Forecasting wildfires in major forest types of India

Severity of wildfires witnessed in different parts the world recent times has posed a significant challenge to fire control authorities. Even when early warning systems have been developed provide quickest warnings about possible wildfire location, severity, and danger, often it is difficult deploy resources quickly contain at short notice. Response time further delayed terrain complex. Early based on physics-based models, such as WRF-FIRE/SFIRE, are computationally intensive require high performance computing data related fuel properties climate generate forecasts intervals (i.e., hourly basis). It therefore that objective develop monthly yearly forecasts, series models seem be useful they lesser computation power limited (as compared models). Long duration preparing an efficient management plan for optimal deployment event forest fire. The present research aimed forecasting number fires types India basis using “Autoregressive Integrated Moving Average” (both univariate with regressors) 25 km × spatial resolution (grid) developing susceptibility maps Geographical Information System. was validated autocorrelation function (ACF), partial ACF, cumulative periodogram, Portmanteau (L-Jung Box) test. Both univariate- regressor-based performed equally well; however, model preferred due parsimony. R software package used run test model. forecasted active counts were tested against original 3 years from 2015 2017. variation coefficient determination 0.94 (for year 1 forecast) 0.64 (when all 3-year considered together) observed tropical dry deciduous forests. These values varied 0.98 0.89 moist 0.97 0.88 evergreen estimate future frequency ratio, which indicator susceptibility.