Modelling daily precipitation occurrence and amount using a first-order Markov chain with distribution fitting: A case of black bush polder and Ebini, Guyana

Abstract

Precipitation simulation models are crucial for understanding, decision-making, and responding to phenomena related to hydrological, agricultural, and water resource management. This is particularly true for these climate-sensitive sectors in countries with high average annual rainfall, as well as those that depend on rainfall for food security and economic resilience. The application of precipitation simulation models in Guyana remains largely unexplored, despite the country’s high average annual precipitation and its population residing mainly along the low-elevation coastal zone. This study aimed to develop and evaluate a stochastic precipitation model capable of simulating daily rainfall patterns for two climatically distinct regions of Guyana. Daily rainfall occurrence was modeled using a first-order Markov chain, while wet-day rainfall amounts were fitted to Gamma, Weibull, and Lognormal probability distributions. The analysis used daily rainfall records from 1981 to 2022, with monthly stratification applied to capture Guyana’s bimodal rainfall regime. The model accurately reproduced key precipitation characteristics, showing high agreement between observed and simulated data. Projections for 2023–2030 closely align with established seasonal patterns, replicating the primary wet season (May–August) and the secondary wet season (November–January). The Gamma and Weibull distributions provided superior fits for most months, reflecting the skewed nature of daily rainfall. This study provides the first empirical framework for stochastic rainfall modeling in Guyana, offering a foundation for hydrological forecasting, climate risk assessment, and agricultural planning. The modeling framework also holds transferability for other Caribbean and South American regions facing comparable climatic variability and limited observational data.