An Assessment of Temperature from Multiple Gridded Climate Datasets in a Region with Strong Physiographic Gradients: Michigan, United States

In this study, three gridded climate datasets are examined for their ability to accurately reproduce temperature climatologies at observing sites across Michigan, United States during the period 1981-2018: Parameter-elevation Regressions on Independent Slopes Model (PRISM, versions D1 and D2), Gridded Surface Meteorological (gridMET), and Topography Weather (TopoWx). Michigan is the focus of this study since a comprehensive assessment of gridded climate datasets in a non-mountainous region with otherwise complex physiography is lacking. Two observational networks are utilized, one not assimilated into any of the gridded datasets (Enviroweather), and the other assimilated into all three datasets (Global Historical Climatology Network-Daily). Overall, TopoWx is found to exhibit the smallest deviation from observed daily temperatures, PRISMD2 is found to exhibit the smallest deviations from observed annual extreme minimum temperatures, and gridMET is found to exhibit the largest estimate–observation differences across all timescales. Maps of contoured PRISM-version temperature differences reveal distinct spatiotemporal patterns. Generally speaking, differences are maximized along the lakeshores, in areas with low station density, and at airport station sites. Despite the challenges involved in assessing gridded climate datasets, results of such studies are critical sources of user information. Knowledge of dataset strengths and weaknesses can help inform users as to which datasets and variables are likely to be most accurate or appropriate for a given application and timescale. Study results are expected to be applicable to other regions where terrain gradients, complex coastlines, and land use patterns are important climate controls.