Prognosticators for precipitation variability adopting principal component regression analysis

Abstract

Precipitation is an intricate phenomenon influenced by several variables. It is extremely important to maintain life on Earth and balance the hydrological cycle. It is such a critical process that its scarcity leads to droughts and abundance leads to flooding, both extremities bring destruction. Nevertheless, predicting it properly through modeling can help attend to the unpredictability of this very important natural phenomenon. This novel research is dedicated to developing an accurate, mathematical model for establishing potential predictors of precipitation using data from the Pakistan Metrological Department (PMD). The study area is selected on its specific topography which is rugged terrain making it susceptible to flash flooding. On the contrary, Baluchistan province has encountered numerous reoccurring droughts and floods in the past few decades, which has destroyed the economy of the province mainly based on agriculture and livestock. Therefore, using precipitation data as a predictor with significant trends, a principal component regression analysis (PCRA) model has been developed for the significant months which are found to be the month of January and June. The Mann–Kendall technique was implemented to find the trend in the monthly precipitation data of 13 stations selected in Baluchistan which shows positive/negative trends in January and June. Principal components of large-scale oceanic and circulation indices, sea water surface temperature (SWST), geopotential height (GPH), sea-level pressure (SLP), relative humidity (RH), outgoing longwave radiation (OLR), and zonal wind (ZW), were the predictors. PCR is more robust than other modeling techniques; it can handle multicollinearity and reduces redundant variables. The current study identified the potential of precipitation variations with the help of two novel climate indices, EQWIN and ENSO-MODOKI, which have not been studied for the study area. The PCR model developed accounts for 73.33% and 95.05% of precipitation variability for January and June. The model successfully passed all pre- and post-estimation tests. The root mean square errors (RMSE) are 10.13 and 3.63 for January and June respectively. The results also show that the (EQWIN) and (EMI-Modoki) have a substantial effect on the precipitation pattern of a large province (Baluchistan), the province that hosts a significant portion of the routes of the Western and Central China-Pakistan Economic Corridor (CPEC) and International Gwadar port. The study addresses 2 SDGs namely SDG # 11.5 (natural disasters) and SDG#13 (climate action). It is also beneficial to the National Disaster Management Authorization (NDMA) and the Pakistan Metrological/Climate Department (PMD) by initiating timely flood alarms, water management indications, and drought threats. PCR modeling will not only reduce the devastation and catastrophe of flash floods due to unprecedented, torrential rain in the rugged terrain. It will save precious human lives, property, livestock, crops, infrastructure, etc. by early warning.