Forecasting Kelp Forest Decline Using Satellite Time Series, Early Warning Signals, and Ocean Climate Drivers

Overview

Kelp forests are highly productive ecosystems along California’s coast. However, kelp forests have been collapsing due to a variety of factors including climate stressors and destructive grazing. Studies point to 40% decrease in stable kelp canopy in California over the past 50 years. Conservation efforts are underway, yet there is a lack of a proactive early warning system that would greatly aid conservation efforts by signaling high risk of canopy collapse. This study investigated whether integrating Critical Slowing Down (CSD) indicators with oceanographic drivers such as sea surface temperature (SST) and nutrient upwelling metrics could provide an early warning system (EWS). I hypothesized that if a model integrates early warning signals with environmental drivers, then it will be able to predict catastrophic regime shifts in kelp canopy biomass, because these statistical metrics detect the critical slowing down that occurs when environmental stressors push the ecosystem toward a tipping point. I analyzed 40 years of Landsat-derived kelp biomass time series throughout the coast of California, using CSD indicators, SST, and upwelling indices to analyze correlation and create predictive models. I developed a novel Geographic-Neighbor Leave-One-Region-Out (LORO) cross-validation protocol. Under Leave-One-Region-Out cross-validation, the model achieved a mean AUC of 0.800 across 4 regions, all statistically significant (block bootstrap B=2000, p < 0.05). The hypothesis is supported, and the model is robust. It leverages established methodology and aligns with the ecological dynamics of California kelp forests.

Forecasting Kelp Forest Decline Using Satellite Time Series, Early Warning Signals, and Ocean Climate Drivers

Overview

Competition history

Related projects

Santa Cruz Kelp Beds Are Recovering Despite Sea Urchins

Monitoring the Decline of Sea Fan Populations Near Anacapa Island Using a Remotely Operated Vehicle

Hybrid Spatiotemporal Framework for Marsh Climate Forcing: Biogeochemical Thresholds, Flux Prediction, Site Optimization

California Tide Pools: Environment in Crisis? What We Can Learn from Algae and Rocks?

Nature-Based Erosion Prevention Field Data and Modeling of Drag Force on Giant Kelp Under Winter Swell

An Accessible, Low Cost Tool for Citizen Scientists: Using Remote Sensing Techniques to Predict Fire Damage Propensity

Risk-Aware Early Wildfire Detection from Satellite Imagery Guided by Fire Weather Indices

Conditional Controls on Kelvin-Wave-Driven Initiation of Historical El Nino Events

Browse more like this