Development of Statistical Models for Horticultural Crops Research

Response surface models (RSM) for maximizing input-use efficiency in Perennial crops : RSM were constructed using SORD for working out the optimum doses of Evaporation Replenishment (ER,%), Recommended dose of fertilizer (RDF,%) and the corresponding optimum yield of Mango/Passion Fruit/Acid lime/Annatto using three years experimental data

ANN models to express the nonlinear relation among climatic factors vis-à-vis thrips incidence in Capsicum :Artificial Neural Network models (MLP)  were developed to optimize the role of several weather factors on thrips incidence in Capsicum individually for plot data pertaining to high thrips rating (exceeding 3.0) and rating less than 1.0.

A quick method for estimation of leaf area in Fig using non-destructive sampling approach:ANN theory was utilized to develop a statistical model thereby capturing the intrinsic nonlinearity among  leaf parameters and facilitating quick estimation of leaf area in Fig (Cv Deanna and Poona)

Crop-logging models in Banana (CV Granaine):Biometrical models for identifying significant crop-logging parameters (along with their point estimates) across different growth stages of G9 banana crop were developed, having 71 to 83.4% power of predicting banana (G9) crop yield.

Non-parametric index for assessing crop yield stability in vegetable crop improvement program :A combined index was developed based on various yield related characters to suggest a rank based non-parametric measure for assessing crop yield stability in cucumber which may be more practically meaningful to come out with stable lines either for release as variety or for use in crop hybridization trails.

Artificial neural network models in Papaya (Cv Surya):Artificial neural network models capturing inherent nonlinearity among biometrical traits were developed (R² 70 -84%) to identify factors influencing Papaya (Surya) crop yield at vegetative/pre-flowering/fruiting stages, enabling selection/development of markers at early stage.

Bootstrap method to evaluate the sampling efficiency and arrive at optimum sample size for  leaf area estimations in Fig by non-destructive sampling:A bootstrap resampling method was suggested to workout the optimum sample size required for developing leaf area estimation models is Fig using non-destructive sampling approach

Method for revalidation of canopy architecture model in Mango:A Jackknife method for revalidation of canopy architecture models in Mango was suggested with the lowest RMS error (5.16 %) having the optimal number of components to predict the crop yield.