Improving Remote Sensing Derived Aboveground Biomass Predictions Within Inundated Wetlands
Category: Research Poster
Author(s): Megan Hoover
Presenter(s): Megan Hoover
Mentors(s): Jessica O'Connell
[OO1.1]Wetlands provide critical ecosystem services such as carbon sequestration, with potential to mitigate climate warming. Some wetland carbon is stored within plant biomass, which can be measured via remote sensing. Remote sensing of AGB often relies on vegetation indices (VIs) that are mathematical combinations of spectral reflectance bands within the electromagnetic radiation spectrum. The Normalized Difference Vegetation Index (NDVI), composed of near infrared and red reflectance, is commonly used to track vegetation changes. However, in flooded areas, water absorbs longer wavelengths of light which can lead to less reflectance and low AGB estimates. We need a VI that can track AGB across water depth, for more reliable AGB estimates. We hypothesized that utilizing the VARI or Pheno VI in flooded environments would yield better AGB estimates than NDVI, due to a better combination of wavelengths in the index formula. To evaluate this, we built a Generalized Additive Model (GAM) that predicted AGB from NDVI, VARI, and Pheno VI’s, and compared goodness of fit metrics. We also evaluate these result in context of differences in site characteristics. Methods involved obtaining field and Sentinel 2 satellite data from two wetland types (tidal wetlands and depressional freshwater wetland ponds). We processed those data and models in RStudio and completed a statistical analysis. Results indicated that the Pheno and VARI VI were more reliable to use in wet areas and NDVI in drier conditions. These findings will help to quantify carbon sequestration potential via AGB estimates, which can help to guide land management decisions.