Improving Crop Yield Forecasting in Uruguay using Landsat Imagery
Authors: Sunny Ng, Pietro Ceccato
Mentors/Advisers (affiliation): Pietro Ceccato (International Research Institute for Climate and Society, IRI)
Team Location: IRI in Palisades, New York
Abstract: Using field observations of crops in Uruguay, Landsat images were classified into 95 classes for the 2011 and 2012 planting seasons. The SIAM (Satellite Image Automatic Mapper) algorithm classified images based on spectral properties instead of through trained pixels in a supervised classification method. The pixel counts for different crops such as maize and soybean were analyzed in a matrix to measure the level of accuracy of SIAM and plotted on a time series to visualize how pixels evolve over the planting season. If the results show a clear trend, this will improve crop-yield predictions at early stages in the planting season, which will be beneficial to Uruguay’s Ministry of Agriculture.
What is the community concern for early crop yield predictions? Supposing SIAM was able to detect a poor yield, what could be done to protect Uruguay's Ministry of Agriculture?
Would it be more beneficial to create maize and soybean yield predictions from imagery with a higher spatial or spectral resolution than the data your team utilized?
@amadson Improving spatial resolution and spectral resolution would certainly help. For example, using hyperspectral sensors may help us determine differences in phenology for different crop types.