Quantifying temporal mismatches in satellite and in situ data for aquatic environments

Accurate interpretation of satellite-derived water quality algorithms requires comprehensive error quantification. Most errors in satellite-derived water quality estimates are attributed to satellite data processing or algorithm refinement, and in situ data are often considered the reference standard for validation. However, temporal mismatches between satellite overpasses and in situ sampling introduce an additional and distinct source of uncertainty that is frequently acknowledged but remains poorly characterized. A review of Sentinel-2 inland water chlorophyll-a (chl-a) algorithms found sampling time discrepancies ranging from 1 hour to 5 years, with mean absolute errors spanning 0.11 to 39.48 μg L^-1. To better understand this source of error, we simulated temporal mismatches using a long-term chl-a dataset from buoys on B. Everett Jordan Lake, North Carolina. Results show that the mean absolute deviations increased nonlinearly with rising chl-a concentrations and longer temporal mismatches. The probability of a satellite-derived chl-a value being within ±20% of an in situ measurement ranged from 60- 100% on the same day and declined to 44.44-88.89% after 11 days, depending on chl-a concentration. These findings highlight the npotential impact of temporal mismatches on algorithm performance and the need for consistent quantification and reporting standards in future studies.