Rainfall and Mekong River flow along the riverfront of upper Northeast Thailand drive agriculture and flood risk, but forecasts for the region rarely address data gaps, nonstationarity or skill relative to climatology. This study develops NAGA-RF, a mathematical model with three components: NAGA-R, a trend-adjusted periodic mixed-Gamma model for monthly rainfall; NAGA-Q, a season-al autoregressive model with exogenous rainfall for monthly discharge; and NAGA-P, a block-bootstrap projection model. Records from four Thai Meteorological Department stations (1995–2025), gap-filled from 3-hourly data, were combined with gridded data and discharge at Nakhon Phanom and tested with Mann–Kendall, Sen’s slope and Pettitt methods. Minimum temperature rose by 0.42–0.46 °C per decade, whereas annual rainfall showed no robust trend. Mekong dry-season flow rose by 49.0% from 2013 and wet-season flow fell by 25.4% from 2014, without a matching change in local rainfall. For 2025, NAGA-R reached a Nash–Sutcliffe efficiency of 0.89 against 0.85 for climatology at the three long-record stations, with well-calibrated 80% intervals, and NAGA-Q reached 0.80 against 0.48 for persistence. Minimum temperature is projected to be 0.9, 1.1 and 2.0 °C above the 1995–2024 normal by 2030, 2035 and 2055.