Assessing Sea Level Rise Impacts on California Mussel Beds

Of the input variables, elevation was the strongest predictor of mussel location, followed by distance to bluff. The model explained 80% of the variation in our results, though we expect that we could make our model even more accurate by adding more variables.&nbsp;

Our project seeks to assess how projected sea level rise might impact mussel distributions on the California coast, specifically Natural Bridges State Beach. We used ArcGIS Pro to create a random forest model combining NOAA sea level rise data, USGS bathymetry data, and historical site-based mussel counts from the Monterey Bay National Marine Sanctuary's LiMPETS program. The model created a series of heatmap rasters representing predicted mussel zonation at zero feet, three feet, and ten feet of sea level rise. This method allows us to extrapolate observational data from a single transect across a much larger section of coastline.

Slope Raster - we used this elevation raster as the input for the Slope tool, creating a second raster representing the slope, in decimal degrees, of every pixel. These variables are important because mussels, which thrive in the middle intertidal zone, like a fairly narrow depth range.

We hypothesized that mussels would move upshore with rising tides because they are dominant in the mid-intertidal zone. Subquestions included:

We established historical mussel counts at the Natural Bridges site using survey data collected through the LiMPETS program along a 30-meter vertical transect multiple times a year between 2006 and 2019. We mapped the transect in ArcGIS Pro with each point feature representing a quadrat at three-meter intervals to mimic the real life sampling points. Then, we gave each point a table of values representing the count of mussels observed at that quadrat on each survey date.

Overall, our results suggest that the range of potential mussel habitat may actually expand with sea level rise. We think that this could be a result of increased inundation over rocky intertidal areas lowering the risk of desiccation for mussels that would currently be susceptible to dangerous levels of sun exposure. Sea level rise could expand mussel’s upper limit closer to bluffs where they currently can’t survive.LiMPETS vertical transect data exists for 27 species at 27 sites up and down the California coast. Thus, our methods could be used to model the impacts of sea level rise on the zonation of mussels and other species elsewhere. Our recommended next step would be to couple this analysis in tandem with other species that interact with mussels in the rocky intertidal zone in order to create a whole ecosystem analysis in relation to sea level rise. We hope our project can provide a foundation for further study.

NB_MSL_0

NB_MSL_3

NB_MSL_10

NBMLayers_0ft

NBMLayers_3ft

NBMLayers_10ft

The sea mussel, genus Mytilus, is a bivalve mollusk found throughout the west coast of North America. Mussel beds are dominant in the middle intertidal zone, providing habitat and shelter for various invertebrates and algae. They are also a food source for larger animals like starfish, crabs, and sea otters. Their stable populations and relative immobility, combined with their importance in the ecosystem, makes them well-suited as an indicator species for how sea level rise might impact other species.

Our objective was to answer this research question:

We would like to see how upper and lower limits of these particular species within the rocky intertidal ecosystem have shifted or adapted to changing ocean conditions. Many intertidal organisms occupy a narrow habitat niche, making them a great tool for spatial analysis. This, combined with our robust dataset, allows us to observe deviations from the normal distribution and identify any drastic ecological changes that have already occurred or that may be expected to occur.

The study area is a rocky intertidal reef at Natural Bridges State Beach which has been monitored through the LiMPETS program since 2006. The beach is within the Monterey Bay National Marine Sanctuary. We chose this site due to its particularly robust dataset of historical mussel counts, which provided a strong foundation for our random forest model.

Stephenson, T. A. “Causes Of Intertidal Zonation.” Nature, vol. 150, no. 3796, 1 Aug. 1942, pp. 158–158., doi:10.1038/150158a0.

This project is an analysis of how the distribution of sea mussels (Mytilus sp.) in rocky intertidal zones along the California coast might shift with projected sea level rise. We undertook this research for the Advanced GIS course in the MS Environmental Management program at the University of San Francisco, with advising from Professors David Saah and Fernanda Lopez-Ornelas.

Long-term Monitoring and Experiential Training for Students

How will sea level rise affect the zonation of sea mussels at Natural Bridges State Beach?

A machine learning analysis of changing intertidal habitat suitability at Natural Bridges State Beach, California

start

The California Mussel, Mytilus californianus, is a dominant rocky intertidal species (Blanchette 2007) that exist in temperate oceans, and typically occupy lower and middle intertidal zones. Mussel beds form in predictable and defined areas (Paine 1974), and their spatial patterns are hypothesized to be attributed to temperature, productivity, and wave exposure (Blanchette 2007). Some studies show that the primary control agent for zonation is a combination of exposure and temperature, as too much sun exposure will result in dessication. (Stephenson 1942). There are limited amounts of studies that look into how much sea level rise could ultimately impact mussel bed distribution, and we think that this gap in research could be studies using random forest modeling with GIS.

Distance to Bluff Raster - To establish the location of the bluffs above the beach, we used the Raster to Polyline tool to extract a subset of the values from our elevation dataset into a line feature representing the steep slope of the cliff. Then, we manually adjusted the vertices of the bluff line to match satellite imagery for the site. Finally, we used the Euclidean Distance tool to make a raster for the distance to that line from every point in the study area.

<li>We based our analysis on citizen science data collected at Natural Bridges State Beach through the Long-term Monitoring Program and Experiential Training for Students (LiMPETS) from 2006-2019. This dataset contains close to 30 years of transect distribution and zonation data collected via the Long-term Monitoring Program and Experiential Training for Students (LiMPETS) program through the Greater Farallones Association. We will use this data to establish the observed zonation of our study species over several decades, then will estimate random values for other parts of the coast.&nbsp;</li><li>NOAA sea level rise projections: We will also use NOAA shapefiles of past and projected sea level rise to establish average distance-to-shore for each species by year and to predict future intertidal zonation as climate change progresses.&nbsp;</li><li>USGS Bathymetry data: We used this bathymetry dataset to contextualize the zonation of the LiMPETS survey data within the actual topography of the intertidal zone. This enabled us to predict where intertidal zonation may shift more quickly or more slowly relative to seafloor slope and the presence of other geomorphological features. We hope to use intertidal bathymetry to derive slope, which can be used as a predictor variable.</li>

Paine, R. T. “Intertidal Community Structure. Experimental Studies on the Relationship between a Dominant Competitor and Its Principal Predator.” Oecologia, vol. 15, no. 2, 1974, pp. 93–120. JSTOR, <a href="http://www.jstor.org/stable/4214949" target="_blank">www.jstor.org/stable/4214949</a>.

Next, we created four rasters to serve as predictive variables for our random forest model: elevation, slope, distance to bluff, and distance to mean sea level.

<li>Can we use machine learning classification to extrapolate transect-based observational data across a larger area of similar nearby coast?</li><li>Can we use this model with other sites and/or other species to identify high-risk areas for changing intertidal habitat suitability?</li>

Rocky Intertidal reefs function as a dynamic ecosystem, providing habitat and food for many invertebrates and shallow ocean dwellers. They pose as great study areas for population and community dynamics due to the slow moving or sessile nature of the species that occupy them. The vertical gradient of rocky intertidal areas can best be explained by the upper and lower limits that each species can tolerate, whether it is physical, chemical or behavioral boundaries. Factors that control a species zonation include their ability to tolerate prolonged periods of inundation, solar radiation, temperature or wave exposure, and species interactions like predation and habitat exclusion (Stephenson 1942). One common hypothesis is that tides control the spatial limit of most intertidal species (Connell 1972).&nbsp;

The result of our analysis was three raster surfaces representing the predicted zonation of sea mussels at Natural Bridges State Beach with zero feet, three feet, and ten feet of sea level rise. Essentially, each pixel of each raster has an estimated value for mussel count based on the combined predictive value of the four input variables.

Connell, Joseph H. “Community Interactions on Marine Rocky Intertidal Shores.” Annual Review of Ecology and Systematics, vol. 3, 1972, pp. 169–192. JSTOR, <a href="http://www.jstor.org/stable/2096846" target="_blank">www.jstor.org/stable/2096846</a>.

Elevation Raster - projected lidar bathymetry data from the US Geological Survey into a UTM raster, then classified the symbology to get discrete bands representing ranges of depth. This raster contains a value for elevation at every unique pixel in the study area.

The final geoprocessing step was to generate a predictive raster for mussel zonation at 0 feet, three feet, and ten feet of sea level rise. For this, we used the Forest-based Classification and Regression tool with the transect point features and all four variable rasters as inputs, switching out the distance to mean sea level raster for each increment of projected sea level rise that we wanted to model. The result was three raster surfaces -- one for each sea level rise scenario -- representing the predicted zonation of sea mussels.

Distance to Mean Sea Level Raster - To get a raster for distance to this line, we first had to map a polyline representing MSL values along the coast. We used a polygon from NOAA sea level rise projections to establish MSL at the study site in 2016, then converted it into a line feature representing baseline mean sea level. Then we used the Euclidean Distance tool to create a raster representing the distance to that line from every point in the study area. In order to predict mussel zonation in the future, we also repeated the process to create distance-to-MSL rasters for NOAA data representing feet and 10 feet of sea level rise.

Blanchette, Carol &amp; Helmuth, Brian &amp; Gaines, Steven. (2007). Spatial patterns of growth in the mussel, Mytilus californianus, across a major oceanographic and biogeographic boundary at Point Conception, California, USA. Journal of Experimental Marine Biology and Ecology. 340. 126-148. 10.1016/j.jembe.2006.09.022.&nbsp;