Stochastic modelling of the daily precipitation occurrence in the Tocantins-Araguaia hydrographic region, Amazon, Brazil

Evaluating the temporal and spatial variation of the occurrence of dry and rainy days is essential for planning the use of water resources. This article evaluated the behaviour of daily precipitation in the Tocantins-Araguaia hydrographic region (TAHR). The 1st-order stationary Markov chain (MC) of two states was used without considering the spatial correlation between the rainfall gauge stations. The developed model was efficient in estimating the occurrences of daily precipitation for the entire TAHR, as the maximum errors found were equal to RMSE: 14.1568, RSR: 0.0149, NSE: 0.9973 and PBIAS: -1.7650. However, the areas located in the Amazon-Cerrado Biome (Ecotone) require more attention due to the greater occurrence of consecutive drought days. Therefore, it is suitable to use MCs of higher orders for greater accuracy. In addition, the transition probabilities that led to rain conditions were greater when they were closer to the Atlantic Ocean, while the transition probabilities that led to drought conditions were greater in the areas located in the Cerrado biome.