MSc

Rice is the staple food and the major field crop cultivated in Kerala. Its production is highly influenced by unfavourable weather events and climatic conditions. Thus it poses a challenge to farmers, crop planners and government owing to varying production of grains. Reliable crop yield forecasts are highly essential to estimate crop production, to assist farmers, exporters and government in decision making for efficient resource allocation, price adjustment and export planning. It also helps to reduce various secondary risks associated with local and national food systems.
The present investigation “Comparison of different weather based models for forecasting rice yield in central zone of Kerala” was carried out at the Department of Agricultural Meteorology, College of Horticulture, Vellanikkara during 2017-18, to compare the accuracy of different weather based models developed using five years’ rice crop data collected from previous studies at the department for forecasting rice yields in central zone of Kerala and to validate them using the present experimental data. The field experiment was conducted at Agricultural Research Station, Mannuthy during the kharif season of 2017. Split plot design was adopted with five dates of planting viz., 5th June, 20th June, 5th July, 20th July and 5th August as the main plot treatments and two varieties viz., Jyothi and Kanchana as the sub plot treatments. The number of replications for the experiment was four.
Daily observations of weather during the crop period were made which showed an increase in the maximum and minimum temperature and decrease in rainfall and relative humidity towards the end of the crop period. Different growth and yield attributes like plant height, dry matter accumulation, number of tillers, panicles, spikelets, filled grains, grain yield, straw yield and the duration of different phenophases were also noted. Correlation analysis was carried out using the weather, yield and phenological data of 5 years in both the varieties. The various growth indices such as leaf area index, net assimilation rate, leaf area duration and crop growth rate were worked out to analyze the growth and development of the crop.
Plant height was found to be higher for Jyothi compared to Kanchana. Dry matter accumulation, yield attributes except straw yield were found varying between five dates of planting. Yield and yield attributes were influenced by different weather parameters during different dates of planting. With delay in dates of planting the duration of different phenological stages were reduced in both the varieties. Jyothi took more number of days to attain different growth stages compared to Kanchana. The highest yield in Jyothi and Kanchana were obtained for June 5th planting.
Crop weather models using statistical techniques were developed using five years’ weather and crop yield data by adopting four different methods for Jyothi and Kanchana separately. The methods were (i) based on weekly weather variables (ii)based on fortnightly weather variables (iii) based on crop stage wise weather variables and (iv) based on composite weather parameters. Each crop weather model was fitted by stepwise regression analysis using SPSS software. CERES-Rice model also was run for Jyothi and Kanchana by creating weather file, soil file, crop management file and experimental files separately for each year.
For comparing the accuracy of the developed crop weather models and simulation model for Jyothi and Kanchana, and for their validation, mean absolute percentage error (MAPE) was calculated for each model using the observed and estimated yield data. The model with least mean absolute percentage error (MAPE) is considered as a better model for yield prediction. In the case of Jyothi, lowest MAPE (4.00%) was obtained for model based on 5 fortnightly weather variables. In Kanchana also, the model developed using 5 fortnightly weather variables was selected with an MAPE value 7.62%.
All the crop weather models are showing very good results out of which crop weather model using 5 fortnightly weather variables which coincide with flowering stage has given a good forecast compared to the other models for both Jyothi and Kanchana.