Identification of dry and wet events in karst and non-karst areas of Guangxi based on a novel optimization method involving the antecedent precipitation attenuation coefficient

Abstract

The determination of the traditional antecedent precipitation (AP) attenuation coefficient k is somewhat random and empirical, which is not conducive to the evaluation of AP influence in different regions and the comparison of results. In this study, two traditional calculation methods of AP are coupled to derive the formula for k applicable to different time scales in karst and non-karst areas. A novel optimization method incorporating an artificial neural network for determining the number of antecedent days N and the value of k is proposed. Dry and wet events in karst and non-karst areas of Guangxi based on hydrometeorological data at four time-scales (3 h, 6 h, 12 h, and 24 h) obtained from the Global Land Data Assimilation System (GLDAS) are identified. The results show that: (1) the mean optimal N value (Nopt) in karst areas of Guangxi is approximately 28 days, which is lower than the mean N value (approximately 59 days) in non-karst areas; the optimal k value (kopt) in karst areas from January to March and from August to December is lower than the k value in non-karst areas, whereas the kopt in karst areas in May and June is larger than the kopt in non-karst areas; (2) the recognition ability of kopt at the daily scale for drought events is better than the recognition ability of the other three time-scales, and kopt has better recognition ability than the traditional fixed k value for dry and wet events; (3) this study innovatively proposes that the kopt optimization method can be effectively applied to the recognition of dry and wet events in karst and non-karst areas of Guangxi, which has important theoretical value and practical significance and can serve as an effective reference for related research in other regions.