10 min read
Faculty Advisors:
Lisa Haber, Virginia Commonwealth University
Brandon Alveshere, Virginia Commonwealth University
Graduate Mentor:
Kayla Preisler, University of Arizona
Terrestrial Fluxes Group Introduction
Rice Rivers Center Director Chris Gough and Graduate Mentor Kayla Preisler
Quinn Koch
Monitoring Postfire Ecosystem Recovery With Spectral Indices and Eddy-Covariance Flux Towers
Quinn Koch, University of California, Los Angeles
Fire is a common ecological disturbance in forest ecosystems, leading to changes in forest structure and function that have implications for the Earth's carbon budget. Observations of post-fire carbon fluxes provide insight into the trajectory of forest recovery and its future as a carbon sink. Eddy-covariance flux towers measure high frequency greenhouse gas exchange between forests and the atmosphere, yielding measurements of net ecosystem exchange (NEE), gross primary productivity (GPP), and ecosystem respiration (Reco). While flux towers are the gold standard for quantifying ecosystem scale fluxes, vegetation indices derived from remote sensing are highly correlated with tower flux data and may provide broader spatial scale understanding of how carbon fluxes vary following fire and other disturbances. The objective of our study is to examine the relationship between tower carbon flux data and NASA Landsat-derived spectral indices at five sites in the United States and Australia that were disturbed by severe fire. Specifically, we evaluated changes following fire in two Landsat-derived spectral indices, Normalized Difference Vegetation Index (NDVI) and Normalized Burn Ratio (NBR), examining whether spectral indices paralleled temporal variation in NEE, GPP, and Reco. We found that the recovery of spectral indices outpaced the recovery of NEE and GPP at sites that experienced severe fire, highlighting how lags in structural and functional responses to disturbance may decouple vegetation indices from carbon fluxes. This suggests that a temporal lag should be considered when using vegetation indices as a proxy for carbon fluxes in post-fire ecosystems compared to unburned systems. This analysis represents a small snapshot of ecosystems worldwide; therefore, continuing to monitor these trends at future burned flux tower sites will be crucial to further understanding this relationship.
Sara Typrin
Characterizing Forest Response Pathways in the Blackwater National Wildlife Refuge
Sara Typrin, Carleton College
Coastal forests along the Chesapeake Bay are rapidly becoming marshes due to sea level rise and extreme weather events. Predicting these ecosystem shifts is essential for climate adaptation responses. Previous studies have employed Normalized Difference Vegetation Index (NDVI) time series trends to characterize the resilience of coastal ecosystems; however, few have assessed NDVI variation trends within the Chesapeake Bay coastal region, where rates of sea level rise far exceed the global average. This study examines the spatial distribution of forest response pathways in relation to elevation within Maryland's Blackwater National Wildlife Refuge and the surrounding Eastern Shore region. We used the Landsat 8 record (2014-2024) to extract NDVI values for areas classified as upland forest. We calculated trends in NDVI and NDVI variation using Kendall's τ (rank correlation) to characterize each 30m pixel into one of four ecosystem shift trajectories: abrupt transition, gradual transition, recovering, or stable. We found that 14.7% of the study area is in abrupt transition, 27.4% in gradual transition, 17.3% is in recovery, and 40.6% is stable. Mapping these regions qualitatively shows that in the BNWR, areas closer to the coast tend to experience abrupt or gradual transitions, and areas farther from the coast are typically stable or in recovery. Recovering forests have higher and more variable elevations than other pathways in a subset of BNWR's southwest region. Future work can examine how elevation and distance to the coast relate to forest response pathways at a regional scale.
Austin Jeffery
Structural Characteristic Variation Between Upland Forests and Forested Wetlands
Austin Jeffery, The University of Texas at Austin
Forested wetlands are important for regulating the Earth's climate, cycling nutrients, and providing vital habitats, but are far less studied than upland forests. Prior work in upland forests has illustrated that canopy structural traits vary widely within and across forest types, and that these traits affect crucial ecosystem functions and services such as primary production and carbon sequestration. However, how canopy structure varies within and across forested wetlands has not been thoroughly explored. This study uses waveform lidar data collected during the 2024 SARP East flight campaigns over the Chesapeake Bay region using the LVIS (Land, Vegetation, and Ice Sensor) airborne platform. The LVIS Facility L2 Geolocated Surface Elevation and Canopy Height Products were used to investigate how canopy structure varies across forested wetlands and to compare canopy structural variation between forested wetlands and upland forests. To analyze the data, each lidar granule was first divided into upland and wetland forests by overlaying the granules over a USGS NLCD land use map and a USFS forest type map. Then, 20 plots were created of 100 granules each based on four tree species and whether it was an upland forest or forested wetland plot. Two upland and two wetland species were used with 5 plots each. Then, the data were used to assess variation in structural characteristics, including canopy height and vertical complexity, among forested wetlands and upland forests. The analysis resulted in a significant statistical difference between forested wetlands and upland forests structural characteristics. Additionally, forested wetlands showed a general larger variance in canopy structural complexity suggesting variation in canopy height, canopy density, layering, and forest age. This study serves as a benchmark for LiDAR-based structural characterization of forested wetlands, and informs management and conservation of forested wetlands in the mid-Atlantic region.
Ellery Moore
Arctic Ecosystem Carbon Dynamics: Comparing Greenhouse Gas Measurements in Alaska and Northern Canada Using MODIS Satellite Data and Atmospheric Flask Samples
Ellery Moore, Colby College
As global temperatures continue to warm, the International Panel on Climate Change (IPCC) has called attention to thawing permafrost as a potential tipping point leading to "irreversible" change to Earth's ecosystems. Currently, permafrost holds an estimated 1,400 Pg of carbon, which will be released primarily as greenhouse gases (GHGs), methane (CH4), and carbon dioxide (CO2), through microbial activity as temperatures continue to rise, thus exacerbating the atmospheric GHG effect and further warming. In Alaska and Northern Canada, permafrost underlies most of the land, with regions determined by the percentage of frozen soil: continuous (90-100%) and discontinuous (50-90%). Upon examination of spatial maps, the continuous region tends to correspond to the tundra ecosystem, and the discontinuous region to the boreal forest ecosystem. We quantified the permafrost regions using Moderate Resolution Imaging Spectroradiometer (MODIS) derived Normalized Difference Vegetation Index (NDVI) and land surface temperature (LST). In this study, we aim to determine if CO2 and CH4 concentration measurements differ between the two ecosystems using atmospheric flask samples collected during the Arctic Boreal Vulnerability Experiment (ABoVE) in 2017. Overall, the results showed a positive correlation between NDVI and LST, with the boreal forest characterized by higher NDVI and LST than the tundra. Additionally, higher CO2 concentrations were associated with lower NDVI and LST. However, when separating the samples into the two ecosystems, no difference was seen in their diurnal cycles. In general, CH4 measurements did not show a clear relationship with NDVI and LST, but predominantly higher measurements were seen in the tundra when separating the samples by ecosystem. The different CH4 concentrations could be influenced by other environmental sources not considered in this study, such as thermokarst lakes and anthropogenic factors. Further work to differentiate the ecosystems and confirm findings can be done by examining soil moisture samples and comparing permafrost active layer thicknesses. Additionally, to better understand the rates of carbon release, eddy covariance measurements could be examined between the tundra and boreal forest over time.
Rayyane Matonding
San Francisco BVOC Emissions: The Role of Urban Vegetation in HCHO/NO2 Ratios
Rayyane Matonding, University of San Francisco
Biogenic Volatile Organic Compounds (BVOCs) influence local air quality, especially during summer when emissions and photochemical activity peak. BVOCs can oxidize to form ground level ozone, which poses respiratory health risks. Formaldehyde (HCHO), a key photooxidation product of BVOCs, serves as a useful proxy for biogenic emissions in remote sensing studies. Likewise, nitrogen dioxide (NO2) indicates combustion-related activity and anthropogenic VOC influence. This study examines the relationship between urban tree cover and BVOC-related ozone formation using the HCHO to NO2 photochemical regime, which reflects the balance between biogenic and anthropogenic sources. HCHO and NO2 data were obtained from NASA's TEMPO instrument, and tree cover data from SF OpenData. San Francisco was selected due to its urban greening efforts, high anthropogenic emissions, and prevalence of invasive tree species. Two neighborhoods were selected, Sunnyside with approximately 22 percent canopy cover and Potrero Hill with approximately 2 percent canopy cover, to compare temporal trends in HCHO to NO2 ratios using time series plots. These neighborhoods were chosen based on the availability of hyperlocal weather data, which allowed for more localized atmospheric analysis. No consistent relationship between tree cover and HCHO to NO2 ratios was observed, except during 15:11 and 18:11 on June 18, 2024, which may be associated with elevated photolysis. When weather variables such as zonal wind, meridional wind, and temperature were included in the analysis, no significant correlations were found. Further research should include other cities, additional time periods, and tree species information.
Emmanuel Kaiser-Veyrat
Vegetation Traits to Methane Fluxes: A Machine Learning Approach Across Diverse Wetlands
Emmanuel Kaiser-Veyrat, Cornell University
Wetlands are the largest and most uncertain biological source of CH4, a greenhouse gas with 56 times the radiative forcing of CO2 over a 20-year time horizon. Given the spatiotemporal constraints of these dynamic ecosystems for consistent on-site observations, remotely sensed vegetation indices (VIs) offer a scalable approach to capturing the biophysical and biochemical conditions that govern CH4 exchanges. However, their reliability in wetland environments is challenged by signal saturation in dense vegetation as well as spectral mixing of water, soil, and plants. Seeking to quantify these limitations, we employ the machine learning algorithm, Random Forest Regressor (RFR), to answer the question: Can remotely sensed vegetation traits predict CH4 fluxes across freshwater and saltwater marshes? VIs from the Index DataBase are derived from the Landsat Collection 2 Level-2 products for Landsat-7 ETM+ and Landsat-8 OLI. The FLUXNET-CH4 Community Product yields 17 wetland sites across the contiguous U.S. with daily mean methane flux values spanning some or all of the 2011 to 2018 interval. Generalized flux footprints were computed for every site adopting a uniform approach scaling fetch with increasing measurement height. Extracting feature importances from RFR, we found the Green Vegetation Moisture Index (GVMI) to consistently outperform all other indices, including two meteorological covariates measured from flux tower sites: air temperature and shortwave radiation. Grouping the VIs into five categories (moisture and water, greenness and productivity, structure and soil, pigments, and burn), we found that moisture and water indices consistently scored higher in feature importance than all other categories combined.
