Abstract:
An effective pre-harvest rice yield estimation method is very important for assessment
of seasonal rice production for strategic planning purposes. In Sri Lanka, a conventional
method is used to estimate seasonal rice production and it fails to forecast rice yield
before harvest because the experiment is conducted during the harvest. This study was
focused to identify cultivated paddy lands and forecast rice yield using free satellite
data. 8-day composite images (250m spatial resolution) from Moderate Resolution
Imaging Spectroradiometer (MODIS) sensor onboard the NASA EOS Terra satellite were
used from 2007 to 2014. In this study, a new method has been suggested to identify
cultivated paddy lands by analyzing temporal dynamics of the paddy cultivation. Then
linear and exponential yield forecasting models were built at different times of the
season based on NDVI and EVI2 vegetation indices. Accuracy assessment results show
that the suggested cultivated paddy lands identification method has ability to identify
cultivated paddy lands with 74% average accuracy. According to the comparison
between estimated yield and national statistical data, both NDVI and EVI2 based models
give more reliable estimations about 96 days after beginning time of the season. But,
EVI2 based model (derived at 96 days) give more reliable estimations than NDVI based
model with 92% average accuracy. Therefore seasonal rice yield can successfully be
forecast before one month to harvest time using EVI2 based model. Improvement of
this study will help to forecast national level yield estimations in the country by
illuminating problems faced with current system.
