Salt marshes offer important ecosystem services to coastal populations and are key organismal habitats, but are under threat as a result of drowning related to sea level rise. The extent to which any given marsh is resilient to sea level rise depends on its ability to produce vertical accretion. This is primarily driven by belowground biomass (BGB) production, which explains why BGB may serve as an early warning sign of vulnerability to marsh drowning. Declines in plant productivity may occur in BGB before aboveground biomass (AGB), indicating that BGB may serve as an early warning sign of vulnerability to marsh drowning. However, landscape assessments of BGB are rare, as BGB is difficult to measure and has high spatiotemporal variability. The Belowground Ecosystem Resiliency Model (BERM) is a geospatial informatics tool to estimate whole-plant biomass (AGB and BGB) with satellite, climate, tide, and elevation data at a 30 m spatial scale and monthly time step. BERM was built using machine learning algorithms and extensive ground-truth calibration datasets in U.S. Georgia Spartina alterniflora marshes. Here, we aimed to characterize landscape salt marsh resilience with BERM. To do this, we generated S. alterniflora AGB and BGB predictions across the Georgia coast, covering an area of 691 km2, from 2014-2023 and identified biomass trends. We found broad declines in BGB alongside gains in AGB. A total of 74% of the marsh experienced a decrease in BGB, with an average annual trend of -0.91%. Simultaneously, 88% of the marsh increased in AGB at an average rate of 0.66% per year. We classified much of the marsh (27% of area) as vulnerable to drowning (defined as a decline in BGB that exceeded model error). We also investigated biomass trends against flooding frequency, where flooding was derived via a remote sensing-based model. BGB losses were greater with increasing flooding frequencies, suggesting that accelerated SLR will further reduce productivity. Based on BERM predictions, early stage marsh drowning is likely widespread, and management actions to conserve ecosystem services are an urgent need.

Tropical freshwater lakes are well-known for their high biodiversity, and the East African Great Lakes in particular are renowned for their endemic cichlid fish adaptive radiations. While comparative phylogenetic analyses of extant species flocks have revealed patterns and processes of their diversification, evolutionary trajectories within lineages, impacts of environmental drivers, or the scope and nature of now-extinct diversity remain largely unknown. Time-structured paleodata from geologically young fossil records, such as fossil counts and particularly ancient DNA data, would help fill this large knowledge gap. High ambient temperatures can be detrimental to the preservation of DNA, but refined methodology now allows data generation even from very poorly preserved samples. Here, we show for the first time that fish fossils from tropical lake sediments yield endogenous ancient DNA (aDNA). Despite generally low endogenous content and high sample drop-out, high-throughput sequencing and in some cases sequence capture allowed for taxonomic assignment to family or tribe level and phylogenetic placement of individuals. Even skeletal remains weighing less than 1 mg and up to 2700 years of age could be phylogenetically placed. We find that the relationship of degradation of aDNA with the thermal age of samples is similar to that described for terrestrial samples from cold environments adjusted for elevated temperatures. Success rates and aDNA preservation differed between the investigated lakes Chala, Kivu and Victoria, possibly caused by differences in water oxygenation at deposition. Our study demonstrates that sediments of tropical lakes preserve genetic information on rapidly diversifying taxa over time scales of millennia.