Habitat fragmentation in California: Impacts of land use, climate change, and invasive species

Annotated Bibliography for GEOG 560

Compiled by Jacob Dixon

Ackerly, D. D., Loarie, S. R., Cornwell, W. K., Weiss, S. B., Hamilton, H., Branciforte, R., & Kraft, N. J. B. (2010). The geography of climate change: Implications for conservation biogeography. Diversity and Distributions, 16(3), 476–487. https://doi.org/10.1111/j.1472-4642.2010.00654.x

This study assesses changes to climates in California and Nevada under two future climate scenarios (warmer-drier and warmer-wetter) including elimination of current climates and introduction of novel climates. From baseline climate PRISM data, the authors mapped future climate distributions and tabulated mean temperatures and precipitation on a logarithmic scale across the study area. The authors also visualize areas where directional movement responses to offset climate change differ between climate variables. In the warmer-drier scenario, responses for temperature and precipitation are generally in the same direction but impacts on organisms that cannot move would be greater. The opposite is true for the warmer-wetter scenario; responses to remain at the same temperature or precipitation level are in opposite directions. The authors discuss topo climates (0.01-1 km variations) and microclimates (<10 m variations) as potential areas where heterogeneity can provide refuge to shifting climates in the short term and may show greater resilience to climate change than shown in their mesoclimate (>1 km variation) scale approach.
Relation to my research interest
As noted in the study, climate change is causing shifts in suitable habitats for plants and animals across California. The data presented here can be used to guide conservation strategies by connecting areas of current habitat with areas of likely future habitat. Additionally, novel climates represent an area of likely increased invasive presence that should be managed. High resolution elevation and climate data are important for accurate management decisions because topo climates are potential areas where heterogeneity can provide refuge to shifting climates.

Backus, G. A., Rose, M. B., Velazco, S. J. E., Franklin, J., Syphard, A. D., & Regan, H. M. (2025). Population Decline for Plants in the California Floristic Province: Does Demography or Geography Determine Climate Change Vulnerability? Diversity and Distributions, 31(8), e70067. https://doi.org/10.1111/ddi.70067

In this study, authors selected two each of oak, ceanothus, and pine species to evaluate their vulnerability to climate change based on changes in local carrying capacity, determined by amount of suitable habitat. Habitat distribution was measured from CDFW baseline data projected into the future to 2085 with two climate models and two emissions models. Demographic models for each species were implemented from previous studies based on metrics of survival, reproduction, and local seed or vegetal dispersal. Both the distribution and the demographic models were combined in a simulation which predicted relative abundance of the 6 selected species across California over time. To answer the research question, demographic models for all 6 species were applied to the distribution model for each species.
Authors concluded that all 6 species would ultimately decline in the study's timeframe due to climate and land use change; however, it was also concluded that the difference between each species’ demography was a significant factor for determining relative abundance in local populations of each species.
Relation to my research interest
This study shows a method to integrate two separate processes: the change of habitats spatially over time, and the ability for species populations to successfully track those changes. Due to differences in life history and reproductive characteristics, the future distribution of plant populations in relation to proposed species distribution models can vary. This method can be adapted to other species of interest.

Beier, P., Spencer, W., Baldwin, R. F., & McRAE, B. H. (2011). Toward Best Practices for Developing Regional Connectivity Maps. Conservation Biology, 25(5), 879–892. https://doi.org/10.1111/j.1523-1739.2011.01716.x

Presented is a 7-step process for developing regional connectivity maps based on the author’s experience creating state and regional maps for California and other areas.
The first step presented is to identify narrow goals for the map. Common goals are to identify areas where connectivity can be addressed, or where projects that would reduce connectivity should be mitigated. The authors caution that an extensive, low-resolution map should only identify where connections should be implemented, rather than be an implementable design.
Further steps focus on defining the region of interest and expanding the region to include linkages to neighboring regions (e.g. neighboring states). Tradeoffs between large and small regions of interest are identified.
The authors recommend identifying natural landscape blocks. Landscape blocks are areas of high ecological integrity (high level of natural cover, low road density, and limited edge effects); they may be protected areas and can be created using expert opinion to draw polygons by hand or use algorithms to identify areas that meet quality thresholds.
Later steps are to identify pairs of landscapes that would benefit from connectivity and to depict connectivity if resources are available. If resources are not available for detailed linkage designs, then placeholder linkages should be used, and linkage designs will be developed over time.
The final step is to provide supporting documents for future users to implement the connectivity map. This can include data for agencies to create priorities, information on focal species, and communication strategies to explain the map.
Relation to my research interest
The steps describe an easy-to-follow framework for developing connectivity maps. Together with models predicting necessary movements due to habitat change (Ackerly et al. 2010), areas that facilitate movement between current and future habitat can be identified with this framework.

Derugin, V. V., Silveira, J. G., Golet, G. H., & LeBuhn, G. (2016). Response of medium- and large-sized terrestrial fauna to corridor restoration along the middle Sacramento River. Restoration Ecology, 24(1), 128–136. https://doi.org/10.1111/rec.12286

The study monitored species richness at remnant, old restoration (prior to 2000) and young restoration (after 2003) habitat segments along the Sacramento River using trail cameras over two years. The young restoration sites had the highest species richness of predator species and equal richness for other species of interest. As a result, the authors conclude that restoration effects are observed rapidly, and heterogenous landscapes including younger and older vegetation are important for supporting predator species.
Relation to my research interest
While linear riparian restoration is important, it is equally important to restore natural stream processes that allow meandering and flooding. Seasonal flooding supports young vegetation establishment on upper flood plains. The study reaffirms importance of heterogeneity of restoration sites in addition to linear restoration.

Gillespie, L. E., Ruffley, M., & Exposito-Alonso, M. (2024). Deep learning models map rapid plant species changes from citizen science and remote sensing data. Proceedings of the National Academy of Sciences, 121(37), e2318296121. https://doi.org/10.1073/pnas.2318296121

In this study, the authors introduce a new species distribution model (SDM), Deepbiosphere. The model is localized to California and combines iNaturalist citizen science species observations (n>650,000) with National Agricultural Imagery Program (NAIP) imagery (1m resolution) in a convoluted neural network (CNN), TResNet. (Ridnek et al., 2021). Genetically and spatially related species observations are grouped to improve accuracy for under-observed species, and the model is passed through a climate data filter using WorldClim’s Bioclim variables. Deepbiosphere more accurately predicted species occurrences than TResNet or Bioclim networks on their own, and outcompeted common SDMs (Inception V3, Maxent, Random Forest).
The model was tested in several scenarios. The first scenario observed patterns of old growth and second growth forest in Redwoods National Park and compared remote sensing classification to human annotation of remotely sensed imagery. Human annotators very accurately identified old growth redwood stands, but Deepbiosphere was able to additionally identify second growth redwood forests that annotators missed. The authors also explored modeling understory species (Oxalis oregana & Rubus ursinus) associated with these two redwood age classes, finding that Oxalis preferred mature stands and Rubus preferred young stands.
Relation to my research interest
This study presents a tool for managing species distribution change over time. Citizen science data has tremendous potential for modeling species distribution, especially in a “a species-rich and data-dense state with abundant high-quality remote sensing imagery” (Gillespie et al., 2024, p. 1) as the authors describe California. The method is very accurate and has excellent resolution due to the use of NAIP imagery. It is potentially limited by the fact that the model was only trained in California, NAIP imagery is constrained to the contiguous United States, and it is constrained temporally because NAIP data is collected biannually (annually in some states). It would be interesting to see if a similar framework using lower spatial resolution but greater coverage and temporal resolution, such as Sentinel-2 data, would be plausible. (Sentinel-2 – Documentation, n.d.).

T. Ridnik et al., “Tresnet: High performance gpu-dedicated architecture” in Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision, R. Farrell et al., Eds. (Institute of Electrical and Electronics Engineers, New York, NY, 2021), pp. 1400–1409.

Sentinel-2 – Documentation. (n.d.). Retrieved March 2, 2026, from https://documentation.dataspace.copernicus.eu/Data/SentinelMissions/Sentinel2.html.

Gregory, A., Spence, E., Beier, P., & Garding, E. (2021). Toward Best Management Practices for Ecological Corridors. Land, 10(2), 140. https://doi.org/10.3390/land10020140

This study describes how to best manage an ecological corridor once regional connectivity maps and linkage designs have been implemented. The authors provide management recommendations regarding linear barriers, riparian areas, and management of different land use types, including agricultural lands. Of note are recommended buffer widths for protected streams to allow for seasonal flooding and riparian vegetation above the high-water mark. Also recommended is targeting conservation program enrollment within the corridor to mitigate agricultural land uses.
Relation to my research interest
Developing connectivity maps, as described by Beier et al. (2011), is the first step of restoring ecological connections, but long-term maintenance and iterative connectivity are required for success of linkages. Recommendations included in this study can be combined with common conservation programs and easements (Rissman and Merenlender, 2008) to prioritize habitat improvements and reductions in wildlife conflicts within the ecological corridor.

Howard, J. K., Barnett, A. R., Fesenmyer, K. A., & Anderson, M. G. (2025). From fragmentation to resilience: Connectivity and habitat diversity as drivers of fish persistence in California watersheds. PLoS One, 20(12), NaN-NaN. https://doi.org/10.1371/journal.pone.0339212

This study focuses on the impact of linear fragmentation of streams on habitat quality, habitat diversity, and fish community changes. Fragmentation was measured by dividing National Hydrography Dataset (NHD) stream reaches into “Functionally Connected Networks” (FCNs). Stream reaches were broken into FCNs at natural (waterfalls) and anthropogenic (dams) barriers. The analysis found FCNs increased from 2,711 historical networks to 4,321 current networks and diversity declined significantly in all areas of California. However, streams with limited damming in the San Joaquin Valley also showed habitat degradation, potentially related to water quality impacts and channelization of streams when they are replaced with canals.
Relation to my research interest
The method described in this study produces detailed quantification of the degree and effect of fragmentation. This method is made possible by historic fish count data as well as the relative simplicity of linear features and clear distinction between dammed and free streams compared to fragmented terrestrial areas where barriers take the form of a spectrum of permeability. Results related to habitat degradation in streams with less dams highlight the need for a historical streams dataset to quantify changes in stream shape over time and the effect channelization has on species diversity.

Huang, S., Ramirez, C., Kennedy, K., Mallory, J., Wang, J., & Chu, C. (2017). Updating land cover automatically based on change detection using satellite images: Case study of national forests in Southern California. GIScience & Remote Sensing, 54(4), 495–514. https://doi.org/10.1080/15481603.2017.1286727

This study presents a method to automatically update land cover datasets over large areas to consistently monitor plant growth patterns over a heterogeneous landscape. Datasets that are updated yearly, such as the National Land Cover Database (NLCD), focus on broad land cover categories not designed for plant cover analysis. This study uses USFS baseline Forest Inventory and Analysis (FIA) data published for 2009 to produce land cover classification for a target year 2002-2010. The algorithm developed, called AutoLCD, was tested on the three National Forests in Southern California. The chosen forests are diverse in physical and biological characteristics, and disturbance from wildfires, pests, and drought play a major role in vegetation cover making them good demonstration sites.
The authors reclassified 2009 data into three subcategories for each original class. Landsat data from target years was compared to 2009 baseline to detect if a change occurred. Pixels with similar landscape and spectral attributes were found for changed pixels. The most representative classification was chosen if it met a minimum candidate count (20 pixels) and confidence level (classification = 60% of candidate pixels). This method was also applied to any 2009 FIA pixels outside of a chosen standard deviation from mean Landsat 5 band values to improve the baseline land cover data quality. A rules-based approach to ensure that enough representative pixels were available and limit classification categories for special scenarios (e.g. following severe fire) was applied. Results of land cover classification were able to show successional stages and pre-burned vegetation cover following a wildfire in 2003.
Relation to my research interest
High quality datasets that show temporal changes are not available for every area of interest. The method presented here is a tool to work backwards or forwards from a high-quality dataset to determine and classify land cover and analyze change over time. Testing this method over a broader area would provide high quality land cover change data over time, but processing initial land cover dataset to maximize accuracy is crucial for accurate output from the algorithm.

Iverson, A. R., Humple, D. L., Cormier, R. L., & Hull, J. (2023). Land cover and NDVI are important predictors in habitat selection along migration for the Golden-crowned Sparrow, a temperate-zone migrating songbird. Movement Ecology, 11(1), 2. https://doi.org/10.1186/s40462-022-00353-2

In this study, authors used small GPS devices to study fine scale movements of short-bout migrating passerine birds. Objectives of the study were to measure chosen stopover locations using NDVI (MODIS/ 16-Day;1km resolution), North American Land Change Monitoring System (30 m resolution), and climactic characteristics. Migration length and duration were described by GPS recordings every other day during migration and every 13th day during the breeding season when the animals moved less frequently to conserve battery life on the very small GPS trackers. Sensitivity analysis was conducted to test chosen habitat to random habitat pixels near the migration route.
The study found that NDVI was positively correlated with habitat selection when shrubland and needleleaf forest land classifications were excluded but negatively correlated with either habitat type present. Shrublands were the most selected stop along the migration route, needleleaf forests were the second most selected and the most common among random points near the migration route. This indicated that needleleaf forests were common but not always favored. Both shrubland and needleleaf forest have a broad range of allowed cover in the NALCMS (20-100% woody cover). Grasslands were also commonly chosen for stopover sites, particularly in Fall.
Relation to my research interest
As this study indicates, land classification alone is not always a clear indicator of habitat selection. The negative correlation between NDVI level relative selection strength on shrubland and needleleaf land classes indicates that relatively more open shrublands and forests are selected for. Reclassifying land uses into more specific categories using NDVI or other remote sensing methods may be useful for better understanding habitat selection. Higher resolution NDVI and GPS collection may be an option for future studies and may help gain more insight into specific route selection, especially in a longer study period.

Kimball, S., Goulden, M. L., Suding, K. N., & Parker, S. (2014). Altered water and nitrogen input shifts succession in a southern California coastal sage community. Ecological Applications, 24(6), 1390–1404. https://doi.org/10.1890/13-1313.1

This paper evaluated the impacts of water additions/reductions and nitrogen inputs on the succession process following fire in coastal sage scrub (CSS). The study found that increased nitrogen inputs and reduced water both significantly increased the length of time the plant community had high grass cover following fire, before returning to coastal sage scrub. However, inputs of nitrogen to mature CSS stands have not been found to produce change in plant cover in mature CSS stands. Overall, the authors conclude that transition from shrub cover to annual grass cover is a factor of both disturbance and nitrogen input.
Relation to my research interest
Nitrogen pollution distribution is concentrated around urban areas and roadways, which are also common sources of wildfire ignition. The study indicates that coastal sage scrub near urban areas may be particularly at risk for type conversion due to these factors.

Lambrinos, J. G. (2006). Spatially variable propagule pressure and herbivory influence invasion of chaparral shrubland by an exotic grass. Oecologia, 147(2), 327–334. https://doi.org/10.1007/s00442-005-0259-1

This paper studied the impact of rabbit herbivory on resistance of coastal chaparral to invasive grasses. By planting both unprotected and protected invasive seedlings at increasing distances into a chapparal stand, the researcher found that herbivory successfully deterred invasive seedling establishment. Additionally, they studied seed dispersal and observed that chaparral cover acted as a physical barrier to wind dispersal.
Relation to my research interest
Invasive species are commonly found at habitat edges where resident herbivores are unwilling to leave protected cover to browse. Fragmentation from roads increases perimeter of habitat, even if the relative area reductions are not large. As a result, habitat edges are important to study and manage as they are areas of increased invasive species pressure.

Lawson, D. M., Regan, H. M., Zedler, P. H., & Franklin, J. (2010). Cumulative effects of land use, altered fire regime and climate change on persistence of Ceanothus verrucosus, a rare, fire-dependent plant species. Global Change Biology, 16(9), 2518–2529. https://doi.org/10.1111/j.1365-2486.2009.02143.x

The study focuses on a uniquely vulnerable plant species whose US range is limited to urban and near urban San Diego County, under threat from climate change, and unable to disperse from occupied sites to suitable, unoccupied habitats due to fragmentation and seed dispersal methods. The authors inventoried existing species populations through field data and aerial photography. Based on this baseline, they estimated future populations under warmer-drier and warmer-wetter future climate trends and found that climate impacts are more of a threat to the species than further fragmentation and shortened fire regimes.
Relation to my research interest
Ideal habitat for species will move spatially due to climate change effects, but some species, such as C. verrucosus, are not able to distribute to available habitat because of severe fragmentation. Highly fragmented species are unable to recolonize an area by seed after the resident population is eliminated due to drought or fire. Studies of the current distribution of plant species may overestimate the persistence of a population at a site if it is fragmented from nearby seed sources.

McIntyre, P. J., Thorne, J. H., Dolanc, C. R., Flint, A. L., Flint, L. E., Kelly, M., & Ackerly, D. D. (2015). Twentieth-century shifts in forest structure in California: Denser forests, smaller trees, and increased dominance of oaks. Proceedings of the National Academy of Sciences, 112(5), 1458–1463. https://doi.org/10.1073/pnas.1410186112

This study discusses the effects of climactic water deficit on forest density and overall biomass in California. The comparison of historical versus modern forests was completed using Vegetation Type Map (VTM) data from the mid-20th century with Forest Inventory Analysis (FIA) data from the early 2000s, creating a 10 arc-minute raster of tree density and basal area change, and comparing it to climactic water deficit during the same period. Presented results indicate that tree density has increased, but canopy height and biomass have decreased.
Relation to my research interest
As the study discusses, plant community changes in response to climate change are not always indicated by a change in dominant species. Structural changes, such as those observed in tree density and basal areas may not be captured by land classification, but ecosystem services are impacted, nonetheless.

Park, I. W., Hooper, J., Flegal, J. M., & Jenerette, G. D. (2018). Impacts of climate, disturbance and topography on distribution of herbaceous cover in Southern California chaparral: Insights from a remote-sensing method. Diversity and Distributions, 24(4), 497–508. https://doi.org/10.1111/ddi.12693

In this study, the authors use Landsat derived NDVI to identify the distribution of herbaceous cover in historically chapparal-dominated landscapes in the Angeles National Forest (ANF). The authors selected pixels classified as shrublands by National Land Cover Database (NLCD). They excluded pixels outside of typical chapparal elevation bands, precipitation ranges, and those that intersected fire perimeters in the previous 20 years because chapparal vegetation may be returning but is not yet distinguishable.
The study found that remotely sensed data accurately predicted herbaceous cover compared to field methods. The greatest predictor of herbaceous cover was precipitation (negative correlation), but herbaceous cover is also correlated with proximity to forest boundaries, roads, and other urban land uses. The authors only considered paved roads, citing a lack of high-resolution data of unpaved roads that dissect National Forests. Because the study eliminated areas burned within the last 20 years, impacts of increased fire intervals on herbaceous cover were not fully studied.
Relation to my research interest
This study provided an overview for predicting herbaceous cover under current conditions. The method presented here can be refined to model potential invasions into chapparal dominated landscapes in the future. As noted in the study, roads and land use boundaries are major sources of invasion and disturbance in the form of increased fire regimes. Further study could include unpaved roads which are common throughout National Forests. Additionally, it would be interesting to differentiate between native and invasive herbaceous cover and test this model using higher resolution NAIP data.

Rissman, A., & Merenlender, A. (2008). Ecology and Society: The Conservation Contributions of Conservation Easements: Analysis of the San Francisco Bay Area Protected Lands Spatial Database. Ecology and Society, 13(1). https://doi.org/10.5751/ES-02329-130140

This study compares the spatial distribution and characteristics of conservation easements and “fee-simple properties”. Conservation easements allow landowners to retain some control over land while prohibiting development and may dictate management goals of the property depending on the easement. Fee-simple properties are protected areas owned by state, federal, or non-government organizations. The study compared land cover from the Fire and Resource Assessment Program (FRAP, 100 m), 30 m DEM, Census, and categorization data from local land use plans. The analysis found that both property types had steeper slopes than non-protected areas, but conservation easements were on flatter slopes than fee-simple areas. Overall, connectivity – both within and between land use types – was high, reflecting selection criteria for connected lands and the concentration of land available for protection.
Relation to my research interest
Conservation easements provide a crucial steppingstone between more traditional publicly owned protected areas and contain vegetation types not frequently found on fee simple properties. As they are more likely to be developable land, the land that conservation easements cover is more at risk of land use conversion. Managing conservation easements to properly benefit connectivity while maintaining agricultural productivity is a challenge. However, compared protecting a property by purchasing, the reduced price and management required by agencies make them a key piece in connecting landscapes. The connectiveness assessment in this study could be used to identify areas of priority for future conservation easement or conservation planning. As in other studies, higher resolution data for determining land cover would be useful for a more accurate assessment of vegetation represented.

Stemle, L. R., Laabs, D., & Searcy, C. A. (2025). Land use change, fragmentation, and sea level rise create escalating viability concerns for an imperiled salamander. Ecosphere, 16(5), e70284. https://doi.org/10.1002/ecs2.70284

This study studied the migration of Santa Cruz Long Toed Salamander (SCLTS) from breeding ponds to upland, the barriers to successful habitat connection, and future climate impacts on extant ponds. The purpose of the study was to find breeding ponds with adequate and accessible habitat which could be used for release of captive-bred salamanders and identify imperiled ponds to be the target of restoration. The authors defined ponds using satellite imagery, created buffers based off migration distance to contain 50% and 95% of the SCLTS, and created barriers at any road with >1000 average daily vehicle traffic. Proportion of habitat in the resulting buffers was calculated using National Land Cover Database (NLCD) data. The authors conclude that only 24% of breeding ponds are likely to support SCLTS long-term.
Relation to my research interest
The study found that many ponds had suitable habitat within the buffer area, but that it was not accessible due to road barriers. Barriers to upland habitat were a greater limitation for northern ponds than southern ponds which are mostly limited by land use. The most beneficial methods to defragment the landscape such as upland restoration or road mitigation vary spatially. Forest land cover was classified as habitat, but the authors note that NLCD does not differentiate between native-dominated and invasive forests, and invasive-dominated forests in the study do not provide good habitat for SCLTS. Study of accessible habitat would benefit from remote sensed vegetation data or a different land classification dataset to differentiate between forest types. The authors note that habitat fragmentation has a slow, serious effect on species populations, but there is an opportunity to defragment the landscape before the new equilibrium population is set.

Syphard, A. D., Regan, H. M., Franklin, J., Swab, R. M., & Bonebrake, T. C. (2013). Does functional type vulnerability to multiple threats depend on spatial context in Mediterranean-climate regions? Diversity and Distributions, 19(10), 1263–1274. https://doi.org/10.1111/ddi.12076

This study discusses whether grouping species by functional traits such as life history and response to disturbance can help assess global change impacts. While grouping species by functional traits may be useful for studying climate change related impacts, spatial variation of land use changes may have a more immediate effect on species distribution. The authors combine the effects of altered fire regime, climate change, and land use changes on two functionally similar but geographically distinct Ceanothus species. C. verrucosus gains territory under the warmer-wetter climate model, when land use is not considered. Both species show reduction in projected abundance if fire return interval is shorter than 10-20 years. The study shows that while functional groups are useful when projecting climate change impacts on a species, geography and land use conflicts must be considered as well.
Relation to my research interest
There are many complex processes interacting in the model presented including urban growth and makeup related to fire frequency, multiple climate models, and unclear future development trends. Additionally, threats themselves are spatially distributed and will not impact species uniformly. As the authors note, distribution of threatened species needs to be analyzed with references to changes in land use, the degree of climate shifts in an area, and the interaction of multiple threats.