Deep-water oxygen loss is a shared decline across 195 temperate lakes, but lake-specific trends did not persist between 1990–2003 and 2004–2017 or predict 2018–2025

Jane et al. (2021) reported a widespread decrease of late-summer dissolved oxygen (DO) in temperate lakes. We used the public data (EDI edi.698.4, edi.624.1) to test two questions: is the network-wide decrease robust, and do the trends of individual lakes persist in time? We reproduced the published medians (deep DO -0.121, surface -0.110 mg/L per decade). A Python rebuild of the profile pipeline matched the published lake-year means (median difference 0.003 mg/L). The deep-water decrease is robust. With a bootstrap over 23 regions (US states or countries), the median deep trend is -0.12 mg/L per decade (95% CI -0.27 to -0.06), and all 10 regions with at least 5 lakes are negative. A fixed-depth layer with a sampling-date correction gives -0.15 (CI -0.24 to -0.08). But lake-specific trends do not persist. Deep-water lake slopes in 1990-2003 do not predict slopes in 2004-2017 (r = -0.10, 158 lakes; with a 4-year gap r = -0.15, CI -0.29 to -0.0004). A simulation with lake-specific AR(1) noise rejects a persistent lake trend SD of 0.3 mg/L per decade or more (p = 0.033). A random-effects meta-analysis of full-period slopes gives 0.30. For surface DO, the associations of trends with surface temperature and latitude change sign between the periods. New 2018-2025 profiles from the same monitoring sites (Water Quality Portal, NTL-LTER) reproduced the published values exactly at 140 surface and 71 deep lakes. In 95 surface and 16 deep lakes with enough data, the median change from 2008-2017 to 2018-2025 was -0.03 mg/L (CI -0.14 to +0.08) and -0.10 mg/L (CI -0.27 to +0.41). These agree with the network trend but do not confirm it. Extrapolation of each lake's surface trend predicted 2018-2025 worse than the 2008-2017 mean (MAE +0.18 mg/L, CI +0.10 to +0.28). Limits: the persistence bound is for 14-year linear trends with independent AR(1) noise. The out-of-sample lakes are mostly in Minnesota, New Hampshire and Wisconsin, and the deep test has low power.

lake deoxygenation · dissolved oxygen · hypolimnion · trend persistence · out-of-sample test · reproduction · Water Quality Portal · NTL-LTER · regional clustering · temperate lakes · long-term monitoring · random-effects meta-analysis