SARP East 2024 Hydroecology Group

Faculty Advisors:
Dr. Dom Ciruzzi, College of William & Mary
Graduate Mentor:
Marley Majetic, Pennsylvania State University

Marley Majetic, graduate mentor for the 2024 SARP Hydroecology group, provides an introduction for each of the group members and shares behind-the scenes moments from the internship.

Jordan DiPrima, University of New Hampshire
Land subsidence is a frequently overlooked geologic hazard that is caused by natural processes and, more recently, anthropogenic stressors. The goal of this study is to observe subsidence trends and hotspots among land cover types on Virginia’s Eastern Shore and Long Island, New York. This study utilizes interferometric synthetic aperture radar, or InSAR, satellite data from Sentinel-1 to map vertical land motion from 2017 to 2023. Land cover data were sourced from Landsat 8 satellite imagery. Subsidence was mapped within the following land cover types on the Eastern Shore: urban, wetland, cropland, temperate or sub-polar grassland, temperate or sub-polar shrubland, mixed forest, and temperate or subpolar needleleaf forest. These land cover types have mean vertical velocities ranging from -0.2 mm/yr to -5.2 mm/yr. Results suggest that land subsidence is most severe in cropland areas on the Eastern Shore, with a mean vertical velocity of -5.2 mm/yr. In contrast, wetlands display the most subsidence on Long Island with a mean vertical velocity of -2.1 mm/yr. Long Island lacked distinct trends among land cover types and instead showed evidence of subsidence hotspots. These hotspots exist in the following land cover types: temperate or sub-polar grassland, barren lands, wetland, cropland, and temperate or sub-polar broadleaf deciduous forest. Overall, Eastern Shore croplands and Long Island wetlands were determined to be the most susceptible land cover types. These findings highlight regions at risk of sea level rise, flooding, and coastal erosion as a result of subsidence. With further research, we can map subsiding landscapes on a global scale to improve resource allocation and mitigation techniques.

Isabelle Peterson, University of Wisconsin – River Falls
Thermokarst landscapes have and will continue to change as the arctic landscape warms due to climate change. Permafrost underlies much of these arctic landscapes, and as it melts, thermokarst landscapes are left behind. The Seward Peninsula in Alaska has an abundance of these landscapes, and thermokarst lakes are present in the northernmost portion. Several lakes have come and gone, but with increasing climate instability and warming of the area, there is a possibility of more permafrost melting, creating more of these lakes. To capture these changes, Harmonized Landsat Sentinel-2 (HLS) imagery were used to create annual lake maps of the northern portion of the Seward Peninsula from 2016 to 2024. Much of the methodology was informed from Jones et al. (2011); however, their study used eCognition, while the present study used ArcGIS Pro. This caused some differences in results likely due to the differences in software, satellite imagery, and the proposed study area. Lake number changes were observed annually. From this annual change, several 10 to 40 ha lakes disappeared and reappeared within the study period, along with smaller lakes filling in where larger lakes once were. Thermokarst lake drainage is a process described by Jones and Arp (2015) which has devastating geomorphological impacts on the surrounding area, creating large drainage troughs which diminish surrounding permafrost in a quick time frame. To capture these events and overall changes, satellite imagery is essential. This is especially true in remote regions which are hard to reach by foot and require flight missions to be scheduled over the area for aerial photography. However, LVIS and other higher resolution aerial instruments would provide higher accuracy when identifying smaller lakes, as satellite imagery does not accurately capture lakes below 1 ha in the study area. This assertion is made due to conflicting results compared to Jones et al (2011). While the methodologies of this study have been executed manually, Qin, Zhang, and Lu (2023) have proposed the idea of using Sentinel-2 imagery to map thermokarst lakes through automatic methods. While automatization has not yet been perfected, the potential is there and can be used to analyze thermokarst areas effectively. With more satellite imagery, annual, monthly, and potentially daily changes can be captured in favorable months to monitor changing landscapes in arctic regions. Thermokarst lakes have been changing, and monitoring them can help in the process of understanding the changing climate in arctic areas, especially through the lens melting permafrost.

Emmanelle Cuasay, Miami University
Refugia are areas that are characterized by stable environmental conditions that can act as a refuge for species as Earth’s climate warms. In this study, fourteen Harmonized Landsat Sentinel-2 images from February 2014 – March 2024 of the northern Philippines region were used. The region of interest is the terrestrial biome by Lake Taal. Normalized Difference Vegetation Index (NDVI) maps were created from all fourteen images to determine the NDVI 25th highest quartiles of the long-term average NDVI images and of a dry and wet year NDVI image. These values were then used to create refugia and non-refugia maps using ArcGIS Pro. Land cover data from Sentinel-2 and a digital elevation model (DEM), using the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER), were plotted in ArcGIS Pro to determine the slope and aspect of the area. Global Ecosystems Dynamics Investigation (GEDI) data were used to look at forest height of the study area, and the distribution of forest height, slope, aspect, and elevation were plotted to determine their probability densities in refugia and non-refugia areas. Results of this study show increased biomass in refugia areas. This suggests that conservation practices are crucial to aid in the preservation of biodiversity and biomass within these refugia areas.

Jayce Crayne, University of Louisiana at Monroe
Saharan dust storms serve an important role in the western Atlantic’s climate in their contribution to Earth’s radiation budget, modulating sea surface temperatures (SSTs), fertilizing ecosystems, and suppressing cloud and precipitation patterns (Yuan et al., 2020). However, Saharan dust storms are expected to become less frequent in this region as SSTs continue to rise (Yuan et al., 2020). Predicting the climate response to this change requires a keen understanding of how the presence of these storms affect evapotranspiration (ET) and its indicators. This study utilizes site-based observational data from an AmeriFlux tower near Gainesville, FL recorded during a large dust storm in late June 2020. The storm’s progression was documented using satellite imagery from Aqua and Terra and aerosol optical depth (AOD) measurements from an Aerosol Robotic Network (AERONET) station co-located with the AmeriFlux tower. Indicators of ET such as surface air temperature, vapor pressure deficit, photosynthetic photon flux density, and net radiation were analyzed. Findings were compared to modeled ET and latent energy flux reanalysis data provided by the Global Land Data Assimilation System (GLDAS). Both model simulations and on-site observations support that ET decreased during the days dust concentrations were heaviest and for a short time thereafter. Cloud cover data adopted from meteorological aerodrome reports (METARs) provided by an automated surface observing system (ASOS) located in Gainesville showed that clouds were not a major contributor in decreasing ET during the days of heaviest dust. The results of this study show a considerable decrease in ET as a result of dust aerosols. Further research is necessary to determine whether changes in ET due to Saharan dust storms are significant enough to alter climates in the western Atlantic and, if so, what the climate response will be if the frequency of storms decreases.

Brandon Wilson, North Carolina State University
Biodiverse regions across California remain vulnerable to harmful wildfires year round. Quantifying and measuring these regions’ wildfire resilience is necessary for understanding where/how to allocate environmental resources. Several ecological wildfire studies have been conducted utilizing artificial intelligence and remote sensing to analyze and predict biodiversity damage across wildfire prone regions, including Northern Algeria and Arkansas, USA. The current case study aims to analyze biodiversity damage from the 2023 Csarf Smith River Complex Fire in Six Rivers National Forest, California and predict the difference in Normalized Burn Ratio (dNBR) and difference in Normalized Difference Vegetation Index (dNDVI) for 2025 and 2028 using remote-sensing-based random forest (RF) regression. Furthermore, to observe, holistically, a practical method California has implemented to address state-wide wildfire damage, the 2019 California Wildfire Fund (AB 1054 and AB 111) was evaluated using the synthetic control method (SCM). For this case study, remote sensing data from the United States Geological Survey (USGS) and NASA (Landsat 9 Satellite C2 L2, TerraClimate and the Land Data Assimilation System) were utilized for processing relevant spectral indexes for the RF. Data from NOAA, Energy Information Agency, International Monetary Fund and Bureau of Economic Analysis were utilized as synthetic control datasets to evaluate the effects of the 2019 California Wildfire Fund. Elevated topography in this study area is susceptible to high severity burn effects, while less elevated topography burns less. This result affected dNBR and dNDVI predictions as elevated areas seemingly did not have strong resilience to rampant burns. This demonstrates a direct correlation to potential lower transpiration rates for elevated areas, warranting further analysis. Results of low variance, post-treatment, between the treated unit and the synthetic control unit, poses concern for the positive effect of the 2019 Wildfire Fund.

Carrie Hashimoto, Rice University
Climatic warming and high tree density have caused larger and more severe wildfires to occur in western United States forests over time. Wildfires affect both the hydrology and ecology of forests via alterations to the water balance (e.g., evapotranspiration, streamflow, infiltration, and more) and could shift vegetation communities and subsequent ecosystem structure and function. This project explores ecological characteristics of a landscape that predict the extent to which the Creek Fire in the southern Sierra Nevada has affected evapotranspiration. Strides in understanding of consequential evapotranspiration changes can create pathways to address emerging forest health challenges posed by similar western fires. For analysis, various remote sensing and modeled data were collected from OpenET, the North American Land Data Assimilation System, TerraClimate, Harmonized LandSat Sentinel-2 data, and the Shuttle Radar Topography Mission. Multiple linear regression and generalized additive models were constructed. Relative change in evapotranspiration served as the response variable. Model covariates included average temperature, total precipitation in the preceding months, average soil moisture, elevation, slope, aspect, northness, latitude, pre-fire normalized difference vegetation index (NDVI), and post-fire change in normalized burn ratio (dNBR). Best subset selection with cross validation demonstrated minimization of cross-validation error with a 7-covariate model. This reduced model yields lower complexity and more interpretability while sustaining an adjusted R2 of 0.626, compared to the full model’s adjusted R2 of 0.663. A reduced generalized additive model (GAM) with interaction terms drawn from the linear model variable selection demonstrated an adjusted R2 of 0.695, indicating a better fit that comes at the cost of reduced interpretability and higher computational requirements than the linear models. The goal of this work is to disentangle environmental indicators of post-fire evapotranspiration change, such that predictive modeling of future wildfire impacts on evapotranspiration can be achieved.