Overview of commercial fishing spatial data modeling exercise

illustration of a fisherman holding out two large lobsters, blue and red

We developed a method to refine coarse-scale commercial fishing information using best-available spatial data to model fishing patterns in and around California’s MPAs before and after MPA network implementation.

Key Takeaways

As this MPA Human Uses study was conducted to evaluate sociological impacts from implemented MPAs, we found it valuable to also evaluate and model available landings fishing data provided by CDFW to approximate fine-scale spatial fishing patterns. While we assert that this modeling approach can be quite useful to determine fishing patterns, we also recommend that fishery managers establish methods and programs to gather actual data on finer-scale spatial and temporal scales, rather than relying on modeled datasets long term. 

To model this data, we transformed California commercial fishing data from 10×10 nautical mile area to 1×1 nautical mile area using spatial data collected during the development and implementation phases of California’s MPA network. The methods used here are exploratory and should be considered, as all models are, an approximation of reality.

We applied our model methodology to the California Spiny lobster, red sea urchin, and nearshore finfish fisheries and found our methods successfully transformed the California Commercial fishing landings information from coarse- scale into fine-scale micro-blocks at rates of 87% – 98% fidelity.

Background

We needed to break down the commercial fishing landings data collected by the California Department of Fish and Wildlife (CDFW) to assess it on a finer scale from port to port. While the original data  is useful for high level analyses, the spatial scale of these data is too coarse to evaluate the performance of specific MPAs. Most CDFW fisheries landings data are summarized to 10 x 10 nautical mile blocks, (100 square-nautical mile area) which is much larger than individual MPAs (avg. 6.5 mi2). 

We developed a method that provides a new way of looking at fishing patterns in and around California’s MPA network. By combining CDFW commercial fishing landings data with our fishery survey fine-scale data, we were able to create a new dataset that provides year-on-year spatial fishing effort data at a 1 x 1 nautical mile scale — a scale that is more appropriate for MPA performance evaluation studies. The resulting modeled spatial layers refine the coarse-scale fisheries data and provide a useful proxy using the best-available data for commercial fishing data.

Modeling Methods

To complete this spatial modeling process we used two data sources, including commercial fishing landings receipts from California Department of Fish and Wildlife (CDFW) in the years 2005-2020, as well as spatial layers of fishing areas developed by Ecotrust. This data was based on in-person interviews conducted during 2006-2012 to support  MPA network development and implementation processes.  

To develop spatial layers of fishing areas, we interviewed fishermen across the state and asked them to physically map the areas that each fisherman identified as important to their fishing operations through an exercise. For each fishery a fisherman  targeted, they were asked to distribute a hypothetical set of 100 pennies among their fishing grounds to give each area a weighted importance. For example, a fisherman could draw one 100-penny area or one hundred 1-penny areas. 

After completing this exercise, we applied an economic value based on an average of the five most recent years of fishing. For example, if an area was assigned a relative importance of 60 pennies by a fisherman targeting nearshore finfish, we applied 60% of their gross landing receipt total for that fishery to that area. We then aggregated this information across fishermen and fisheries to create a dataset that provided a representation of the relative importance of any particular fishing area. Here we assumed that the more important fishing areas also yielded higher ranges of fish pounds landed. 

The data were then aggregated together and a map was produced showing the relative importance of the fishing grounds, by fishery-port complex and by fishery in the regions designated by the Marine Life Protection Act (MLPA) process (North Coast, North Central Coast, Central Coast, and South Coast).

Application of modeling method to MPA context – assumptions, caveats, nuancesillustration of a school of small fish swimming

Time-scales, pre- and post-MPA implementation

For this modeling effort, we defined “pre-MPA years” as 2005-2009 and “post-MPA years” as 2010-2020, in order to  simplify the model. We recognize that MPAs were implemented in a phased approach by region. This methodology can be applied to other years, however since this effort was focused on methods development and Ecotrust’s spatial data were collected within this timeframe, we determined these years were the most appropriate for modeling.

Focus on non-migratory species

We focused this modeling effort on non-migratory species for several reasons. We assume that fishing patterns are mostly consistent year-to-year, that fishing effort is focused  on areas that yield the highest success, and that non-migratory species persist in generally the same locations year to year. Therefore, we assume effort is consistent in a specific location over time. Conversely, migratory species range over large areas, thus the locations of fishing effort  changes over time. So we assert that generally, non-migratory species are the best suited for this modeling approach.

Fishery groupings to single species

In our fishermen interviews for the MPA implementation phase, we collected spatial data in fishery groupings that align with CDFW fisheries management, though part of our goal for this modeling was to produce spatial layers that represent single species. We included the single species modeling to support the other long-term MPA monitoring projects that are part of this decadal review. After testing the model outputs, we feel confident this methodology can be applied to single species, assuming certain caveats. For example, the landings data for black rockfish is highly spatially coincident between CDFW and Ecotrust data and the species represents 27% of the total fishery. This resulted in an average redistribution of 92% over 16 years for balck rockfish. In other words, Black rockfish was successfully redistributed because it represents a large portion of the fishery and the Ecotrust interview data included those fishers who target this species. 

It should be noted that there were two species where this methodology did not work well:  California Scorpionfish and Quillback Rockfish, likely due to these species representing only 1% of the total nearshore finfish fishery, and likely due to a spatial mismatch between the Ecotrust data and the reported are in the CDFW data. We recommend that other investigators using this modeling method should verify model outputs of single species redistribution.

See report for full findings.

illustration of rockfish

Assessing modeling method success and shifts in fishing patterns over time

To assess the outputs of this spatial modeling approach, we aimed to answer two primary questions: (1)  how well did this approach work in transforming landings data from the 10nm x 10nm blocks to the 1nm x 1nm micro-blocks; and (2) what does this tell us about fishing patterns in and around California MPAs?

Determining successful transformation of landings data
Our results show a high-level of fidelity when compared between the total pounds per year summarized to the 10nm blocks and the output of our analysis. The average percentage of pounds transformed for lobster, urchin, and nearshore finfish (all species) is 97%, 98%, and 87% respectively. We view this as a successful transformation of the data because it shows both a spatial match and the capacity to redistribute the pounds within the fishing areas.

The individual nearshore finfish species also refactored with moderate to high fidelity to the original landings data. Out of 17 species modeled, 15 captured an average 60% or higher of pounds landed; and 10 out of those 15 had 70% or higher average.

Analyzing shifts in spatial fishing patterns over time, from pre- to post- MPA implementation

When we look at the transformed/modeled dataset of fine scale spatial landings data, we see an increase in the percentage of pounds harvested in the micro-blocks immediately adjacent to the State Marine Reserves. For lobster, these micro-blocks averaged 10% of catch in pre-MPA years and increased to 13% in the post implementation years. The same level of increase can be seen for urchin, averaging 19% in pre-MPA years increasing to 22% in post-MPA years. But for nearshore finfish (all species) the increase is greater. In pre-MPA years the average catch in the adjacent areas was 7% but increased to 15% in the post-MPA years.