Applicability of logistic regression model of groundwater levels to drought forecast

초록

During a dry season, groundwater levels hardly react to a small amount of rain whereas it can represent a long-term drought. To calculate the probability of heavy drought, which means lower than -2 in SPI index, two logistic regression models were established using SPI index and groundwater levels for the period, from March to May, as the input variables. The first logit, which was derived from the relationship between the ratio of monthly-averages of groundwater levels in March for the average and drought years and SPI index, is "LOGIT(DTW) = -5.072 + 6.594 x Ratio of depth to water levels in March". The second logit, which was derived from the relationship between the ratio of groundwater levels' slopes from March to May for the average and drought years and SPI index, is "LOGIT(SLOPE) = 0.411 - 0.919 x Ratio of groundwater levels' slopes". Despite assuming no pumping around the monitoring wells and using limited data of the year 2001, these logistic regression models are proven to be effectively applicable for a heavy drought period and improved drought forecast will be possible with an accumulation of available data.

키워드