Technical note: Evaluation and bias correction of an observation-based global runoff dataset using streamflow observations from small tropical catchments in the Philippines

Abstract

Abstract. Even in relatively wet tropical regions, seasonal fluctuations in the water cycle affect the consistent and reliable supply of water for urban, industrial, and agricultural uses. Historic streamflow monitoring datasets are crucial in assessing our ability to model and subsequently plan for future hydrologic changes. In this technical note, we evaluate a new observation-based global product of monthly runoff (GRUN; Ghiggi et al., 2019) for 55 small tropical catchments in the Philippines with at least 10 years of data, extending back to 1946 in some cases. Since GRUN did not use discharge data from the Philippines to train or calibrate their models, the data presented in this study, 11 915 monthly data points, provide an independent evaluation of this product. We demonstrate across all observations a significant but weak correlation (r2=0.372) between the GRUN-predicted values and observed river discharge, as well as somewhat skillful prediction (volumetric efficiency = 0.363 and log(Nash–Sutcliffe efficiency) = 0.453). GRUN performs best among catchments located in climate types III (no pronounced maximum rainfall with short dry season) and IV (evenly distributed rainfall, no dry season). There was a weak negative correlation between volumetric efficiency and catchment area, and there was a positive correlation between volumetric efficiency and mean observed runoff. Further, analysis for individual rivers demonstrates systematic biases (over- and underestimation) of baseflow during the dry season and underprediction of peak flow during some wet months for most catchments. To correct for underprediction during wet months, we applied a log-transform bias correction which greatly improves the nationwide root mean square error between GRUN and the observations by an order of magnitude (2.648 mm d−1 vs. 0.292 mm d−1). This technical note demonstrates the importance of performing such corrections when determining the proportional contribution of smaller catchments or tropical islands such as the Philippines to global tabulations of discharge. These results also demonstrate the potential use of GRUN and future data products of this nature after consideration and correction of systematic biases to (1) assess trends in regional-scale runoff over the past century, (2) validate hydrologic models for unmonitored catchments in the Philippines, and (3) assess the impact of hydrometeorological phenomena to seasonal water supply in this wet but drought-prone archipelago.