Newly collected data across Alaska reveal remarkable biases in solar radiation products

Data on surface solar radiation are scarce in high‐latitude regions, and few studies have evaluated the performance of reanalysis products in estimating solar radiation in those regions. Here, an extensive solar radiation dataset is compiled from 98 stations across Alaska to evaluate 11 different surface solar radiation products (seven reanalysis and four observation‐derived). No product can capture all aspects of the ground‐based observations, and there is ample room for improvement; root mean square errors (RMSEs) of daily, monthly, and annual average comparisons of the products against observations are 38–65, 19–39, and 11–17 W⋅m−2, respectively. ERA5, MERRA2, and ERA‐Interim performed the best in Alaska. Daily records from all products show large RMSEs of 60–108 W⋅m−2 during May–July, equivalent to 30–55% of the observed solar radiation during this season. The sparseness of Alaskan observations, cloud cover, and algorithm issues may be potential sources of bias. An extensive solar radiation dataset is compiled from 98 stations across Alaska to evaluate 11 different surface solar radiation products. No product can capture all aspects of the ground‐based observations, and there is ample room for improvement. Daily records from all products show large root mean square errors of 60–108 W⋅m−2 during May–July, equivalent to 30–55% of the observed solar radiation during this season.