Refining remote sensing precipitation datasets in the South Pacific with an adaptive multi-method calibration approach

Abstract. Calibration techniques refine numerical model outputs for climate research, often preferred for their simplicity and suitability in many climate impact applications. Atmospheric pattern classifications for conditioned transfer function calibration, common in climate studies, are seldom explored for satellite product calibration, where significant biases may occur compared to in situ meteorological observations. This study proposes a new adaptive calibration approach, applied to the Tropical Rainfall Measuring Mission (TRMM) precipitation product across multiple stations in the South Pacific. The methodology involves the daily classification of the target series into five distinct weather types (WTs) capturing the diverse spatio-temporal precipitation patterns in the region. Various quantile mapping (QM) techniques, including empirical quantile mapping (eQM), parametric quantile mapping (pQM), and generalized Pareto distribution quantile mapping (gpQM), as well as an ordinary scaling, are applied to each WT. We perform a comprehensive validation by evaluating 10 specific precipitation-related indices that hold significance in impact studies, which are then combined into a single ranking framework (RF) score, which offers a comprehensive evaluation of the performance of each calibration method for every weather type. These indices are assigned user-defined weights, allowing for a customized assessment of their relative importance to the overall RF score. Thus, the adaptive approach selects the best performing method for each WT based on the RF score, yielding an optimally calibrated series. Our findings indicate that the adaptive calibration methodology surpasses standard and weather-type-conditioned methods based on a single technique, yielding more accurate calibrated series in terms of mean and extreme precipitation indices consistently across locations. Moreover, this methodology provides the flexibility to customize the calibration process based on user preferences, thereby allowing for specific indices, such as extreme rainfall indicators, to be assigned higher weights. This ability enables the calibration to effectively address the influence of intense rainfall events on the overall distribution. Furthermore, the proposed adaptive method is highly versatile and can be applied to different scenarios, datasets, and regions, provided that a prior weather typing exists to capture the pertinent processes related to regional precipitation patterns. Open-source code and illustrative examples are freely accessible to facilitate the application of the method.