Seasonal quantification of aquatic macrophytes in small boreal lakes with multiscale remote sensing

Small lakes are common across the Boreal-Arctic zone. Due to shallowness and high shoreline-surface area ratios, they are abundant in aquatic macrophytes. Vegetated littoral zones have been suggested to count as wetlands when quantifying carbon sinks and sources, but the actual magnitude of aquatic vegetation is seldom quantified. Remote sensing has potential in this quantification but has rarely been applied to macrophyte abundance and phenology in Boreal-Arctic lakes. We examined the abundance of emergent and floating-leaved vegetation in small, mainly oligotrophic lakes at the northern boreal-subarctic ecotone in Finland with remote sensing. We collected uncrewed aerial vehicle imagery to quantify aquatic macrophytes in five small lakes at total, growth-form, and genus levels in terms of 1) binary presence and 2) percentage coverage. We produced seasonal estimates across a 105 km² catchment by training stochastic gradient boosting models with Sentinel-1, Sentinel-2, and combined multi-sensor data. In a sequential classification-regression approach, the multi-sensor models identified 39-93 % of occurrences of aquatic vegetation and explained 26-93 % of variation in coverage of different target classes. Across 45 small lakes in the catchment, emergent and floating-leaved macrophytes were present in 9 % and 25 % of surface area, respectively, and both life forms covered 2 %. Total coverage by lake ranged 3-97 %. While denser Carex spp. and Nuphar lutea occurrences were well predicted, sparse and erect Equisetum fluviatile stands were challenging to capture. Dominant species varied across the lakes, although N. lutea was the most common. Multispectral Sentinel-2 data universally outperformed Sentinel-1 SAR data in estimating aquatic macrophytes, detecting coverages above ca. 1-5 %. Shoreline effects prevented accurate prediction in lakes with <0.25 ha areal extent. Based on our results, the seasonal mapping of aquatic macrophytes at the genus level with Sentinel data is feasible, though requires passive instrument data, which is only available intermittently due to cloud cover.