Identification of Inconsistencies in Historic Rainfall Records of the Himalayas of Bhutan

Abstract

This study focuses on revisiting the old technique of testing the consistency of historic precipitation data from the Himalayas of Bhutan through a double mass curve (DMC) and residual curve (RC). The method of least square and regression analysis is employed to supplement by providing a statistical means to quantify and justify the inconsistency in the data. Inconsistency in the dataset may arise due to natural or anthropogenic activities and implies the change of precipitation regime over the region compared to the meteorologically homogeneous region or the change in method to collect data. As the precipitation data is pivotal for an assessment of water resources over the Himalayan catchments, testing the historical dataset for consistency is imperative when the data was collected manually. Precipitation data is limited in the country in both the time and space domain as the installation of dense rain gauge for precipitation measurement is challenging. Moreover, maintaining it would be extremely costly and difficult owing to the tough Himalayas terrain. It is vital to investigate the quality of available limited datasets and recommend the best method for rectifying the inconsistency thus capitalizing on the limited resources before progressing with the assessment. The test results are presented in four categories: Best, Good, Satisfactory and Poor category. It is observed that maximum stations fall in Good (32%) and least in Satisfactory (15%) with a sufficient number in Best (29%). It can be concluded that more of the Class C stations shows discontinuity when compared with class Astations AWS (Automatic Weather Station) stations. Further, it is proven that consistent data are also statistically homogeneous. Furthermore, this research puts forward an innovative approach of grouping stations for consistency tests based on the frequency of rainfall events. The result provides an avenue to apply robust algorithms to estimate the missing data since the stations are categorically tagged with their quality.