Wheat yield prediction using CERES-wheat V4.6 an operational approach for Uttar Pradesh, India

Abstract

CERES-Wheat model embedded in DSSAT v4.6 was used to simulate the yield under irrigated condition for the eight locations representing different agroclimatic zones of Uttar Pradesh. The genetic coefficient required for running the CERES-Wheat v4.6 model was derived for commonly grown cultivars i.e. PBW-343 (22nd Nov, 29th Nov and 06th Dec) and Malviya-234 (13th Dec, 20th Dec and 27th Dec) at different agroclimatic zones. The simulated average yield is 3430 for F2 (pre flowering stage) and 3560 kg ha-1 for F3 (pre maturity stage) stage during 2014-15. The simulated yield production forecast during 2014-15 depicts increase grain yield by-9.3% more compare to year 2013-14. The maximum yield (F2=4824 & F3=4846 kg ha-1) in Western agroclimatic zone (Modipuram) followed by lower yield (F2=2743 & F3=2893 kg ha-1) in North eastern plain zone (Bahriach) in 2014-15. The validation of the simulated yield is highest correlation (R2=0.98 and RMSE value 384, 372, 367, 355, 271, 269 and 196 kg ha-1 respectively) for Kanpur, Jhansi, Lucknow, Varanasi, Baharich, Modipuram and Faizabad, followed by Allahabad (R2=0.93 and RMSE 110 kg ha-1). The validation results indicate that the hind cast wheat yields for most of the districts are within the acceptable error limit (±10%) in all the years of validation; however, prediction was marginally higher in the year 2011-12 for Central Plain zone (Kanpur), 2012-13 for Buldelkhand Zone (Jhansi) and 2013-14 for Eastern Plain zone (Faizabad). The simulated yields are clearly able to indicate the yield variability due to rise of temperature encounter during booting or grain filling state in the wheat crop. The model validated for previous year grain yield showed deviation 9 to 10% which can be improved by further fine tuning the genetic coefficient and other input data at district level.