Supporting Online Material for
Strong Top-down Control in Southern California Kelp Forest
Benjamin S. Halpern,* Karl Cottenie, Bernardo R. Broitman
*To whom correspondence should be addressed. E-mail: firstname.lastname@example.org
Published 26 May 2006, Science 312, 1230 (2006)
This PDF file includes:
Materials and Methods
Tables S1 and S2
Supporting Online Material
The U.S. National Park Service has been conducting annual surveys of algal,
invertebrate, and fish abundances at 16 different kelp forest sites around the five islands
in the Channel Islands National Park since 1982 as part of the Kelp Forest Monitoring
Program (KFMP; (S1)). Here we focused on data from the KFMP for only four years,
1999-2002 (for our purposes, a year runs from July the previous year to June of the
current year), as these were the only years for which oceanographic data were available
(see below), and only for the 4 northern Channel Islands (Fig. 1). Abundance data were
available for 49 species, with trophic classifications of these species as elsewhere (S2).
Three detritivorous invertebrate species were excluded from analyses. Full details on
how these data were processed prior to analyses are provided elsewhere (S3).
Spatial, temporal, and environmental data were also used in analyses, as described
elsewhere (S3). The KFMP recorded hourly temperature at each site (environmental
variable), latitude and longitude of each sites was used to calculate spatial variables, and
4 potential temporal patterns were modeled using Principal Components of Neighboring
Matrices (PCNM) techniques (S4) to represent temporal variables.
We estimated primary production around the KFMP monitoring sites from
satellite-based observations of chlorophyll-a biomass (mg m-3) from the Sea-viewing
Wide Field-of-view Sensor (SeaWiFS) between September 1997 and October 2002. The
quality of SeaWiFS observations in coastal waters off California has been extensively
validated ((S5), see also (S6)), and primary production around the islands has been shown
to be negatively correlated with SST (S7), which in turn is strongly and negatively
correlated with nutrient levels (S8). Thus primary production is a strong, albeit indirect,
metric for local nutrient levels. The satellite observations used in the present analyses
were collected at a nominal resolution of 1.1 km, such that each of the KFMP sites was
assigned to ~1km2 pixels, and averaged to monthly means. From the satellite composite
we then calculated the long-term annual (July to June) and winter (January to March)
means and standard deviations for each site.
We did not include data on fishing effort (ultimate top-down control) as high-
resolution fishing data for the Channel Islands (species and location-specific catch rates)
do not exist. Furthermore, recreational and commercial fishermen target species in all
trophic levels (e.g., giant kelp, urchins, rockfishes and kelp bass) and only heavily target
one species (lobster) from those identified in the forward selection procedure, and so
humans act as herbivores and primary, secondary and tertiary predators in this system. In
theory, although fishing could mask bottom-up regulation of the kelp forest communities
by harvesting any changes in biomass distribution among trophic levels that resulted from
differences in primary production, this is not likely the case around the Channel Islands.
Recreational fishing around the islands is heterogeneously distributed in a similar pattern
to primary production, since the warmer-water islands (Anacapa and Santa Cruz Islands)
are also much closer to mainland harbors, and commercial fishing primarily targets
market squid and sea urchins. If anything, fishing pressure should accentuate differences
in biomass distribution among trophic levels between low and high-productivity systems
if productivity were the main driver of those differences.
Variation decomposition analysis (VDA) is a statistical technique that partitions
the amount of variation explained by each variable (S9). The technique produces an F-
statistic that is an asymptotically pivotal reference statistic and compares the amount of
explained variation with the residual error (scaled for the appropriate degrees of
freedom). The significance of the F-statistic is obtained through permutations of the data,
and the degrees of freedom are not used directly for the computation of the p-values
associated with the F-statistics (as is common practice with RDA results; (S9). To
investigate the unique amounts of variation of a particular trophic level explained by top-
down versus bottom-up processes, we restricted our dependent species matrix to contain
only the species abundances of that particular trophic level. We also included data on
and accounted for the location (longitude and latitude were transformed into third-degree
spatial polynomials, creating 9 spatial variables), temporal patterns in the data (12
Principal Coordinates of Neighboring Matrices variables; (S4), and a variety of
environmental variables (regional wave height, ENSO index values, local temperature;
see (S3) for additional descriptions of all three types of variable). Our aim in including
these latter variables was to remove their influence on species abundances and isolate
true top-down and bottom-up effects.
S1. Davis G.E., D.J. Kushner, J.M Mondragon, J.E. Mondragon, D. Lerma, and D.V.
Richards. 1997. Kelp forest monitoring handbook volume 1: sampling protocol.
National Park Service, Channel Islands National Park, Ventura, California.
S2. Micheli, F. and B.S. Halpern. 2005. Low functional redundancy in coastal marine
assemblages. Ecol. Lett. 8: 391-400.
S3. Halpern, B.S. and K. Cottenie. in press. Little evidence for climate effects on local-
scale structure and dynamics of California kelp forest communities. Global Change
S4. Borcard, D., P. Legendre, C. Avois-Jacquet, and H. Tuomisto. 2004. Dissecting the
spatial structure of ecological data at multiple scales. Ecology 85: 1826-1832.
S5. Otero, M.P. and D.A. Siegel. 2004. Spatial and temporal characteristics of sediment
plumes and phytoplankton blooms in the Santa Barbara Channel. Deep-sea Research:
Studies in Oceanography 51: 1129-1149.
S6. Ware, D.M. and R.E. Thomson. 2005. Bottom-up ecosystem trophic dynamics
determine fish production in the Northeast Pacific. Science 308: 1280-1284.
S7. Blanchette, C.A., B.R. Broitman, and S.D. Gaines. 2006. Intertidal community
structure and oceanographic patterns around Santa Cruz Island, California. Marine
Biology: DOI 10.1007/s00227-005-0239-3.
S8. Dayton P.K., M.J. Tegner, P.B. Edwards, and K.L. Riser. 1999. Temporal and spatial
scales of kelp demography: the role of oceanographic climate. Ecological
Monographs 69: 219-250.
S9. Legendre L and P. Legendre. 1998. Numerical ecology. Elsevier, New York.
Table S1. List of all species included in analyses. Species are grouped by trophic levels.
Eisenia arborea, Pterygophora californica, Laminaria farlowii, Macrocystis pyrifera
Strongylocentrotus purpuratus, Strongylocentrotus franciscanus, Lytechinus anamesus,
Haliotis corrugata, Haliotis refescens, Haliotis fulgens, Lithopoma undosum, Megathura
crenulata, Aplysia californica
Crassedoma giganteum, Stylaster californica, Urticina lofotensis, Corynactis californica,
Balanophyllia elegans, Serpulorbis squamigerus, Astrangia lajollaensis, Lophogorgio
chilensis, Muricea fruticosa, Tethya aurantia, Diaperoecia californica, Phragmatopoma
californica, Dioptra ornata, Styela montereyensis
Chromis punctipinnis, Sebastes mystinus
Pisaster giganteus, Pycnopodia helianthoides, Kelletia kelletti, Panulirus interruptus
Sebastes atrovirens, Embiotoca jacksoni, Embiotoca lateralis, Oxyjulis californica,
Damalichthys vacca, Hypsypops rubicundus, Alloclinus holderi, Rhinogobiops nicholsii,
Lythrypnus dalli, Semicossyphus pulcher
Sebastes serranoides, Paralabrax clathratus
Table S2. Results from variation decomposition analyses for the relative strength of top-down versus bottom-up forces in controlling Download full-text
different trophic levels. Results indicate percentage of variance explained for each variable when controlling for all other potential
variables. Degrees of freedom, F-statistics, and p-values are derived from VDA tests, as described in the Methods.
All predators Primary predators only Secondary predators only
Herbivores + Planktivores