Rice yield prediction with agro-spectral model

Abstract

Rice yield prediction with agro-spectral model

B. Sreedevi
Scientist(SS) Agronomy, Directorate of Rice Research,
Rajendranagar, Hyderabad -500 030, India.
Email: sreepala_bsd@rediffmail.com

L. Venkataratnam
Group Director, Agriculture & Soils, National Remote Sensing Agency
Balanagar, Hyderabad- 500 037, India.
Email: venkataratnam_l@nrsa.gov.in

Abstract
Pre-harvest estimates of crop production is indispensable for national food security including policy determinations of import/export plans and prices. Eventhough Rice (Oryza sativa L ) is the major staple food crop of the world in general and Asian countries in particular, the number of remote sensing investigations are meager, when compared to the number of investigations on wheat and maize, due to, much higher diversity in growing conditions as well as yield, sensitivity of spectral indices to water background etc.

Yield prediction models based on meteorological variables and agronomic parameters have been extensively used till recently. The use of satellite based multi-date multi-band spectral data in crop discrimination, crop area estimation and crop stress monitoring has proved to be a great success and attempts have been also made to employ remote sensing data for yield estimation of field crops. Spectral vegetation indices derived from ground based or satellite sensors are used to predict crop yields. The spectral data being the manifestation of integrated effects of weather, soil productivity and cultural practices are expected to be good indicators of crop growth and yield. With this background the present study was carried out to predict the rice yield by combining agronomic parameters and spectral vegetation indices.

Field experiments were conducted in wet season for two years with 8 different varieties of medium duration, under irrigated conditions. Spectral measurements and biometrics observations were taken at an interval of 10 days from 10 days after transplanting till harvest. IR/R ratio and Normalized Difference Vegetation Index were computed for each observation stage. The agronomic variables are related to spectral parameters at maximum vegetative growth stage. Simple correlation was worked out between the spectral indices and biometrics parameters with rice yield. From the correlation studies, the factors which gave highest R value was selected and used for developing multiple regression equation. The resultant equation did have the highest R value and was chosen for predicting rice yields. This model is useful to standardize the prediction models before yield forecasting systems based on satellite data.