellite remote sensing of rainfall, temperature, land surface, and atmospheric conditions provide near real-time data and contribute to data archives for historical trend and probabilistic analysis. Satellite information is particularly useful in data-poor regions where weather stations are spatially sparse with temporal gaps of missing data. Gridded weather products and regional climate models provide short-, mid-, and long-term predictions with uncertainty metrics at spatiotemporal scales that enable the assessment of crop growth and yield. Computer models for simulating crop production are evolving and ease of use continues to improve. The accessibility of data required for weather risk mapping - such as soils, climate, and farm management- grows largely due to the digital revolution and adaptation of Geographic Information Systems (GIS). GIS has become a popular tool used to store, organize, visualize, and analyze geospatial data in digital formats. Geostatistical approaches for linking these databases continues to progress with open source data and tools promoting their use. Recent focus by the international community on visualization and delivery mechanisms is enhancing access with web and mobile technologies that remove traditional barriers to utilizing this information to make informed decisions.
Methods for weather risk mapping and crop risk management are also evolving with the growth of geospatial data. Harnessing these growing geospatial data platforms requires careful consideration, multidisciplinary expertise, and, ultimately, financial support. The development of these methods involves expertise from the fields of geography, finance, climatology, computer modeling, agronomy, policy, outreach, land ownership, among others. Interdisciplinary efforts that interlink all these data and expertise are helping to advance agricultural risk management. Weather risk metrics, crop insurance, agroclimate zonal mapping, crop suitability planning, and vulnerability assessment are a few of the specialized thematic applications that are improving in large part due to the growth of geospatial data. A good illustration of using geospatial data for weather risk mapping and crop risk management recently took place in Mozambique. The agricultural sector in Mozambique is dominated by 3.2 million smallholder families, the majority of whom grow food crops. A smaller number (16%) also participate in cotton and tobacco out-grower schemes. Food and cash crops are the primary support for livelihoods, especially in rural regions where rural smallholders provide about 95% of agricultural Gross Domestic Product (GDP). The growth of the agricultural sector has helped reduce poverty and build rural infrastructure from both food and cash crops. However there remains significant opportunity to enhance efficiency as only 10% of Mozambique’s 36 million arable hectares are currently cultivated and technologies and organization are lacking (i.e., little irrigation, low productivity). However, weather variability presents a tremendous threat to Mozambique’s agricultural sector and little has been done to address risk or establish a risk management framework.
"Satellite information is particularly useful in data-poor regions where weather stations are spatially sparse with temporal gaps of missing data."
Working under the umbrella of the World Bank and Government of Mozambique, we developed an agricultural weather risk mapping tool for Mozambique. The tool is designed to be easily scalable, transferable, and, thanks to readily available geospatial data, can be operationalized. To begin to address weather risks and develop adaptation and mitigation strategies, such as crop insurance, fundamental information on homogenous agroweather zones is required. To do this we utilized rainfall and temperature data from satellite sensors (TRMM, MODIS) at multiple scales over the past decade. These climate products were combined with soil (HWSD) attributes (e.g., soil texture, organic carbon, moisture) and elevation (SRTM) derivatives (e.g., slope, aspect, profile convexity). Principle Components Analysis, diagnostic geostatistics, and K-means clustering were then applied to map homogenous agroclimate zones across Mozambique. The next step was to assess crop suitability across the unique agroclimate zones. We used the Denitrification-Decomposition (DNDC) agricultural model to quantify crop susceptibility to weather conditions and scenarios such as drought.
Using climate, soils, and management data (enhanced by field work and local expertise), crop suitability was modeled and mapped for all major crops including beans, cassava, cotton, groundnut, maize, millet, potato, irrigated rice, rainfed rice, sorghum, and tobacco. We then developed a Crop Suitability Index (CSI) to show which crops grow effectively in different regions based on the geospatial data. For example, maize was found to have relatively low suitability in the semi-arid and arid south whether using conventional or optimized fertilization schemes.
A vulnerability and risk assessment was then carried out by simulating a range of selected climate scenarios. These scenarios were taken from trends found in the geospatial data, regional climate models, and local expertise. A second product, the Crop Vulnerability Index (CVI), highlights changes in crop growth and yield due to different temperature and precipitation patterns relative to baseline weather data as observed by satellites. The CVI can be used to determine what crops face the most risk under different weather conditions in different regions of Mozambique. Some examples of the findings from this work include:
- Gaza, Inhambane, southern Manica, and southern Sofala (the south-central relatively arid region of the country) tend to be the most vulnerable with yields decreasing between 15-50% with a 2.5° C temperature increase and a 10% reduction accumulated rainfall relative to baseline for the top four crops by area.
- Key commercial cotton and tobacco regions were shown to be vulnerable assuming no shifts in management strategies.
The coastal regions were found to become vulnerable with reductions in total accumulated precipitation, timing, or intensity of precipitation.
The resultant activities and programs arising from this work are very positive. Information generated from the results of this project is now being linked with efforts to build a more comprehensive framework for addressing agricultural weather vulnerability. With this foundation built, crop insurance planning is also underway in the region to combat weather risks, crop failure, and cycles of poverty. In addition, data products can be designed to be made available cheaply and quickly on open source web-GIS platforms to enable easy access and decision making. Whereas we have highlighted recent work carried out for Mozambique for mere illustrative purposes, geospatial data and agricultural weather risk mapping is a field that is evolving very rapidly. Exciting developments show promise to bring an easy accessible complexity of knowledge for applications in risk management for the agricultural sector in developing economies.