Author ORCID Identifier
https://orcid.org/0000-0003-0943-1301
Date of Award
10-1-2024
Document Type
Thesis (Master's)
Department or Program
Earth Sciences
First Advisor
Carl E. Renshaw
Second Advisor
Francis J. Magilligan
Third Advisor
Jonathan W. Chipman
Abstract
Monitoring river suspended sediment concentration (SSC) is critical for environmental challenges such as understanding the fate of thawed permafrost sediment and its impact on global carbon cycling. However, traditional SSC monitoring using Landsat imagery is limited by spatial and temporal constraints, particularly for narrow rivers in cloudy and/or snowy regions.
This study investigates the use of higher spatial (3 m) and temporal (daily) resolution satellite imagery from the PlanetScope constellation to estimate SSC in remote rivers such as those in the Arctic. I compare the performance of PlanetScope’s spectral resolution (4 and 8 bands) with Landsat 7. Merging data from both sensors was applied to test if this improved model accuracy.
In comparison to previous site-specific satellite-based estimates of SSC, I found that applying log transformations of the SSC values and normalizing the Landsat surface reflectance data improved the accuracy of the model from ~49% absolute relative error to 30% to 39%. However, I found that site specific SSC regressions are sensitive to overfitting when only a few in situ measurements are available, with low regularization in the LASSO regression sometimes causing the model to behave similarly to ordinary least squares regression, which is known to potentially overfit when the number regressors exceeds the number of calibration data. In the absence of overfitting, the 4-band PlanetScope sensor performs as well or better than 4-band Landsat models. Without sensor harmonization, merging data from different sources generally did not improve SSC prediction accuracy. Because the performance of PlanetScope SSC regressions is no worse than those for Landsat, PlanetScope’s higher spatial and temporal resolution may offer significant advantages in SSC monitoring compared to Landsat for narrow rivers in cloudy and/or snowy regions.
Recommended Citation
Rykin, Ivan, "Challenges of using more precise temporal and spatial resolution of remote sensing data for surface water quality monitoring" (2024). Dartmouth College Master’s Theses. 180.
https://digitalcommons.dartmouth.edu/masters_theses/180
Included in
Climate Commons, Data Science Commons, Environmental Indicators and Impact Assessment Commons, Environmental Monitoring Commons, Fresh Water Studies Commons, Geographic Information Sciences Commons, Glaciology Commons, Hydrology Commons, Remote Sensing Commons, Sedimentology Commons, Spatial Science Commons
