Forecasting of pest induced yield loss in rice using Damped Trend Exponential Smoothing model for Balesore District of Odisha

Introduction

In India, majority of the population consumes rice as a staple food and rice is the primary agricultural commodity. It is the farming fraternity's oldest source of income. The production of rice needs to be increased to meet the demand of growing population. Earlier agriculture production was carried out in large cultivated areas, whereas now-a-days, urbanization diminishes the cropping area as a result production becomes less. Moreover, disengagement of the farmer fraternity from doing agriculture for several reasons like environmental calamities (drought, flood, etc.), ecological stress (pest and disease attack), stress due to nonavailability of marketing place, increase in cost of cultivation and lack of stimulation to the farming community are significant causes of less production. Odisha, one of the potential states of India, constitutes ten agro-climatic zones with thirty districts out of which there were only 13 in numbers before 1993. The farmers of the State show less interest in agricultural activities due to shared environmental and biological hazards. Cyclones hit Odisha almost every year. The climate change scenario in Odisha is reflected very well in the form of erratic rainfall, late onset of monsoon, rise in temperature, which readily affects the rice life cycle from sowing to harvesting stage. With the development of new crop varieties, pests of crop mutate themselves according to the level of hazards. The erratic climate change pattern also helps the pests to grow andmultiplicate faster. Different pests infest the crop at different growing stages, viz., vegetative, reproductive, and ripening. The list of major pests with their damaging growth stages along with stage of crop infestation are enlisted in Table-1. The insect pest attack severely damages the crop by accounting for around 25% of crop loss (Singh; Dhaliwal & Arora, 2015) [1] in the case of rice crops. Particular for any particular pest the economic threshold level (ETL) for a pest determines the pest management steps. Above ETL directs to adopt proper management practices to prevent the economic injury level (EIL) of crop due to pest. The pest incidence density defines the ETL for a particular pest, whereas the lowest population of pests that will cause economic damage is called EIL. In Odisha, based on the severity of the pest attack, the cultivated area is classified into three categories, i.e. (i) Low intensity, (ii) Moderate intensity, and (iii) High intensity. The low infestation intensity is determined below ETL, whereas the rest two categories are classified above ETL, which signifies taking the necessary steps to prevent EIL.

Materials and Methods

The yield data of Balesore district has been obtained from Department of Agriculture and Farmers' Empowerment, Government of Odisha (GoO). The yield loss data have been estimated and the detailed procedure are discussed later. The period of yield loss data is considered as 2001 to 2018.

Analytical Procedure Estimation of yield loss of rice due to pest

Government of Odisha has published two kharif manual in the year 2001 and 2007 respectively. This manual contains the rice varieties are suitable and grown in different agroclimatic zones along with their standard yield. Based on that for the selected districts we calculated average standard yield by averaging the standard yield of varieties grown in that districtin two periods viz. one from 2001 to 2007 and other one from 2008 to 2018. The difference between the average yield and observed yield will be considered as yield loss due to all other factors. Out of the total loss, 25% the total loss will be considered to be the loss due to pest (Dhaliwal et al., 2015) [1] . Likewise for each district from 2001 to 2018 the estimation of yield loss of rice due to pest was carried out.

Holt's Damped Trend or Damped Trend with Exponential Smoothing

This is an extension of Holts Linear trend that adds a damping parameter to regulate the trend from extrapolating too aggressively into the future. In this method the time series divided into three components viz., level L t , trend b t and damping parameter 𝜑whose ranges are 0 to 1.

The equations for individual components are

Finally, the forecast equation for m periods ahead is given by

For 𝜑 = 1, this becomes Holts Linear Trend and for 𝜑 = 0, trend dies out completely and forecasts converge to level.

Results and Discussions Estimation of rice yield loss due to pests

The kharif manual published by Department of Agriculture and Farmers' Empowerment, Government of Odisha (GoO) during 2001 and 2008 suggest the varieties recommended for different districts of Odisha. After obtaining the standard yield for each district, the yield loss has been obtained by differencing the observed yield from standard yield. The pest induced yield loss has been obtained by considering 25% of total yield loss. So, the yield loss due to pests is obtained from 2001 to 2018. Oni and Y.O. Akanle (2018) [4] studied on application of smoothing model to forecast cassava production. In their study they found that damped trend model exhibited better results than double exponential method. The damping factor furter smoothened the unnecessary spike in the dataset resulting better model fitting. Implementing, the final chosen parameters to whole data set, the future forecasting values can be generated. For Balesore district, next sevenyear (2019-2025) pest induced yield loss were predicted successfully along with 95% confidence interval to curb the uncertainty.

Conclusion:

Diminishing the yield loss in rice is one of the method to increase the production. In this study, yield loss has been obtained by differencing the standard yield of the district and actual yield of the district. The pest induced yield loss has been estimated using the Dhaliwal's method. According to his method, 25% of total yield loss is due to pest. Based on the time series data of yield loss due to pests, the future yield loss due to pests has been forecasted using Holt's. Damped Trend method. The best parameter combination having least information criterion and error metrics will be shortlisted to predict the future yield loss due to pests. The parameter combination α = 0.7, β = 1, φ = 0.7 has least information criteria and error metrics. It has obtained 16.29% MAPE for testing data set. Using this configuration, future forecast from 2019-2025 has been made with the 95% confidence interval.

References

1. Dhaliwal GS, Arora R. Integrated pest management: concepts and approaches. Ludhiana: Kalyani Publishers; 2015.
2. Gardner ES Jr. Exponential smoothing: the state of the art-Part II. Int J Forecast. 2006;22:637-66.
3. McKenzie E, Gardner ES Jr. Damped trend exponential smoothing: a modelling viewpoint. Int J Forecast. 2010;26:661-5.
4. Oni OV, Akanle YO. Comparison of exponential smoothing models for forecasting cassava production. Int J Sci Res Math Stat Sci. 2018;5(3):65-8.
5. Rudhran PRA, John N. Time series forecasting using hybrid Holt's exponential smoothing model with damping parameter and neural network. Int Sci J Eng Manag. 2025;4(4).
6. Sbrana G. Damped trend exponential smoothing: prediction and control. J Quant Econ. 2012;10(2)