GEOSCAN Search Results: Fastlink


TitleCrop yield forecasting on the Canadian Prairies using MODIS NDVI data
AuthorMkhabela, M S; Bullock, P; Raj, S; Wang, SORCID logo; Yang, Y
SourceAgricultural and Forest Meteorology vol. 151, no. 3, 2011 p. 385-393,
Alt SeriesNatural Resources Canada, Contribution Series 20181620
PublisherElsevier BV
Mediapaper; on-line; digital
File formatpdf
Subjectsgeophysics; remote sensing
ProgramCanada Centre for Remote Sensing Divsion
AbstractAlthough Normalised Difference Vegetation Index (NDVI) data derived from the advanced very high resolution radiometer (AVHRR) sensor have been extensively used to assess crop condition and yield on the Canadian Prairies and elsewhere, NDVI data derived from the new moderate resolution imaging spectroradiometer (MODIS) sensor have so far not been used for crop yield prediction on the Canadian Prairies. Therefore, the objective of this study was to evaluate the possibility of using MODIS-NDVI to forecast crop yield on the Canadian Prairies and also to identify the best time for making a reliable crop yield forecast. Growing season (May-August) MODIS 10-day composite NDVI data for the years 2000-2006 were obtained from the Canada Centre for Remote Sensing (CCRS). Crop yield data (i.e., barley, canola, field peas and spring wheat) for each Census Agricultural Region (CAR) were obtained from Statistics Canada. Correlation and regression analyses were performed using 10-day composite NDVI and running average NDVI for 2, 3 and 4 dekads with the highest correlation coefficients (r) as the independent variables and crop grain yield as the dependent variable. To test the robustness and the ability of the generated regression models to forecast crops grain yield, one year at a time was removed and new regression models were developed, which were then used to predict the grain yield for the missing year. Results showed that MODIS-NDVI data can be used effectively to predict crop yield on the Canadian Prairies. Depending on the agro-climatic zone, the power function models developed for each crop accounted for 48 to 90%, 32 to 82%, 53 to 89% and 47 to 80% of the grain yield variability for barley, canola, field peas and spring wheat, respectively, with the best prediction in the semi-arid zone. Overall (54 out of 84), the % difference of the predicted from the actual grain yield was within ±10%. On the whole, RMSE values ranged from 150 to 654, 108 to 475, 204 to 677 and 104 to 714kgha-1 for barley, canola, field peas and spring wheat, respectively. When expressed as percentages of actual yield, the RMSE values ranged from 8 to 25% for barley, 10 to 58% for canola, 10 to 38% for field peas and 6 to 34% for spring wheat. The MAE values followed a similar trend but were slightly lower than the RMSE values. For all the crops, the best time for making grain yield predictions was found to be from the third dekad of June through the third dekad of July in the sub-humid zone and from the first dekad of July through the first dekad of August in both the semi-arid and arid zones. This means that accurate crop grain yield forecasts using the developed regression models can be made one to two months before harvest.

Date modified: