Research background
The goal of my graduate research project is to help build a more comprehensive picture of native Olympia oyster spatial distribution and abundance in Yaquina Bay, Oregon. Olympia oysters have experienced a major decline in population in the Pacific Northwest; it is estimated that the original population has been reduced by 90-99% (Beck et al. 2011). The species is an important ecological and cultural resource, which has garnered it support from resource management agencies, tribes, and local environmental groups. Yaquina Bay is one of only three Oregon estuaries that has historically supported Olympia oysters and may be the only estuary to have maintained a continuous population since Native American settlement (Groth and Rumrill 2009). Because of this, it has been identified by local shellfish biologists as a good candidate for species enhancement. However, current information about exactly where and how many wild oysters exist in Yaquina Bay is inexact and anecdotal. My research project aims to fill that data gap so that appropriate, informed actions can be taken for species enhancement and recovery.

Question asked
Through this course, I have been experimenting with spatial analysis to see if I can create an accurate map of habitat suitability for Olympia oysters in Yaquina Bay. To do this, I’ve been focusing on three environmental parameters: salinity, substrate, and elevation. I then compared the map of predicted suitable habitat with field observations of known oyster locations. The main research question I examined in this project was:

How does the spatial pattern of Olympia oyster location data correspond to the spatial pattern of suitable habitat in Yaquina Bay?

Description of the dataset

Habitat parameters
I collected one raster layer for each of the three habitat parameters, then overlaid them on one another to establish suitable habitat. I ranked the subcategories within each parameter from highest to lowest likelihood to be suitable for oyster presence.

• Salinity: Salinity data was derived from an existing dataset available through the Oregon Department of Environmental Quality (DEQ). DEQ operates a publicly accessible online water monitoring data portal where results water quality monitoring stations are posted. Wet-season salinity was used to represent the extreme of salinity stress, since Olympia oysters are more capable of withstanding higher salinity waters than fresher water. The salinity dataset is based on historical measurements from 1960-2006 and represents a gradient from highest salinity (~32psu) to fresher water (<16psu).
• Substrate: Substrate data was taken from the Estuary Habitat Map of Oregon using the Coastal and Marine Ecological Classification Standard (CMECS). CMECS is a national standard; publicly available data for Oregon’s coastal areas was made accessible through an effort by the Oregon Coastal Management Program (Department of Land Conservation and Development). The CMECS Estuarine Substrate Component layer is an Oregon-wide raster that I clipped to Yaquina Bay for this analysis. This layer characterizes substrate type broadly and is fairly low resolution.
• Elevation: Elevation is represented through a bathymetric dataset from 2001-2002, which is the most current estuary-wide available dataset. This raster was sourced from the Newport, Oregon office of the Environmental Protection Agency (EPA) through contact with a staff geographer.

Oyster location data
This spring, I completed several field surveys of the intertidal zone of Yaquina Bay during low tide series. These were conducted in collaboration with my advisor and field staff from the Oregon Department of Fish and Wildlife (ODFW) on April 19-20 and May 17, 2019. We used a stratified random sampling design to randomly select survey locations. Once at the field survey locations, we timed our search for five minutes and assessed as much area as possible within that time. Oysters were characterized as ‘present’ if evidence of at least one oyster, living or dead, was detected within the search area. The oyster or oyster shell needed to be attached to a substrate to qualify. These data points were collected using a handheld Garmin GPS unit and field notebook, then transferred to Google Earth for quality control, and finally to ArcGIS Pro for analysis.

Hypotheses
I originally hypothesized that the spatial pattern of Olympia oyster location data would correspond with the spatial pattern of suitable habitat as established by analysis of the three habitat parameters (salinity, substrate, elevation). I did not think it would be exact, but would at least allow for a somewhat accurate prediction of the where the oysters were located in the estuary. I felt this model would be most useful in determining the spatial range of the oysters by finding the extreme ends of the population upstream and downstream. Of the three habitat parameters, I hypothesized that salinity would be the ultimate predictor of where oysters could be found and that substrate would be the limiting parameter.

Approaches
Exercise 1
I did not have any oyster location data to work with yet, so I first focused on the three habitat parameters to identify and rank where habitat was least to most suitable. I ranked the subcategories within each parameter raster layer by looking through the available literature on the subject and consulting with shellfish biologists to get an idea of what conditions oysters prefer. I then combined them using the ‘Weighted Overlay’ tool in ArcGIS Pro to determine where overlap between the layers shows the best environmental conditions for oysters to survive.

Exercise 2
After collecting field data to determine where oysters were present or absent within the intertidal zone of Yaquina Bay, I used the data points to compare against the map of suitable habitat. I first uploaded the data points collected in the field into Google Earth where I could easily verify the locations against my field notes and photos, as well as perform some minor quality control. The points were imported into ArcGIS Pro where I used the ‘Buffer’ and ‘Zonal Statistics’ tools to analyze the neighborhood surrounding each data point. Looking at these neighborhoods, I determined what habitat suitability type that the majority of surrounding habitat fit into. Finally, I made a confusion matrix to statistically assess which data points fit into each habitat type.

Exercise 3
Based on the results of the confusion matrix in Exercise 2, I wanted to expand on the habitat suitability analysis to see if I could more accurately predict oyster locations. I decided to see if the spatial pattern of oyster location data could be better described by manipulating the spatial pattern of one of the habitat parameters. I decided to change the rank values of the salinity parameter to alter the map of suitable habitat. Once I had a new output for habitat suitability, I ran the same neighborhood analysis that I used in Exercise 2 and created another confusion matrix.

Results
Exercise 1
After applying the ‘Weighted Overlay’ to the input raster layers, this map was generated as a representation of habitat suitability:

Exercise 2
This neighborhood analysis map shows the majority habitat type surrounding each data point:

This table below shows how many data points fell into each habitat type:

This confusion matrix further examines the ‘errors’ in the data, or the observations that did not correlate with my model of predicted suitable habitat. For ease of interpretation, I removed habitat suitability type 1 since there were not observations in this category, and type 3 since it fell in between high and low-quality habitat. Decimals reported indicate the proportion of total observations (n = 35) that fell within this category:

Exercise 3
This second neighborhood analysis map shows the majority habitat type surrounding each data point after the habitat suitability model was altered (by manipulating salinity):

This table below shows how many data points fell into each habitat type:

I created another confusion matrix to further examine the ‘errors’ in the data. I removed habitat suitability type 1 since there were not observations in this category. Decimals reported indicate the proportion of total observations (n = 35) that fell within this category. The confusion matrix shows that the model fit nearly all observations (31) into the type 3 habitat category:

Significance
From this analysis, I feel that these three habitat parameters do a good job of predicting the spatial distribution of oysters generally. This approach can be used as a helpful tool for guiding resource managers and shellfish biologists to likely oyster locations, but cannot be the only method for validating their presence or absence at this point. Given budget cuts or competing priorities for today’s resource managers however, a map predicting where an oyster population is might be enough for making management decisions in a pinch.

In collaboration with my advisor at ODFW, I am using the results of this spatial analysis to fill an important data gap about where and how many Olympia oysters can be found in Yaquina Bay.  This information will be used to develop a repeatable monitoring protocol to be conducted by ODFW’s Shellfish and Estuarine Assessment of Coastal Oregon (SEACOR) program. I hope that my research can also help estimate population enhancement potential by identifying possible sites for habitat restoration. With increased accuracy, I’d also like to analyze potential oyster distribution under future environmental conditions and under different management scenarios.

Recovery of a robust population of Olympia oysters is important to Yaquina Bay and the surrounding community in a number of ways. The species provides many benefits to the ecosystem, including water filtration and habitat for other marine creatures. It is culturally significant to local tribes, including the Confederated Tribes of Siletz. While Olympia oysters are not currently listed as threatened or endangered, possible listing would trigger a number of mitigation and conservation measures that will be difficult and expensive for agencies and private landowners. Additionally, there’s been some exploration that if the population can become robust again, there is potential to grow and harvest this species as a specialty food product.

Software and statistics learning
Most of the spatial analysis tools I used were familiar to me, though I had only really used them in exercises in other classes. I have not been able to use my own data to conduct an analysis, so that was exciting, but also a bit of a learning curve. The confusion matrix was a new tool that was a very simple and useful way to statistically assess my data.

Some issues worth noting
Throughout this process, I identified several small issues that I will need to consider or modify to improve my research analysis:

1. Regarding the neighborhood analysis, the size of the buffer around each point affects how the majority of habitat type is calculated. It might be worthwhile to try another analysis where buffers are smaller or multiple sized buffers are assessed.
2. A major issue for creating the habitat suitability map was the resolution of the base raster layers for each parameter, especially the substrate layer. The current layer does not reflect the nuances of small patches of available substrate and therefore is not a true indicator of available substrate for oysters. The resolution of the substrate layer is not sophisticated enough to capture the true extent of the oysters, because they’re known to be opportunistic in finding suitable substrate.
3. The area covered by each field survey (5-min search at a randomly selected starting point) was not equal at every location. Difficulty of travel at the site impacted the rate at which the survey could be conducted. For example, extremely muddy locations or locations with a lot of large boulders made traversing through the survey site more challenging and slowed the survey team down. Additionally, more ground could be covered if little substrate was available for examination.
4. During field surveys, no distinction was made for adult or juvenile oyster, or living or dead, for determining presence/absence. Separating out this information on the map might be interesting in determining where oysters are more likely to survive long-term. For example, it has been noted that young oysters can opportunistically settle on substrates under less ideal environmental conditions, but may not be able to survive into adulthood.
5. The simplistic approach for ranking the subcategories within each habitat parameter (on a scale from 1-4) ensured that the extreme ends of the rank value were always represented; there may not always be a definitive 1 or 4 value within each category. For example, maybe high salinity and low salinity both rank as 2’s rather than deciding one is slightly more appropriate than the other.

References:

Beck, M. W., R.D. Brumbaugh, L. Airoldi, A. Carranza, L.D. Coen, C. Crawford, O. Defeo, G.J. Edgar, B. Hancock, M.C. Kay, H.S. Lenihan, M.W. Luckenbach, C.L. Toropova, G. Zhang, X. Guo. 2011. Oyster Reefs at Risk and Recommendations for Conservation, Restoration, and Management. BioScience 61: 107-116.

Groth, S. and S. Rumrill. 2009. History of Olympia Oysters (Ostrea lurida Carpenter 1864) in Oregon Estuaries, and a Description of Recovering Populations in Coos Bay. J. Shellfish Research 28: 51-58.