Multiple Linear Regression Analysis to Estimate Hydrological Effects in Soil Rn-222 at Ghuttu, Garhwal Himalaya, India: A Prerequisite to Identify Earthquake Precursors

Various geophysical parameters including soil radon (₂₂₂Rn) are being conT_inuously monitored at Ghuttu, Garhwal Himalaya, India since 2007 as a part of earthquake precursor studies. To analyse the earthquake precursory changes in soil radon, it is essential to clean the soil radon data from other effects. For this, we used data for the period of nine years from 2011 to 2019 and assessed the relationship of soil radon with five other parameters using regression analysis. These parameters are water level, atmospheric pressure, rainfall, air temperature and soil temperature at 10 m depth. We also added one more parameter, i.e. the difference of air temperature (T_out) and soil temperature at 10 m depth (T_in). From the observed six parameters, four showed strong correlation with soil radon. These are (i) water level (correlation coefficient (CC) = –0.9), (ii) atmospheric pressure (CC = 0.6), (iii) air temperature (CC = –0.6) and (iv) temperature difference (T_out – T_in; CC = 0.5). For regression analysis, data during the period 2011–2014 were used for training, while data during 2015–2019 were used for tesT_ing purpose. Based on different models, the one developed using all the six input parameters suggests lowest errors and highest correlation. The observed values of ischolar_main mean square error, mean absolute error and CC were 0.332, 0.281 and 0.931 respectively. The regression coefficients obtained from this model were used to calculate the theoretical radon and residuals. By this approach, the effects of hydrological and atmospheric parameters were found to be reduced to a great extent.