Dataset accompanying manuscript submitted to Communications Earth & Environment titled "Higher but more variable annual CO2 emissions from lakes in drier Arctic landscapes" by Hazuková et al.

Database Compilation: Lakes in this dataset are located across Alaska, Siberian Lowlands, Northwest Territories and Nunavut in Canada, West Greenland, and Scandinavia—spanning gradients of precipitation, temperature, and ecotypes (Fig. 1A). We used the Cryosphere Inland Water Greenhouse Gases Database (Song et al., 2024) and added data from more recent publications and data releases (Karlsson et al., 2024; Saros et al., 2024; Verheijen et al., 2022). When available, we also collated dissolved organic C (DOC) concentrations, specific ultraviolet absorption coefficient at 254 nm (SUVA254), total phosphorus (TP) concentrations, and pH from publications and/or authors. Climate Data: We used the European Centre for Medium Range Weather Forecasts’ ERA5 reanalysis monthly resolution products on a 0.25° grid. Lake watersheds are smaller than grid cell areas; hence, we extracted the closest gridded value of monthly accumulated total precipitation, and potential evapotranspiration until 2023 for each lake center coordinate. We subtracted potential evapotranspiration from total precipitation (P-PET) and used it as a metric of aridity and water balance in the landscape (Woo, 2012). Here, we classify lakes with long-term (1994-2023) negative summer (June – August) P-PET as ‘drier’ and lakes with positive summer P-PET as ‘wetter’; we use a time span that covers collection of all measurements included in the dataset. We used seasonal temperature data (1994-2023) to calculate Conrad’s Continentality Index (CCI) for each catchment (Conrad, 1946). Catchment Properties: For each catchment we extracted surface soil organic carbon (C) content (Northern Circumpolar Soil Carbon Database, 0-100 cm Hugelius et al., 2013) and land cover (Global Land Cover, 2019, 100 m resolution Buchhorn et al., 2020). To distinguish between wetland and upland catchments, we classified lake catchments conservatively as wetlands if wetland cover > 0%. We calculated the watershed:lake area ratio (WA:LA) to scale the effect of the size of the catchment in proportion to the lake. Average slope and elevation of each catchment were derived from the processed ArcticDEM layers. To estimate seasonal variability in water surface coverage (hereafter called ‘Seasonal water (%)), we extracted seasonality data layers from Global Surface Water as available and determined the proportion of each catchment that does not represent permanent water during unfrozen periods (Pekel et al., 2016). We also used products from Brown et al., 2002 to identify permafrost condition (continuous, discontinuous, sporadic, and isolated) and ground ice content (visible ice within the first 10-20 m of the soil profile, high > 20%, medium, and low