Predicting subsurface sonar observations with satellite-derived ocean surface data in the California Current Ecosystem | JoVE Visualize

Predicting subsurface sonar observations with satellite-derived ocean surface data in the California Current Ecosystem

Kellie R Gadeken¹, Maxwell B Joseph², Joseph McGlinchy²

¹Bredesen Center for Interdisciplinary Research and Graduate Education, University of Tennessee, Knoxville, Tennessee, United States of America.

Plos One|August 20, 2021

View abstract on PubMed

Summary

Satellite data can sometimes predict marine life distribution, but models showed low performance for forecasting. Further research is needed to improve predictions of underwater biomass using remote sensing.

Area of Science:

Marine Ecology
Remote Sensing
Oceanography

Background:

Vessel-based sonar provides valuable data on marine organism distribution but is limited by cost and spatiotemporal coverage.
Satellite data offers wide spatial coverage of surface ocean conditions, presenting an opportunity for broader marine life monitoring.
Linking satellite observations to sonar measurements could enable more frequent and extensive predictions of subsurface biomass.

Purpose of the Study:

To evaluate the potential of using satellite imagery to predict subsurface biomass, a proxy for marine life distribution.
To assess the effectiveness of random forest models in linking satellite data to sonar-derived biomass in the California Current Ecosystem.

Main Methods:

Utilized random forest models to correlate satellite-derived environmental variables with sonar-measured biomass proxies.
Analyzed model performance across various sonar frequencies and depths within the California Current Ecosystem.
Compared model capabilities in spatial interpolation versus spatial and temporal extrapolation.

Main Results:

Satellite data showed potential for biomass prediction in specific scenarios, but overall model performance was low across tested sonar frequencies and depths.
Spatial interpolation tasks yielded better results than spatial or temporal extrapolation, indicating limited predictive reliability.
The approach is currently not robust enough for reliable forecasting or broad spatial extrapolation of marine life distribution.

Conclusions:

While satellite data shows promise, current models have limitations in accurately predicting subsurface marine biomass and distribution.
The developed approach is not yet suitable for forecasting or large-scale spatial extrapolation due to low performance.
Future work should explore model enhancements and address limitations to improve the integration of remote sensing with sonar data for marine ecosystem monitoring.