GIS based pest-weather model to predict the incidence of Girdle beetle (Oberiopsis brevis) in Soybean crop RAM MANOHAR PATEL*, A. N. SHARMA and PURUSHOTTAM SHARMA

Soybean [Glycine max (L.) Merrill] occupied the prominent place in edible oil economy, accounting for about 40 per cent of total oilseeds production in India and contributing to nearly 25 per cent of domestic vegetable oil production in the country (Sharma, 2016). Longitudinally, it is being cultivated from about 15oN to 25oN (Bhatia et al., 2008) in India and 50oN to 35oS latitudinally worldwide (Gupta et al., 2017; Watanabe et al., 2012). The crop is highly suitable for varying agro-ecological regions covering Madhya Pradesh, Maharashtra, Rajasthan, Chhattisgarh, Telangana, and Karnataka states (Bhatia et al., 2008).Though, the rapid expansion of soybean in central Indian since last five decades in terms of area and production due to its low input cost and comparative profitability to other crops (Sharma et al., 2015) was unparallel, however, its productivity hovers around 1 ton per ha (Patel et al., 2019b; Sharma et al., 2014). Continuous mono-cropping of soybean-wheat/gram sequence for long is one of the leading constraints for low productivity (Joshi and Bhatia, 2003)resulting in increased incidence of biotic and abiotic stresses (Bhatia et al, 2008).Of the biotic constraints, insect-pests are the major ones which potentially obstruct to attain the realized yield and production by increasing the input cost and impairing the quality of the produce (Baburao, 2012). Babu et al. (2017) reported about 26.4% loss of soybean crop due to biotic factors. At present, nearly 275 insects-pest species were infesting the soybean crop (Lokare et al., 2014) and about a dozen of them are infesting severely from sowing to harvesting (Chouhan, 2012).


June 2021
(2010); however, Shrivastava et al. (1972) reported about 13.7-42.2% loss of soybean. Girdle beetle is a polyphagous insect (Tagde, 2015;Garget al., 2014) and infestation may start as early as from 6-8 days after sowing till maturity (Chouhan, 2012). Initial crop stage i.e. nearly 31 st August to 7 th September (35 th to 36 th SMW) is highly prone to girdle beetle infestation and consequently reduces the yield around 16.72% to 18.97% (Malgaya, 2013) and approximately 75% soybean plants were damaged before maturity consequent upon early infestation by girdle beetle (Tirole,2015).The pod and seed yield loss of about 84.4 and 47.2 kg ha -1 respectively has been observed due to infestation of girdle beetle (More et al., 2014a). The change of weather variables significantly affecting the agriculture productivity in many aspects, one among them is changing the insect-pest biology (Shamim et al., 2009;Dhaliwal et al., 2004) which consequently affect the dynamics, distribution and infestation period.
For protecting the soybean crop from insectpests well before the incidence, it's extremely important to understand and investigate the crucial weather parameters which cause the girdle beetle infestation. The information about girdle beetle infestation with changing weather in soybean is very meager. The effectiveness of integrated pest management (IPM) measures depends on how best and timely it has been implemented. Development and use the forewarning model as an insect-advisory (Chattopadhyay et al., 2019;Patel et al., 2020) along with spatial distribution using geo-spatial technology is needed to control and avoid the infestation well in advance.
The insect forewarning in conjunction with geographical information system (GIS) are the measure which are being implemented to thwart the impending incidence, maximize the efficacy and reduce the input cost needed for controlling the insect by spraying the insecticide in the incidence specific location. Rao and Prasad (2020), have mapped the spatial pest risk indices using the DIVA-GIS (an open source geographic information system, downloadable at http://www.diva-gis.org).GIS is the scientific tool which allows the location specific control of the insect and prediction models are used for disseminating insect-advisory to the soybean growers well in advance before causing economic loss. The present study was carried out to comprehend the effect of climate variables on girdle beetle infestation in soybean and develop pest forewarning model.

Study area and Data Collection
Daily village level survey data on girdle beetle incidence on soybean and district level daily meteorological data of twenty districts of Maharashtra, i.e. from 16.71° N, 74.24° Eto 21.15°N, 79.09° E (Fig. 1) Vidarbha Zone was having high incidence due to continuous infestation and with high intensity over the years followed by Madhya Maharashtra Zone had less incidence compared to Vidarbha zone and irregular infestation over the years having larger districts with medium infestation.
Marathwada Zone had least and irregular infestation with less number of districts in different weeks over the years.

Pre-processing of data
Girdle beetle infestation (% damage per meter row length) data of first emergence week to the peak incidence week, along with the current and previous two weeks' data of meteorological parameters of maximum temperature (°C), minimum temperature (°C), average relative humidity (%) and rainfall (mm) of soybean season were used in the analysis.
The variables are of different order of magnitude; hence the variable's transformed weekly data were standardized so that all the variables should have zero mean and one standard deviation using the equation (1) as (1) Where Zi is i th standard normal variate xi, is i th variables, μ is mean of i th variables and σ is standard deviation.
The data were cleansed from outliers and leverage points using cook's D statistics and Student residuals.

Analysis
The collected datasets were split into training dataset (2010-13) and test dataset (2014-15). Multiple regression methodology (Eq. 2) was used to develop forewarning models for Vidarbha, Madhya Maharashtra and Marathwada zones and overall for Maharashtra; and the models were tested and validated using test dataset for accuracy.
(2) Pest-weather model to predict incidence of girdle beetle in soybean Vol. 23,No. 2 where y and Zi are the standard normal variates; βi and βj are the regression coefficients.
The models were formulated by using the peak incidence week (peak model) value (Ghosh et al., 2014) of five year mean of different districts to work out the relationship and effect of weather on girdle beetle infestation for Maharashtra (all zones). However, in mean model, weekly mean data over the years of each district from first emergence week to first peak week used for Zonal models (Vidarbha, Madhya Maharashtra and Marathwada zones). The models were fitted by second degree polynomial equation taking percent damage per meter row length due to girdle beetle as dependent variable and maximum temperature (TMax), minimum temperature (TMin), relative humidity (RH) and rainfall (RF) pertaining to current, 1 st and 2 nd lag week as independent variables (Vannila et al., 2011;Kamakshi et al., 2018).

Model evaluation and validation
The models were developed to evaluate the relative significance of different meteorological variables on the basis of coefficient of determination, R 2 (Kamakshi et al., 2018), RMSE, PRESS statistics. Accuracy of the models were validated using cross-validation approach (LOOCV -Leave One Out Cross-Validation) that is R 2 Pred (Patel et al., 2019b;Montgomery, et al., 2011);comparison of observed and predicted values using RMSE, standardized residual, and two samples 't-test' have been used. SAS Enterprise Guide version 4.3 software (SAS Institute Inc., 2011) was used to carry out all the analyses (Patel et al., 2019a).

Correlation studies
On the perusal of data on girdle beetle incidence on soybean, it has been found that the peak infestation in different districts were mainly during 36-37 th standard meteorological week (SMW). However, there were deviations in girdle beetle incidencedue to spatial and temporal variation in climate variables. Earlier studies also reported the peak activity of girdle beetle during 35-36 th SWM (Netam et al., 2013;Malgaya et al., 2013).
The correlation coefficient has been calculated between girdle beetle infestation and climatic variables of the Whole area pertaining to current, 1 st lag and 2 nd lag week of peak incidence in soybean. The agro-climatic factors played significant role to influence the dynamics of insectpest migration, reproduction and other activities which led to cause the incidence in soybean. Time and place variation were primarily due to the discrepancy in congenial weather conditions. The insects-pestsalter their dynamics and developmental rates that can vary from place and/or time to cope with environmental conditions (Rao et al., 2015).

Development and validation of prediction model
Multiple regression analysis has been employed to formulate the prediction models of Vidarbha Zone, Madhya Maharashtra Zone, Marathwada Zone using mean model Fig. 1: Girdle beetle incidence zones and using peak model for overall Maharashtra. Models were developed to assess the relationship between girdle beetle infestation with meteorological parameters of current, 1 st lag and 2 nd lag weeks using training dataset. The analysis revealed that all the weather variables of three zones and Maharashtra are significant at 5% level of significance except ZRF -1 2 for Maharashtra which is significant at 10% level. The developed models are curvilinear (Table 1) and explained 79.56% (Maharashtra), 80.30% (Vidarbh Zone), 94.62% (Madhya Zone) and 73.56% (Marathwada Zone) variation (R 2 ) of girdle beetle infestation by weather parameters. Current and previous week rainfall, relative humidity, and maximum temperature were found to be significant variables impacting the girdle beetle infestation in Maharashtra. Rainfall of previous week was common significant factor impacting the girdle beetle infestation in all the zones, lag week maximum temperature in Vidarbha and Madhya Maharashtra zones and relative humidity of previous week in Marathwada and Vidarbha zones.
The regression model for Maharashtra was used for forewarning model to work out congenial conditions of girdle beetle incidence on soybean, as the model explained about 80% variations in the data. The cross-validation technique (LOOCV) result of the modelsrevealed the predicted R 2 (R 2 Pred ) variation of 27.55%, 63.12%, 84.54% and 60.46% for Maharashtra, Vidarbha, Madhya Maharashtra and Marathwada zones, respectively. Validation using test dataset has been carried out for comparing thepredicted and observed values of girdle beetle infestation during 2014 to 2015. The root mean square error (RMSE) were1.058, 1.00, 3.37 and 0.82 and two sample t-test p values were 0.087, 0.317, 0.156 and 0.914 > 0.05for Maharashtra, Vidarbha Zone, Madhya Maharashtra Zone and Marathwada Zone, respectively (Patel et.al, 2020). This signified the model predicted with higher accuracy and minimum errors. Also the estimated standardized residuals also were in between -3 to +3 indicated the suitability of the models for predicting the girdle beetle percentage infestation (Akashe et.al, 2016).
The reports revealed that the peak infestation occur between 34 th to 37 th SWM this variation is due to location and time. Our analysis revealed that the peak incidence in most of the districts were during 36-37 th SWM, hence the pre-disposed conditions of whole area were worked out based on 36-37 th SWM to previous two weeks. The congenial conditions of key meteorological parameters which played significant role in girdle beetle incidence are maximum temperature of current week (TMax 0 ) ranged between 28.6-31.6 ºC, current week average relative humidity (RH 0 ) ranged from 85.2-91.8 %, current week rainfall (RF 0 ) ranged from 31.8-119.2 mm, average relative humidity of 1 st lag week (RH -1 ) ranged from 86.3-92.6 %, 1 st lag week rainfall (RF -1 ) ranged from 38.1-76.4 mm, maximum temperature of 2 nd lag week (TMax-2 ) ranged between 27.7-30.8 ºC, and rainfall of 2 nd lag week (RF -2 ) ranged from 23.3-60.7 mm.

CONCLUSION
In the present study, results revealed that the peak infestation of girdle beetle was in 36 th -37 th SMW in most of the districts. The correlation study revealed significantly positive correlation with RH 0 ,RH -2 , and RF -2 but significantly negative correlation with TMax -1 .The congenial weather conditions for the peak infestation of girdle beetle on soybean were observed to be TMax 0 in the range of 28.6 to 31.6 ºC, RH 0 85.2 to 91.8 %, RF 0 31.8 to 119.2 mm, RH -1 86.3 to 92.6 %, RF -1 38.1 to 76.4 mm, TMax-2 27.7 to 30.8 ºC, and RF -2 23.3 to 60.7 mm. respectively. Thus, satisfactorily validated GB Marathwada = -0.996 X ZRH 0 -0.488 X ZRF -1 + 0.866 X ZRH -1 2 R 2 = 73.56%, R 2 Adj = 63.64% R 2 Pred = 60.46%, SE = 0.554 Note: ZTMax 0 , ZTMax -1 and ZTMax -2 are the standardized maximum temperature at current week, 1 st and 2 nd lag week, similarly for other variables whole area model can be utilized to forewarn the farmers two weeks prior to the infestation and disseminate the weather based insect-advisory of girdle beetle incidence on soybean to take precautionary measures. The combination of statistical model with geographical information system can be used to protect the crop from the insect damage by resorting to suitable management practices in the specified locations and to reduce the input cost.