GEOG 566

         Advanced spatial statistics and GIScience

April 5, 2018

Understanding trends in and environmental drivers of bull kelp abundance

Filed under: My Spatial Problem @ 10:30 am

Research Question:

Along the pacific coast of North America, Nereocystis luetkeana is the dominant canopy forming kelp from central California up to Alaska (Springer et al., 2006). This annual seaweed is subject to extensive swings in population size (up to two orders of magnitude) from year to year. This study aims to identify potential environmental drivers of the substantial observed interannual variability in N. luetkeana populations along southern Oregon and northern California. I want to compare trends in population size and correlates of population size in southern Oregon and northern California because there is a biogeographic break between the two areas, which may results in differential environmental control of kelp population size between the two regions.

Supervised classification and multiple endmember spectral mixing analysis were used to quantify canopy area and density of N. luetkeana from Landsat imagery over the past 30 years. Area and density will then be correlated with several environmental conditions expected to influence N. luetkeana population size, including water temperature, nutrient availability, wave height, climate oscillations, and urchin population size.

Datasets I will be using:

I will eventually be using a dataset of N. leutkeana canopy extent from northern California to southern Oregon provided to me from Dr. Tom Bell. This dataset covers 1985-2018 and is derived from 30m resolution Landsat data. These data will not have a set sampling routine, because clouds, tide, waves and malfunctioning sensors often make measurement of the kelp canopy impossible. Additionally, there are only a few months of the year when bull kelp reaches the surface and is quantifiable via remote sensing (roughly July – November). Therefore, temporally these data will be patchy, catching different areas at different points of the growing season in different years. Spatially, although the Landsat data is at 30m resolution, my collaborators and I use Multiple Endmember Spectral Mixing Analysis to estimate kelp percent cover within each 30m Landsat pixel. This improves our grain, but we are still limited to observing larger, more persistent patches of N. luetkeana.

I do not currently have the data, so I have generated a hypothetical stand in data set that has a similar temporal/spatial attributes.

As far as the independent variables I am looking to examine, I have:

  • Temperature data (offshore and onshore) going back to roughly 2004 (source: National Buoy Data Center and PISCO)
  • Nutrient data going back to ~1995 (source: PISCO
  • ENSO, PDO, and NPGO indices going back to the 1980s
  • Upwelling strength going back to the 1980s (source: NOAA)
  • Urchin abundances going back to the 1980s for Oregon. However, as of now this information is sparse. I currently only have information on fishing effort, rather than direct surveys of urchin abundance. When I do have surveys of urchin abundance they are extremely spatially and temporally limited.


I have two sets of hypotheses. One is my initial hypotheses, which I generated before looking at any of the kelp population data. These were that:

  1. Kelp populations in OR would have increased from about 1990-2000 due to urchin overfishing
  2. Kelp populations would have then began decreasing from about 2000 onwards due to the potentially recovery of the urchin populations and increasing SST due to climate change.
  3. California would not have seen as much variation in the kelp populations because urchin fishing did not go through a boom and bust cycle there the way it did in Oregon.

After a little initial data exploration, my more informed hypotheses are that:

  1. Kelp population size will be decreasing over time due to increasing water temperatures, decreasing nutrient availability, and increasing frequency of violent storms.
  2. Summer/fall wave height, temperature and nutrients, and upwelling will have the strongest effects on kelp population size.
  3. Climate oscillations will have weaker effects than those variables previously mentioned.
  4. Kelp population size in Oregon to be more related to upwelling strength since upwelling tends to be more variable in Oregon than in northern California. Conversely, I expect that urchin density will be more correlated with kelp population size in California than in Oregon since Oregon’s kelp populations have not collapsed in response to sea star wasting disease (and the resulting explosion in urchin populations) the way that California’s have.


I’m looking for a lot of help on this front. I’m hoping to use some analyses that will help me pick out temporal trends in kelp extent/density over time at least two spatial scales (coastwide and statewide). This might include linear regression and wavelet analysis. Furthermore, I hope to find the correlations between environmental variables and kelp extent/density at several spatial scales (coastwide, statewide, and sub-state). However, I don’t know how to go about this. My advisor feels that multivariate linear regression would be all I needed to do this. However, I have seen other researchers use methods such as: empirical orthogonal functions, generalized additive models, regression tree analysis, etc. I am overwhelmed by all the potential statistical methods and don’t know how to go about evaluating their relative merits. Also, I would like help figuring out how to quantify and deal with the high levels of correlation between many of the environmental variables.

Expected outcome:

I would like to be able to say something about temporal trends in kelp population size. Additionally, I want to produce models at 2-3 spatial scales that correlate various environmental factors with kelp abundance, and to identify whether those models differ substantially between areas and scales.


Kelp plays a number of roles in coastal ecosystems. It is a prolific primary producer, provides food, substrate, and habitat for many organisms including commercially important species, can modify wave impacts on shorelines, transports nutrients and carbon to both terrestrial ecosystems and aphotic marine ecosystems. On the pacific coast of North America, the dynamics of Macrocystis pyrifera (giant kelp) has been well-studied for decades, while those of Nereocystis luetkeana (bull kelp), the dominant canopy-forming kelp from Santa Cruz, California to Alaska, are far less understood (Springer et al., 2007). Kelp populations worldwide are declining (Krumshal et al., 2015). Understanding what environmental factors correlate with and potentially drive kelp abundance will fill a hole in our understanding of kelp dynamics along the northern pacific coast of North America and will help scientists predict the future dynamics of populations that depend on kelp for food and habitat.

Furthermore, this question is also of interest to resource managers. Bull kelp supports a variety of commercially important fisheries such rockfish and uni (sea urchin gonads), so understanding trends in kelp populations will help managers plan for fishery futures. There is an effort to reintroduce sea otters to Oregon. As sea otters frequently dine on urchins, which in turn are major grazers of kelp, kelp availability may help determine the feasibility of reintroduction. Kelp beds absorb tremendous amounts of wave energy, which is of growing importance to coastal communities battling sea level rise and increasingly violent storms.

Your level of preparation:

I am very familiar with Arc and am usually able to do just about anything with assistance from the Internet. I have used Modelbuilder and GIS programming in Python in the past, but my Python skills are limited and self-taught. I use R regularly but am less comfortable using it for spatial analyses than Arc.


Print Friendly, PDF & Email

1 Comment

  1.   jonesju — April 9, 2018 @ 8:20 am    

    hi Sara,
    Thanks for the thoughtful blog post. As a starter, I suggest that you create a GIS layer of kelp coverage for a particular time period (or several). You will have to address temporal and spatial resolution: what is the frequency of kelp measurements over time, and what is the grain of the spatial information? You will need to attribute the data layer with some information about kelp – will this be density? percent cover? Once this is done we can think about analyses to describe or characterize the data.

RSS feed for comments on this post.

Sorry, the comment form is closed at this time.

© 2019 GEOG 566   Powered by WordPress MU    Hosted by