The Changing California Coast: Relationships Between Climatic Variables and Coastal Vegetation Succession

EarthzineArticles, Climate, Earth Observation, Ecosystems Theme, Forest Resource Information Theme, Original, Sections, Themed Articles

Study area, four counties along the central California coast.

Wei-Chen Hsu1, University of California at Berkeley;

Alex Remar, University of Delaware;

Emily Williams, University of California at Santa Barbara;

Adam McClure, San Francisco State University;

Soumya Kannan, Gunn High School;

Robert Steers, Ph.D., National Park Service, Inventory and Monitoring Program;

Cindy Schmidt1, Bay Area Environmental Research Institute;

J.W. Skiles, Ph.D., NASA Ames Research Center.

1DEVELOP National Program, NASA Ames Research Center, M.S. 242-8, Moffett Field, California 94035,

Abstract ‰ÛÓ The role of the NASA Applied Sciences DEVELOP National Internship Program is to use NASA satellite missions to apply Earth-based questions in collaboration with local, state, and federal agencies. One recent project focused on the use of the Landsat-5 TM satellite images to assess the general vegetation distribution pattern and areas of vegetation change along the California coast. The land-ocean interface along the central coast of California is one of the most diverse biogeographic regions in the state. This area is composed of a species-rich mosaic of coastal grassland, shrubland, and forested vegetation types. An increase of tree colonization into shrublands along this coast has been documented at several study sites in the past. Our study used Landsat-5 TM satellite images from the years 1985 and 2010 to assess the distribution of grass, shrub, and forest vegetation in the study area, identify areas where grassland, shrubland, and forested land cover has changed, and further correlate those changes with various climatic variables that were associated with water availability (i.e. precipitation, evapotranspiration (ET), and fog frequency). Our results showed that forest vegetation was more abundant in the northern region and regions with high mean annual precipitation and high mean summer ET. Forests have expanded in places with relatively high annual precipitation and may be influencing increased summer ET. As land management agencies such as the National Park Service and U.S. Forest Service work to protect the natural resources of this region, understanding patterns of vegetation change and related climatic factors are critical, especially in light of future climate change projections that could profoundly alter vegetation patterns and processes.

1. Introduction to NASA DEVELOP

The DEVELOP Program sponsors paid human development internships at NASA centers including Ames Research Center for students to extend science research results to local communities. Student interns work on Earth science research projects, mentored by science advisors from NASA and partner agencies, to extend research results to local communities. The NASA Ames DEVELOP Program hosts graduate, undergraduate, and high-school students in a 10-week-long summer internship. Students are involved in a wide range of projects such as vegetation mapping, air quality assessments, wetland restoration, water resource management, and water biogeochemical cycling. The DEVELOP program is ‰ÛÏstudent run, student lead‰Û and is supported by science advisors and mentors.

(See also ‰ÛÏNASA Sensors used to Analyze Patterns of Change along the California Coast‰Û)

2. California‰Ûªs Coastal Vegetation

Vegetation succession is defined as the temporal change in species composition and structure that occurs as new species colonize a site [1]. Vegetation succession can be divided into primary and secondary succession. Succession that occurs on newly founded substrates, such as recent lava flows, is called primary succession. Secondary succession occurs on substrates with already established vegetation communities and soil [2]. During the initial stages of succession, species diversity (or structural complexity) generally increases; this process is known as progressive succession. Eventually, species diversity or structural complexity may decline as some species are displaced by others during later stages of succession; this process is known as retrogressive succession [3].

Study area, four counties along the central California coast.

Figure 1: Study area, four counties along the central California coast. Sources: Ames DEVELOP National Program, May 2012. County shapefiles obtained from

Along the Central Coast of California, secondary succession from grassland to shrubland and from shrubland to forest have been documented [4] [5] [6] [7] [8] [9] [10] [11]. Succession can be influenced by disturbances such as grazing, fires, and resource extraction [1] [11]. Grazing has controlled shrub species that would have otherwise colonized grassland, thus preventing succession from taking place [4] [12] [6]. More recently, grazing intensity has decreased in certain areas, allowing rangelands to convert back to shrublands [9]. Additionally, region-wide fire suppression has resulted in long, fire-free periods, allowing forests to colonize shrublands [13] [14].

Two variables associated with faster rates of succession are water availability and soil fertility [15]. On the central coast of California, the influence of water availability and soil fertility on succession rates is assumed to be positive but has not been investigated previously. Water availability is affected by precipitation, fog, cloud cover, temperature, elevation, aspect, slope, and soil type, among other variables. For example, fog can be an important contribution to the water budget along the central coast of California. Coastal forests often depend on fog drip, where fog condenses on the surface of the leaves and drips as liquid water, acting as an additional water source during the otherwise dry and rainless California summer [16] [17] [18] [19] [20] [21]. In addition, fog and cloud cover reduces evapotranspiration (ET) and lowers ambient temperatures through reduction of overall ecosystem water loss [19].

Our study focused on forest (conifer, hardwood, and woodland) transitions along the central coastal area from Sonoma to Monterey counties (Figure 1). The climate in this study area is Mediterranean-like and strongly influenced by the offshore marine environment [21]. The topography of the study area is dominated by the Coast Ranges, which are divided into the North Coast Range and the South Coast Range by San Francisco Bay. Thus, most of the study area is hilly or mountainous, although several large valleys exist, including the Santa Rosa Plain, the San Francisco Bay, and the Salinas Valley, from north to south, respectively.

This study used satellite imagery and remote sensing techniques to: 1) quantify the amount and distribution of grass, shrub, and forest vegetation; 2) identify areas of vegetation transition between the years 1985 and 2010; and 3) investigate the relationship between climatic variables and forest vegetation type distribution and succession. We hypothesized that areas with greater water availability would have more forest cover and a greater percent of progressive vegetation succession of shrublands into forests. As land managers work to protect the natural resources of this region, understanding patterns of vegetation change and dynamics are critical in land conservation and management.

Water variables analyzed in this study including annual precipitation (A), and summer evapotranspiration (B).

Figure 2. Water variables analyzed in this study including annual precipitation (A), and summer evapotranspiration (B) {24}. Sources: PRISM for precipitation, and MODIS for evapotranspiration.

3. Methods

Landsat-5 TM and MODIS imagery from the years 1985 and 2010 were used for our study. The 1985 imagery was used as a baseline, while combined with the 2010 imagery; the changes over the 25 year period were quantified. For each time period, a total of three Landsat-5 TM scenes were used to cover the entire study area that corresponded to the summer/fall months of September and October when visual differences among the major vegetation types are assumed to be greatest.

Remotely sensed imagery and land cover classifications must be calibrated and checked with ground-control points for image verification. We randomly generated ground-control points that were stratified between areas of change (based on a preliminary change detection analysis) and areas of no change. We also visited 34 points during our field campaign to verify vegetation classification. At each point in the field, percent vegetation for shrub and forest were recorded. Forested vegetation included conifer, hardwood forests, and woodlands. The percent cover was estimated based on a qualitative visual estimation within a 30-meter radius from the generated point.

We performed a land cover classification on the 1985 and 2010 Landsat-5 TM images to identify the distribution of grass, shrub, and forest land cover for our entire study area. Using ERDAS Imagine and ArcGIS, we were able to further quantify the area and percentage of grass, shrub, and forest cover that had changed between 1985 and 2010.

Vegetation classification of the 1985 image (left) and 2010 image (right).

Figure 3. Vegetation classification of the 1985 image (left) and 2010 image (right) {24}. Source: Landsat-5 TM images.

We investigated five climatic variables: 1) mean summer fog frequency, 2) mean summer ET (mm), 3) mean annual precipitation (mm), 4) mean annual maximum temperature, and 5) mean annual minimum temperature (å¡C). Mean annual precipitation and summer ET are shown in Figure 2.

The proxy used to determine fog frequency was the average summer cloud cover fraction, derived from MODIS imagery and compiled by the University of California, Santa Barbara [19]. The mean summer ET data were derived from the MOD 16 global ET dataset that was calculated based on [22], which includes multiple MODIS products such as land cover type (MOD12Q1), FPAR/LAI (MOD15A2), and albedo (MCD43B2/B3) products. The precipitation dataset was obtained from the PRISM Climate Group at Oregon State University. The PRISM datasets were averaged from 1971-2000. We used the averaged annual 1971 and 2000 downscaled PRISM data for each ecoregion following U.S. Environmental Protection Agency (EPA) level 4 boundaries [23]. This data offers higher spatial resolution (800 m pixel size) compared with the alternative 4 km resolution of the monthly PRISM dataset. This provided us with an average precipitation value for each ecoregion to better understand vegetation transition for our study area. While the climatic variables might co-relate, we investigated how each variable independently related to vegetation succession to qualify differences between vegetation types.

Percent of 2010 vegetation cover for the total study area, the northern region, and the southern region with three vegetation types shown.

Figure 4. Percent of 2010 vegetation cover for the total study area, the northern region, and the southern region with three vegetation types shown {24}. Source: Landsat-5 TM images.

To better understand how climatic variables are associated with vegetation cover and transition, we divided the entire study area into 33 ecoregions. We further divided the 33 ecoregions into the north (North Coast Range) and the south (South Coast Range) to examine how vegetation changes differed between these two distinct regions. Lastly, succession within northern and southern regions was correlated to fog frequency, ET, and precipitation using a linear regression method as a simplistic, general model.

4. Results

Vegetation land cover types from both 1985 and 2010 were successfully classified using Landsat-5 TM images (Figure 3) [24]. Based on the 2010 imagery for the entire study area, 74.36 percent is covered by vegetation, with the most common vegetation cover type as forest, followed by shrub and grass (Figure 4) [24]. When analyzing the data by north versus south, forested vegetation had more areal coverage in the north while the south has higher shrubland areal coverage.

Both the location and amount of succession or vegetation type transition along the California coast were investigated in this study. Amount of succession also was calculated for the northern and southern regions of the study area, separately. For the entire study area, we found that between 1985 and 2010, approximately 19 percent of the study area had experienced succession, with the northern region experiencing more succession than the southern region [24]. The most common type of vegetation change was reforestation, where there was a net increase of forest and a decrease in shrublands (Figure 5) [24]. Additionally, there was a greater increase in forested land cover in the north (6.63 percent) compared to the south (0.57 percent) (Figure 5). Overall, grassland coverage did not change greatly during the past 25 years [24].

Percent of vegetation change for the total study area, the northern region, and the southern region with three vegetation types shown.

Figure 5. Percent of vegetation change for the total study area, the northern region, and the southern region with three vegetation types shown {24}. Positive values denote a percent gain by the specified vegetation type and vice-versa.

To better understand the relationship between climatic variables and vegetation distribution throughout our study area, we first examined general differences in precipitation, fog, ET, Tmax, and Tmin between grass, shrub, and forest cover in 2010. Both mean summer ET (R2= 0.71, p-value = 7.7 x 10-9) and mean annual precipitation (R2 = 0.53, p-value = 1.41 x 10-6) were highly correlated with forest distribution [24]. Shrubland cover was negatively correlated with minimum temperature (Tmin) although this was a relatively weak relationship (R2 = 0.13, p-value = 0.03651) [24]. Grassland distribution was negatively correlated with mean summer ET (R2 = 0.368, p-value = 4.85×10-4), both annual Tmin (R2 = 0.46, p-value = 1.25×10-5) and annual maximum temperature (Tmax) (R2 = 0.2998, p-value = 9.74×10-4) [24].

Additionally, climatic variables and succession type such as grassland to shrubland, shrubland to forest, and forest to shrubland, were investigated. Conversion from forest to shrubland was positively associated with mean summer fog (R2 = 0.20, p-value = 0.0086), conversion from shrubland to forest was positively associated with mean annual precipitation (R2 = 0.42, p-value = 4.61 x 10-5) and ET (R2 = 0.24, p-value = 0.0072), and conversion from forest to grassland was positively associated with annual Tmin [24]. Net change to grassland was negatively correlated with summer ET (R2 = 0.25, p-value =0.00557). Net change to shrublands was negatively correlated to annual precipitation (R2 = 0.36, p-value =2.15×10-4) and summer ET (R2 = 0.25, p-value =0.00557), and net change to forested vegetation was positively correlated to both annual precipitation (R2 = 0.27, p-value = 0.00193) and summer ET (R2 = 0.20, p-value = 0.0129) [24].

This study revealed that forested vegetation is positively associated with water availability, as measured by annual precipitation, and that the North Coast Range has the most extensive spatial forest coverage [24]. Furthermore, when analyzing the correlation between climatic variables and succession, conversion to forest was most strongly associated with places that experienced relatively high annual precipitation and summer ET. While in our study region, shrublands readily converted to forested vegetation in areas that were relatively moist, studies in shrublands from drier, southern California mountains and hills have shown no successional trends toward woodland or forest vegetation types [25], although this much of the difference also could be attributed to higher fire frequency in southern California. Recently, shrublands along the desert edge of southern California have experienced die-back (a higher mortality rate) and elevational shifts upslope in response to drying conditions since the 1970s [26]. It is likely that if future climate change creates drier climatic conditions on the central coast, then successional rates and the distribution of vegetation types will be dramatically altered.

5. Conclusions

Our study showed that it is feasible to use remote sensing techniques and satellite images to monitor forest vegetation distribution and dynamics along the California coast. Overall, water availability was found to be the most significant factor that contributes to the current vegetation distribution as well as vegetation succession. Specifically, we were able to determine that the percent forest cover along the Californian coast has a relatively strong correlation with annual precipitation and summer ET, and fog frequency might be another important water input for the ecosystem. Furthermore, succession from shrubland to forested vegetation was positively associated with these two variables. While it is less clear how future climate change will influence water availability in our study region, and furthermore, between the north versus south or between the fog belt versus inland areas, a more advanced climatic analysis should be performed. Future research could include wavelet and singular spectral analysis to provide a better understanding on how each climatic variable might contribute to particular vegetation responses in our study area.


[1] Connell, J. H., and R.O. Slatyer. (1977): Mechanisms of succession in natural communities and their role in community stability and organization. The American Naturalist, 111(982), 1119‰ÛÒ1144.

[2] Barbour, M. G., J. H. Burk, W. D. Pitts, F. S. Gilliam, and M.W. Schwartz. (1998): Terrestrial Plant Ecology, Third Edition. Benjamin Cummings, San Francisco, CA.

[3] Mueller-Dombois, D. and H. Ellenberg. (1974): Aims and Methods of Vegetation Ecology. John Wiley and Sons, New York.

[4] McBride, J. and H.F. Heady. (1968): Invasion of grassland by Baccharis pilularis. Journal of Range Management, 21, 106-108.

[5] McBride, J.R. (1974): Plant succession in the Berkeley Hills, California. Madrono, 22, 317‰ÛÒ330.

[6] Williams, K., R.J. Hobbs, and S.P. Hamburg. (1987): Invasion of an annual grassland in Northern California by Baccharis pilularis ssp. consanguinea. Oecologia, 72, 461‰ÛÒ465.

[7] Williams, K. and R.J. Hobbs. (1989): Control of shrub establishment by springtime soil water availability in an annual grassland. Oecologia, 81, 62‰ÛÒ66.

[8] Horton, T., T. Bruns, and V. Parker. (1999): Ectomycorrhizal fungi associated with Arctostaphylos contribute to Pseudotsuga menziesii establishment. Canadian Journal of Botany, 77, 93‰ÛÒ102.

[9] Keeley, J.E. (2005): Fire history of the San Francisco east bay region and implications for landscape patterns. International Journal of Wildland Fire, 14, 285‰ÛÒ296.

[10] Kennedy, P.G. and W.P. Sousa, (2006): Forest encroachment into a Californian grassland: examining the simultaneous effects of facilitation and competition on tree seedling recruitment. Oecologia, 148, 464‰ÛÒ474.

[11] Zavaleta, E.S., and L.S. Kettley. (2006): Ecosystem change along a woody invasion chronosequence in a California grassland. Journal of Arid Environments, 66 (2), 290‰ÛÒ306.

[12] D‰ÛªAntonio, C.M., S.J. Bainbridge, C. Kennedy, J.W. Bartolome, and S. Reynolds. (2004): Ecology and restoration of California grasslands with special emphasis on the influence of fire and grazing on native grassland species. White paper: Department of Integrative Biology, Berkeley.

[13] Dunne, J. A., and V.T. Parker. (1999): Species-mediated soil moisture availability and patchy establishment of Pseudotsuga menziesii in chaparral. Oecologia, 119, 36‰ÛÒ45.

[14] Russell, W.H. J.R. and McBride. (2003): Landscape scale vegetation-type conversion and fire hazard in the San Francisco bay area open spaces. Landscape and Urban Planning, 64, 201‰ÛÒ208.

[15] Prach, K., P. Pysek, and P. Smilauer. (1993): On the rate of succession. Oikos, 66,343‰ÛÒ346.

[16] Ingraham, N.L., and R.A. Matthews. (1995): The importance of fog-drip water to vegetation: Point Reyes peninsula, California. Journal of Hydrology, 164(1-4), 269‰ÛÒ285.

[17] Dawson, T.E. (1998): Fog in the California redwood forest: ecosystem inputs and use by plants. Oecologia, 117, 476‰ÛÒ485.

[18] Burgess, S.S.O., and T.E. Dawson. (2004): The contribution of fog to the water relations of Sequoia sempervirens (d. don): foliar uptake and prevention of dehydration. Plant, Cell and Environment, 27, 1023‰ÛÒ1034.

[19] Fischer, D.T., C.J. Still, and A.P. Williams. (2008): Significance of summer fog and overcast for drought stress and ecological functioning of coastal California endemic plant species. Journal of Biogeography, 36(4), 1‰ÛÒ17.

[20] Limm, E. B., K.A. Simonin, A.G. Bothman, and T.E. Dawson. (2009): Foliar water uptake: a common water acquisition strategy for plants of the redwood forest. Oecologia, 161, 449‰ÛÒ459.

[21] Johnstone, J.A., and T. E. Dawson. (2010): Climatic context and ecological implications of summer fog decline in the coast redwood region. Proceedings of National Academy of Sciences of the United States of America, 107(10), 4533‰ÛÒ4538.

[22] Mu, Q., M. Zhao, and S.W. Running. (2011). The implications of current and future urban global protected areas and biodiversity conservation. Biological Conservation, 141, 1695‰ÛÒ1703.

[23] Omernik, J. M. (2004). Perspectives on the nature and definition of ecological regions. Environmental Management, 34, S27-S38

[24] Hsu, W., A. Remar, E. Williams, A. McClure, S. Kannan, R. Steers, C. Schmidt, and J.W. Skiles. (2012). The changing California Coast: Relationship between climate variables and coastal vegetation succession. Proceedings of the American Society for Photogrammetry and Remote Sensing (ASPRS).

[25] Franklin, J., C. L. Coulter, and S. J. Rey. (2004): Change over 70 years in a southern California chaparral community related to fire history. Journal of Vegetation Science, 15, 701‰ÛÒ710.

[26] Kelly, A.E. and M.L. Goulden. (2008): Rapid shifts in plant distribution with recent climate change. Proceedings of National Academy of Sciences of the United States of America, 105, 11823‰ÛÒ11826.

Wei-Chen Hsu is a student intern for the NASA Ames DEVELOP Program, and has been with the program since spring of 2010. Miss Hsu has been the team lead for a research project using Hyperion hyperspectral satellite data for vegetation and biofilm classification for the south bay salt pond (SBSP) project. She received her B.S. in Environmental Science, and will also be receiving her M.S. in Landscape Architecture & Environmental Planning, from University of California, Berkeley.