Statistical Comparison of 24 Years of Remote Sensed and Water Quality Monitoring Data for Deer Creek Reservoir
Location
ECC 307/309
Event Website
http://water.usu.edu/
Start Date
4-3-2012 11:20 AM
End Date
4-3-2012 11:40 AM
Description
We compared 24 years (1984-2007) chlorophyll-a data from remote sensing and surface water samples collected at three locations for Deer Creek reservoir using multivariate regression, analysis of variance (ANOVA), and t-tests. Since algae blooms peak during the late summer months, we used data from the August time period. Deer Creek Reservoir has historically exhibited eutrophic tendencies caused by phosphorus pollution, made evident by accelerated algae growth. Consequently, chlorophyll-a concentration is often used as a surrogate measure of algae and overall water quality. Remote sensed chlorophyll-a concentrations were computed using the ratio of reflectance bands 1 (blue-green, 450-520 nm) and 4 (near infrared, 760-900 nm) from satellite imagery and a calibration model derived for East Canyon reservoir. This created a data set with chlorophyll-a values at each pixel, we eliminated cloud cover and bad bands data and computed an average chlorophyll-a content for the reservoir. We used two different multivariate linear regression models to fit the surface water data to both remotelysensed chlorophyll-a and field-measured chlorophyll-a for comparison. We used ANOVA to evaluate the spatial and temporal differences in each model and t-tests to compare the raw chlorophyll-a measurements from both datasets and the predicted chlorophyll-a values from the two models. We found no significant difference between models developed using raw remotelysensed chlorophyll-a or raw field chlorophyll-a for August. However, there was a significant difference between the predicted chlorophyll-a based on remote sensing data and the predicted chlorophyll-a based on field measurements. Chlorophyll-a concentrations were higher at the upper end of the reservoir than near the dam in both models. Temporal variation was similar in both models with a reduction from the mid-1980s to the early-2000s and peaking again in the mid-2000s.
Statistical Comparison of 24 Years of Remote Sensed and Water Quality Monitoring Data for Deer Creek Reservoir
ECC 307/309
We compared 24 years (1984-2007) chlorophyll-a data from remote sensing and surface water samples collected at three locations for Deer Creek reservoir using multivariate regression, analysis of variance (ANOVA), and t-tests. Since algae blooms peak during the late summer months, we used data from the August time period. Deer Creek Reservoir has historically exhibited eutrophic tendencies caused by phosphorus pollution, made evident by accelerated algae growth. Consequently, chlorophyll-a concentration is often used as a surrogate measure of algae and overall water quality. Remote sensed chlorophyll-a concentrations were computed using the ratio of reflectance bands 1 (blue-green, 450-520 nm) and 4 (near infrared, 760-900 nm) from satellite imagery and a calibration model derived for East Canyon reservoir. This created a data set with chlorophyll-a values at each pixel, we eliminated cloud cover and bad bands data and computed an average chlorophyll-a content for the reservoir. We used two different multivariate linear regression models to fit the surface water data to both remotelysensed chlorophyll-a and field-measured chlorophyll-a for comparison. We used ANOVA to evaluate the spatial and temporal differences in each model and t-tests to compare the raw chlorophyll-a measurements from both datasets and the predicted chlorophyll-a values from the two models. We found no significant difference between models developed using raw remotelysensed chlorophyll-a or raw field chlorophyll-a for August. However, there was a significant difference between the predicted chlorophyll-a based on remote sensing data and the predicted chlorophyll-a based on field measurements. Chlorophyll-a concentrations were higher at the upper end of the reservoir than near the dam in both models. Temporal variation was similar in both models with a reduction from the mid-1980s to the early-2000s and peaking again in the mid-2000s.
https://digitalcommons.usu.edu/runoff/2012/AllAbstracts/31