How Accurately Are Climatological Characteristics and Surface Water and Energy Balances Represented for the Colombian Amazon Basin?

Abstract

Context: The Colombian Amazon basin (CAB) is characterized by having many areas with few or no in situ observations. To help fix this shortcoming, there are different gridded datasets that represent meteorological variable estimates from in situ, satellite-based, merged, and model-based products. However, before using them, it is necessary to evaluate their performance.Methods: We analyzed the capacity of five grid datasets to represent the annual cycle, the annual means, and the interquartile range of temperature and the variables that intervene in the surface water and energy balances of the CAB. The five datasets were a regional climate model (REMO) and four reanalyses (ERA40, ERA-Interim, NCEP/NCAR, and NCEP-v2). For the temperature and precipitation analyses, the Delaware, GPCP, CPC, CMAP, and CMAP-v2 datasets were also taken into account. Results: The average annual temperature cycle in the CAB has an unimodal behavior, with higher values between October and March. Only Delaware and ERA-Interim describe this behavior and exhibit values close to those of the observations. The average annual cycle of precipitation is also unimodal, with higher values between March and July. Delaware, GPCP, CPC, and ERA-Interim describe this behavior, although ERA-Interim overestimates precipitation. Additionally, the four reanalyses, especially ERA-Interim, show an unexpected trend of increase in mean annual precipitation. As for the water balance, the precipitation and runoff maps of ERA-Interim and REMO show higher values in the foothills and east of 73 degrees W, a pattern that shows better agreement with the observations than those of the other datasets. ERA-Interim, REMO, and ERA-40 meet the closure condition of both the long-term water and the energy budgets. However, REMO and ERA-Interim overestimated precipitation, ERA-40 underestimated it, and ERA-Interim overestimated solar radiation. This implies that some of the other variables for the water and energy balances are also poorly estimated. Conclusions: The five gridded datasets that allow studying surface water and energy balances are not adequate, so better models and reanalyses are required.