SIMoN
  Sanctuary Integrated Monitoring Network
Monitoring Project

Spatial and temporal variability of kelp forest canopies in central California

Principal Investigator(s)

  • Michael Donnellan
    Moss Landing Marine Laboratories, California State University
Start Date: September 01, 2000
End Date: June 30, 2004

Recent advances in computer hardware and software have spurred a renewed look at historical datasets of kelp canopies. Using a time series of aerial photographs spanning 65 kilometers and 6 years, I described the spatial and temporal patterns of kelp canopy coverage in Macrocystis-dominated kelp forests offshore of Monterey, California. The principal findings of this work were that: 1) canopy dynamics were much more predictable in central California than previously thought; 2) the size of the spatial window through which temporal patterns of canopy abundance are perceived is a critical determinant of the observed results; 3) canopies exhibited typical 'patch sizes' of approximately 1.6 kilometers, suggesting that an important process or processes occur at a similar scale; and 4) kelp forests may be classified by their canopy dynamics over large spatial scales using time series of remotely sensed images, provided the classification scheme is validated by focused in situ work.

Summary to Date

This study was unique in the description of both temporal and spatial patterns of kelp forest canopies. I have attempted to lay descriptive groundwork that will be helpful for generating hypotheses to explain underlying causes of variation, designing studies and monitoring programs (and statistical considerations), and putting past and future studies into spatial and temporal context. I identified several spatial and temporal patterns structuring kelp canopies, and possibly the kelp populations and communities with which they are associated. However, it was beyond the scope of this study to deduce the mechanisms for these patterns due to lack of available data (e.g., wave exposure), accessible spatial analytical techniques, and time. Process-oriented correlative and experimental work is left to subsequent investigations.


Large-scale, low-frequency processes (e.g., El Niņos, major storms, global warming) are not amenable to manipulative experimentation, therefore it is imperative that lengthy time series be compiled to observe decoupled processes with sufficient replication to obtain meaningful results and conclusions. Statistical techniques for time series analysis typically require more-or-less consistently-spaced replication with a sample size of at least 30-50, a considerable investment in time, effort, dedication, and financial resources when canopy surveys must be replicated on an annual basis. As illustrated by the major change in the distribution and relative abundance of canopy-forming kelps in the Monterey Bay area during the previous 50 years, some of the most interesting processes occur over broad temporal and spatial scales. Remote sensing is an opportune, sustainable strategy to monitor such changes, ideally in conjunction with complementary in situ work.

Monitoring Trends

  • The size of the spatial window over which temporal patterns of canopy abundance were summarized was a critical determinant of the observed results. At the spatial scale of the entire study area, kelp canopies followed a highly regular, seasonal pattern. Canopies attained maximum development between late July and September of each year and the maximum surface area observed ranged from 979 to 1,185 hectares (mean = 1117 hectares, coefficient of variation = 0.09) from 1986 through 1989. Minimum canopy abundance occurred between February and March each year and ranged from 103 to 221 hectares (mean = 155 hectares, coefficient of variation = 0.35). In contrast, when the temporal abundance of kelp canopies were summarized over small spatial scales, maximum canopy occurred at least once during each calendar month except February. In comparison, minimum canopy occurred during every month of the year, with most occurrences from December - May.
  • Approximately 72% of the variability in the spatial kelp canopy data were spatially structured, leaving 28% of the variability was unexplained (e.g., measurement and georeferencing error, random noise, historical influences, biological interactions). The value of the range parameter (Ao) in variogram analysis was 804 meters, which indicated: 1) the distance beyond which sampling units were no longer spatially correlated; and 2) the patch size of the spatial structure was approximately 1,600 meters in length. The patch size indicated the distance over which an unknown but important process (spore dispersal?) was occurring.
  • Cluster analysis of temporal canopy abundance for 62 defined "beds" within the study area indicated that kelp beds could be classified into either 3 or 7 natural groupings. The success of this technique suggests remote sensing can be used to classify kelp forests into certain "types", although in situ work is required to evaluate its validity.

Study Parameters

  • Disturbance
  • Distribution
  • Stock assessment

Study Methods

Fifty-eight aerial surveys of kelp canopies were done in the study area from November 1985 - December 1991. Surveys were approximately monthly from 1985 - 1989, bimonthly in 1990, and quarterly in 1991. Imagery interpreted to be algal canopy was traced onto maps, scanned into digital format, and spatially registered using a Geographic Information System (GIS). The resultant time series of kelp canopies was analyzed visually and statistically (e.g., non-linear regression), and summary statistics were compiled for timing of annual maximum and minimum canopy abundance. To complement the analysis of temporally explicit, spatially implicit patterns, I determined spatial structuring of kelp canopies using variogram analysis of spatial autocorrelation. To accomplish this, a composite image of all canopy surveys combined (i.e. canopy "persistence") was created, and a digital transect of persistence values was extracted and input into variogram analysis. To assess whether kelp forests could be classified according to their temporal patterns of canopy abundance, kelp beds were subjectively delineated using the persistence GIS layer and classified via cluster analysis (i.e. sites x times).


Figures and Images

Figure 1. Giant kelp (Macrocystis pyrifera) is the dominant canopy-forming kelp along the central coast, particularly in more sheltered sections of the coastline.


Figure 2. Bull kelp (Nereocystis luetkeana) also forms canopies along the central coast of California.