Prediction of mango hopper, Idioscopus nitidulus (Walker) using hybrid modelling in Konkan region

A Y Munj A Y Munj,V N Jalgaonkar V N Jalgaonkar,B R Salvi B R Salvi,A L Narangalkar A L Narangalkar,Bikash Das Bikash Das,S.s Mali S.S Mali,J S Choudhary J S Choudhary,Madhumita Kumari Madhumita Kumari

doi:10.54386/jam.v23i1.89

Abstract

 Mango hopper, Idioscopus nitidulus is the most destructive pest of mango in the India. Thus, aim of the study was to develop precise and easy early population prediction model of mango hopper for tropical mansoon climate conditions. Weekly occurrence data of mango hopper, I. nitidulus during five consecutive years (2014 to 2018) was used for developing hybrid of multiplicative seasonal autoregressive integrated moving average (SARIMA) and artificial neural network (ANN) model. The population of I. nitidulus increases in the month of January-February on flower panicles and October-November on new vegetative shoots in the region. The linearity in the time series data was best fitted with SARIMA (0, 0, 2) × (0, 1, 1)52 model as their correlation values are not outside the confidence intervals (CI) limits. Further ANN modeling was done for fitting the SARIMA residuals. The fitted values of model prediction and the actual values of year 2017-2018 flowering season (SMW36-52 of 2017 and SMW 1-13 of 2018) were used for testing of prediction efficiency. The performance of the two models in respect to model fitting and effectiveness of SARIMA and hybrid SARIMA-ANN model was compared by evaluating diagnostic statistics of MSE, RMSE, MAE and MAPE. The best fitted developed hybrid model in present study and the data predicted by model was matched with actual data of mango hopper incidence during the year 2017-18. Hybrid model developed in this study will help to predict hoppers population in advance, thus provide a direction for planning of timely prevention and development of effective management strategies which will help to minimize the use of hazardous pesticides.

Highlights

Importance of Mango (Mangifera indica L.) in India can be expressed by its allocation as the king of fruits
The data set of mango hopper incidence recorded at Regional Fruit Research Station (RFRS), Vengurla from January 2014 to December 2018 was used for model fitting (Fig. 1)
The autocorrelation function (ACF) and partial autocorrelation function (PACF) plots of best identified seasonal autoregressive integrated moving average (SARIMA) model are displayed in Fig.2 which showing that correlations fell around zero and within their 95% confidence intervals after one order of differencing

Summary

Introduction

Importance of Mango (Mangifera indica L.) in India can be expressed by its allocation as the king of fruits. Among the major insect pests, mango hoppers (Amritodus atkinsoni Leth, Idioscopus clypealis Leth and Idioscopus nitidulus Walk) are causing serious damage all over India (Rahman et al, 2007). Hoppers lay eggs into the midrib of young leaves at the underside and on inflorescence and even in petioles Both the adult and nymphs feed on vegetative flush by sucking the sap mostly from young leaves, tender shoots, inflorescence and the rachis of young fruits. The excessive sap sucking by number of nymphs and adults may cause severe yield loss up to 100% (Bana et al, 2018) They cause damage to the plant by excreting large amount of “Honeydew”; on which black sooty mould (Capnodium mangiferae) develops giving blackish appearance to all infected plant parts (Munj, 2016). The present study was aimed to develop a hybrid prediction model using time series based seasonal autoregressive integrated moving average (SARIMA) and machine learning based techniques, artificial neural network (ANN) to predict occurrence of mango hopper based on long time series data of mango hopper incidence

Methods

Results

Conclusion