Examples of crop yield estimationforecasting with remote sensing

12.15.1 USDA

The USDA publishes monthly crop production figures for the United States and the world. At NASS, remote sensing information is used for qualitative monitoring of the state of crops but not for quantitative official estimates. NASS research experience has shown that AVHRR-based crop yield estimates in the USA are less precise than existing surveys.1 Current research is centred on the use of MODIS data within biophysical models for yield simulation.

The Foreign Agricultural Service (FAS) regularly assesses global agricultural conditions. Images from geostationary satellites such as GOES, METEOSAT, GMS and from microwave sensors such as SSM/I are used to derive precipitation estimates, as input for models which estimate parameters such as soil moisture, crop stage, and yield. NASA

1 http://www.nass.usda.gov/Surveys/Remotely_Sensed_Data_Crop_Yield/ index. asp and the University of Maryland provide FAS with data from instruments on several satellites: NASA's Aqua and Terra, TOPEX/Poseidon, Jason and Tropical Rainfall Measuring Mission.

NOAA-AVHRR and MODIS images are used to monitor vegetation conditions with bi-weekly vegetation index numbers such as the NDVI (Reynolds, 2001). Other satellite data sources include TM, ETM+ and SPOT-VEGETATION. These are often only visually interpreted. A large imagery archive allows incoming imagery to be compared with that of past weeks or years. When a new image is processed and loaded, key information is automatically calculated and stored in a database. Most of the information is available through the FAS Crop Explorer2 providing easy-to-read crop condition information for most agricultural regions in the world. With these data, producers, traders, researchers, and the public can access weather and satellite information useful for predicting crop production worldwide.

12.15.2 Global Information and Early Warning System

The Global Information and Early Warning System (GIEWS) was established by the FAO in the early 1970s and is the leading source of global information on food production and food security. The system continually receives economic, political and agricultural information from a wide variety of sources (UN organizations, 115 governments, 4 regional organizations and 61 non-governmental organizations). Over the years, a unique database on global, regional, national and subnational food security has been maintained, refined and updated.

In many drought-prone countries, particularly in sub-Saharan Africa, there is a lack of continuous, reliable information on weather and crop conditions. For this reason, GIEWS, in collaboration with FAO's Africa Real Time Environmental Monitoring Information System (ARTEMIS), has established a crop monitoring system using near real-time satellite images. Data received directly by ARTEMIS from the European METEOSAT satellite are used to produce cold cloud duration (CCD) images for Africa every 10 days. These provide a proxy estimate for rainfall. ARTEMIS maintains an archive of CCD images dating back to 1988, which allows GIEWS's analysts to pinpoint areas suffering from low rainfall and drought by comparing images from the current season to those from previous years or the historical average. Similarly, since 1998, the Japan Meteorological Agency has been providing the FAO with 10-day estimated rainfall images for Southeast Asia computed from data received from the Japanese GMS satellite. In addition to rainfall monitoring, GIEWS makes extensive use of NDVI images that provide an indication of the vigour and extent of vegetation cover. These allow GIEWS analysts to monitor crop conditions throughout the season. Data obtained from NOAA satellites are processed by the NASA Goddard Space Flight Center to produce 10-day, 8-kilometre resolution vegetation images of Africa, Latin America and the Caribbean. The FAO, in collaboration with the Joint Research Centre (JRC) of the European Commission, has access to 10-day real-time images from the SPOT-4 VEGETATION instrument. These cover the whole globe at 1-kilometre resolution and are suitable for crop monitoring at subnational level.

2 http://www.pecad.fas.usda.gov/cropexplorer/

12.15.3 Kansas Applied Remote Sensing

The Kansas Applied Remote Sensing (KARS) Program was established in 1972 by NASA and the state of Kansas. In 1998 it became the Great Plains Regional Earth Science Applications Center. This focuses on the assessment of grassland condition and productivity, monitoring and projecting crop production and yield, and monitoring changes in land use and land cover. Yield forecasting is based on statistical modelling of NOAA-AVHRR NDVI data.

12.15.4 MARS crop yield forecasting system

Since 1993 the MARS project of the JRC has been running a crop yield forecasting system for the quantitative assessment of the major European crops in all EU member states. The system has also been used outside the EU since 2000. It is based on:

• low-resolution, high-frequency remote sensing products - NOAA-AVHRR, SPOT VEGETATION and MODIS - received every 10 days with world coverage. An archive of remote sensing data has also been maintained with data for Europe since 1981.

• meteorological information received daily from European Centre for MediumRange Weather Forecasts models and from more than 4000 synoptic stations in Europe and Africa. Other parameters are derived from METEOSAT satellite data. More than 30 years of data have been archived.

• additional geospatial information - soil maps, land cover/land use maps, phenology information and agricultural statistics.

Depending on the region and on data availability, different crop monitoring procedures are implemented. During the growing season, crop qualitative assessments are produced looking at anomalies between crop development profiles of the current year and historical profiles. Before the end of the season, for some countries, quantitative crop production estimates are computed using a simple NDVI regression model or a multiple regression with water satisfaction indices (Nieuwenhuis et al., 2006) and NDVI. The model is calibrated with historical agricultural statistics. In addition to the remote sensing approaches, the MARS Crop Yield Forecasting System uses several simulation crop growth models: WOFOST, LINGRA and WARM. Results are published in monthly or bimonthly climatic and crop monitoring bulletins for Europe with quantitative yield forecasts by country and monthly national and regional crop monitoring bulletins for food-insecure regions including qualitative and, for some countries, quantitative estimates (http://mars.jrc.it/mars/).

0 0

Post a comment