Jeffrey R. Nielson

Dust has the potential to play a significant role in the nutrient dynamics of alpine watersheds with important ecological implications. However, little is known about how dust nutrients circulate through the environment and which watershed characteristics facilitate dust impacts on water quality. This study explored the contribution of dust‐deposited nutrients, focusing on a high‐elevation Long Term Ecological Research site, where dust samples have been continuously collected since 2017. We incorporated observed dust nutrient compositions, including fractions of inorganic and organic nitrogen and phosphorus, into a popular hydrological model, the Soil and Water Assessment Tool, and ran simulations for 2019–2021. By comparing simulations with and without dust nutrient inputs, we estimated the impact of dust‐deposited nutrients on individual watershed processes. Results revealed a significant contribution of dust‐deposited nutrients, particularly soluble reactive phosphorus (SRP), to several nutrient cycling and transport pathways. Notably, dust contributed up to 19.3% of the SRP load in annual streamflow (increasing monthly streamflow concentration by up to 10.9 μg ). Spatial analysis of model estimates demonstrated a relationship between topography, soil type, and the cycling and transport of dust nutrients. The largest dust nutrient contributions were found in catchment areas with lower slope and less hydric soils, where other natural mobilization processes may be limited. This comparative modeling approach stresses the importance of including dust nutrients in watershed models, especially in oligotrophic systems, and has potential to validate these findings elsewhere and identify how watershed characteristics may either mollify or accentuate the impacts of dust deposition on mountain freshwater systems.

Environmental DNA (eDNA) sampling is a powerful method for detecting aquatic species at low densities. However, eDNA may remain close to the source in lentic systems, decreasing the effectiveness of eDNA surveys. We conducted cage experiments with salamanders and simultaneous detailed hydrologic and wind measurements to investigate the influence of the physical environment on detection patterns of eDNA in ponds. We found much higher detection rates in the surface layer than at depth, and that aquatic vegetation reduced detection of eDNA produced in open water by 80%–94%. Within the surface mixed layer, detection rates were highest close to the source in the direction of water flow in the bottom half of the layer, and detections farthest from the source occurred when velocities in this sublayer were high. Detections were near zero even close to the source when this sublayer was flowing fast and away from the sampling point. The direction of water flow in this lower half of the surface mixed layer was negatively correlated with wind direction for most of the study. These spatial and temporal dynamics indicate that eDNA transport processes in ponds are highly complex. Sampling away from aquatic vegetation, in the surface mixed layer, and upwind of potential sources, in addition to sampling at many locations within a pond and considering temporal patterns, may improve detection of rare pond species. This work contributes to a growing body of literature characterizing the variability of eDNA detection in lentic systems.

Ponds are characterized by high biodiversity, intense biogeochemical cycling, and susceptibility to anthropogenic impacts. Yet few studies have quantified the water velocities responsible for vertical mixing or lateral transport in ponds. We used high‐resolution observations of velocity to examine mixing and transport during summer in a 50‐m‐long, 2.7‐m‐deep temperate pond. Many observed transport and mixing processes resembled those found in larger stratified lakes. A surface mixed layer was observed, whose depth ranged between ~ 1 m at night and < 0.3 m during the day. Turbulence was usually sufficient to vertically mix the surface layer in 4–12 min, but no turbulence was observed in the hypolimnion. Persistent (2.5‐h‐averaged) currents usually flowed downwind near the surface and returned upwind near the mixed layer base. Surface currents were proportional to windspeed, with root‐mean‐squared speed of (persistent hypolimnion currents were much weaker). Superposed on persistent currents were 30‐ to 100‐min‐period fluctuations resulting from internal seiches. These fluctuations were comparable in magnitude to more persistent currents in the mixed layer and dominated in the hypolimnion. Seiches did not advect particles far across the pond, but did contribute to production of mixed layer turbulence. Seiches also contributed to shear dispersion, which was sufficient to mix near‐surface tracers across the pond in 2–4 d. Theory suggests that hypolimnion bottom boundary layers were laminar during the downslope phase of seiche motion, but became turbulent during the upslope phase as near‐bed water flows created unstable stratification.

Retrieval of integrated water vapor content (WVC) from remote sensing data is often ill-posed because of insufficient observational information. There are many factors that cause WVC changes, which yield instability in the accuracy of many traditional algorithms. To overcome this problem, we developed a novel fully-coupled paradigm for the robust retrieval of WVC from thermal infrared remote sensing data. Through the derivation of the physical radiative transfer equation, we determined two conditions that need to be satisfied for the deep learning retrieval paradigm of WVC. The first condition is that the input parameters and output parameters of the deep learning need to be able to build a complete set of solvable equations in theory. The second condition is that, if there is a strong relationship between input parameters and output parameters, it can be directly retrieved. If it is a weak relationship, we need to use prior knowledge to improve the portability and accuracy of the algorithm. The training and test data of deep learning is composed of representative solutions of physical methods and solutions of statistical methods. The representative solutions of the physical methods were obtained from the physical forward model, and the statistical solutions were obtained from multi-source data which can compensate for the defect that the physical model cannot simulate mixed pixels. MODIS L1B data was used for case analysis of paradigm retrieval, and the analysis indicated that four thermal infrared bands were usually needed as the input parameters of deep learning and the integrated water vapor content as the output parameter. When land surface temperature and emissivity were taken as prior knowledge, the root-mean-square error (RMSE) of the retrieved WVC was 0.07 g/cm². The optimal accuracy RMSE was 0.27 g/cm². When there was a strong correlation between input parameters and output parameters, i.e., if there were two bands that were very sensitive to WVC in the band combination, high-precision retrieval could also be achieved without prior knowledge. All the analyses show that the paradigm of deep learning coupling physics and statistics can accurately retrieve WVC, which is a significant improvement on the traditional method and solves the problem of lack of physical interpretation of deep learning.

In a stratified lake, where an internal seiche maintained an oscillating bottom boundary layer (BBL), processes including strain‐induced periodic stratification (SIPS), a downslope mean flow, and upslope‐propagating internal waves controlled fluctuating BBL mixing rates. To investigate these processes, high‐resolution temperature and velocity profiles were measured within 2 m of the sloping bed, through more than 100 cycles of upslope and downslope flows. The associated seiche‐induced velocity reversals coincided with the arrival of upslope‐propagating cold fronts. Although fronts are sometimes associated with internal wave breaking, in this lake the measured frontal slopes were small and breaking did not occur. When fronts were not propagating past instruments, near‐bed stratification was consistent with SIPS, being intensified by downslope flows and reduced by upslope flows. Owing to a downslope near‐bed mean flow, which was previously shown to cancel an upslope internal‐wave Stokes drift, peak downslope velocities regularly exceeded peak upslope velocities. Given these strong downslope velocities, turbulent dissipation rates were most intense during downslope phases. Therefore, during downslope flow, creation of stratification by SIPS coincided with elevated dissipation, creating conditions for effective mixing. Correspondingly, two simple parameterizations for mixing efficiency suggest that most of the total near‐bed buoyancy flux (≥ 82%), and most of the irreversible flux (87%), occurred during the downslope‐flow phase of internal seiche cycles. The observed mechanism for dominant mixing during downslope flow has potential to act in other environments where internal waves propagate up sloping beds.