Please click on the poster thumbnail above to enlarge the image.
Audio Presentation
Link will open in a new tab.
External Poster Media Link
Link will open in a new tab.
Remote Predictors of Changing Soil Salinity in an Intermittently Tidal Wetland
Abstract:
<p>Soil salinization has increased due to drought, sea level rise (SLR), and changing precipitation patterns,<br>
leading to greater need to measure it. Remote sensing is a promising tool for estimating soil salinity on<br>
large spatial scales. However, the development of a remote sensing estimation approach for wetland<br>
soil salinity must account for: 1) the high spatial heterogeneity of coastal wetlands, and 2) the fact that<br>
soil salinity is the result of multiple hydrological and geomorphic processes, such as elevation. The<br>
Surface Biology and Geology HIgh Frequency Timeseries (SHIFT) dataset provides a unique opportunity<br>
to assess the application of high-resolution imagery to estimate soil salinity and, when combined with<br>
environmental variables, can account for spatial heterogeneity and multiple processes. In this study, I<br>
combined spectral and environmental datasets to predict soil salinity in an intermittently tidal estuary,<br>
Devereux Slough. Random forest regression was used to determine the model fit between predictor<br>
variables and soil salinity. Elevation was the most important predictor while the modified anthocyanin<br>
reflectance index and other vegetation indices held less importance. Solely using elevation has a high<br>
correlation between predicted and observed values (r = 0.92), but results in maps that do not capture<br>
temporal variability well. The spectral variables better capture the temporal dynamics and their<br>
inclusion results in a more accurate representation of soil salinity over time, despite lower correlation (r<br>
= 0.84). These findings have applicability in salinity monitoring and importance for modeling change in<br>
vegetation community or ecosystem productivity in managed and restored wetlands.</p>
Keywords: Remote Sensing, Machine Learning, Soil Salinity, Wetlands, Pickleweed, Salicornia pacifica
Authors:
German D Silva, University of California - Santa Barbara; Submitting Author / Primary Presenter
, ;
, ;
, ;
, {SubmissionForm.presentersAuthors/affiliation#4};
, ;
, ;
, ;
, ;
, ;
, ;
Remote Predictors of Changing Soil Salinity in an Intermittently Tidal Wetland
Category
In-Person Paper Abstract