Despite the development of new and novel techniques for gridding precipitation (P) observations, considerable uncertainty remains among these precipitation products. In this study, an analysis over the Upper Colorado watershed reveals that, although the spatial pattern of the annual averaged P is similar across the three datasets considered (CPC (~28km), Livneh (~7km) and PRISM (4km)), the root-mean-square-error (RSME) of daily precipitation among datasets is comparable to the annual average precipitation in any dataset. Further analysis also shows a clear reporting time offset discrepancy between the observed and station data. Observational uncertainties are more pronounced in the extreme indices at regional scale; for example, annual maximum P differs by 20-40 mm/day throughout California’s Sierra Nevadas. Our analysis further identifies significant uncertainties in extreme indices such as annual maximum precipitation and simple precipitation intensity index (SDII). In general, climate model performance metrics depend heavily upon the choice of observational dataset.

Our analysis further examined the variable-resolution CESM at 28km resolution for both the mean and ETCCDI indices. We observed that the model simulated observations with demonstrable skill, although biases were more pronounced in the higher elevation areas. However, model biases are comparable to the observational uncertainties in some cases. Pointwise intensity, duration, frequency (IDF) curves of precipitation in the California region computed from model and observational data were similar at nearly half of the assessed stations. However, IDF estimates in the Central Valley region were overpredicted due to inadequate representation of horizontal and vertical extent of the valley. Nonetheless, the model performance provides some confidence that it can produce reasonable estimates under future climate scenarios.

Future work will examine model performance in more detail with thorough investigation of the processes resulting into model biases. This work is part of ``Project Hyperion’’, aimed at the development of a comprehensive regional climate model (RCM) data assessment capability.