Obligate resprouting, obligate seeding, and facultative seeding shrub species in California's Mediterranean-type climate region

Mediterranean-climate region (MCR) shrublands have evolved a set of regeneration strategies in response to periodic, high-intensity wildfires: obligate seeding (OS), obligate resprouting (OR), and facultative seeding (FS) species. Spatial variation is seen in different regeneration strategies. In California, previous studies have found a higher abundance of OR species in mesic environments and OS species in xeric environments (Meentenmeyer et al. 2001). To date, however, data on their spatial distribution at a regional scale in California is limited and presents a significant information gap for resource managers of shrub-dominated landscapes. We developed a multinomial model using temporally dynamic and static variables to predict the distribution of the three shrub post-fire regeneration strategies, plus trees and herbs, in southern California. Cross-validation showed 50% of the predicted values for each of the five plant groups were within 8–24 percent of the actual value (Underwood et al. 2023). Spatial data for OS, OR, and FS provide an important contribution to resource management to help quantify carbon storage of shrublands and prioritize areas for post-fire restoration.The file archive here is named SoCal_postfire_regen_type_v1.zip and is composed of five files that cover the southern California ecoregion (Figure 1): SoCal_ObResprouters_pp_v1.tif = proportion of obligate resprouter biomass per pixel SoCal_ObSeeders_pp_v1.tif = proportion of obligate seeder biomass per pixel SoCal_FacSeeders_pp_v1.tif = proportion of facultative seeder biomass per pixel SoCal_Tree_pp_v1.tif = proportion of tree biomass per pixel SoCal_Herb_pp_v1.tif = proportion of herb biomass per pixel Apply a scale factor of .01 to the raster values to convert to proportions. The dimensions of each 8-bit geotiff raster file are 20632 rows by 15602 columns and the bounding box coordinates are 36.79, -121.96 (upper left) and 32.47, -115.23 (lower right), in decimal degrees. Pixel size is 30 meters and the files are projected in the California Albers Equal Area projection (EPSG 3310). These rasters are encoded as bytes with a NoData value of 0. Areas outside of the study area extent and those pixels where the biomass proportion is 0 are represented as NoData. Intended users of this dataset include resource managers and researchers who are assessing carbon storage and shrublands and prioritizing post-fire shrubland restoration. This dataset is made available under a CC0 license.Funding provided by: California Department of Forestry and Fire ProtectionCrossref Funder Registry ID: http://dx.doi.org/10.13039/100020524Award Number: Funding provided by: USDA Forest Service Western Wildlands Environmental Threat Assessment Center*Crossref Funder Registry ID: Award Number: Funding provided by: USDA Forest Service Pacific Southwest Program*Crossref Funder Registry ID: Award Number:Our study area consists of shrublands within a 31,069 km2 (7,677,317 acres) footprint that encompasses all 36 Level IV ecoregions (Omernik and Griffith, 2014) that overlap with the Angeles, Cleveland, Los Padres, and San Bernardino National Forests in southern California, USA (Figure 1). Vegetation types in the study region are dominated by shrubland (62%), grassland (16%), broadleaf woodland (8%), and conifer and mixed conifer-broadleaf forests (8%). The predominant shrublands communities (following Wildlife Habitat Relations classification, Barbour et al., 2007; https://wildlife.ca.gov/Data/CWHR/Wildlife-Habitats [CWHR]) are: mixed chaparral (29%; dominated by scrub oak [Quercus berberidifolia], various species of Ceanothus and manzanita [Arctostaphylos], and other mostly resprouting shrub species); sage scrub (12%; dominated by California sagebrush [Artemisia californica], purple sage [Salvia leucophylla], black sage [Salvia mellifera], and California buckwheat [Eriogonum fasiculatum]); and chamise/redshank chaparral (6%; dominated by chamise [Adenostoma fasciculatum] and/or redshank [Adenostoma sparsifolium]). We analyzed 222 plots from the USDA Forest Service Forest Inventory and Analysis (FIA) program (Burkman, 2005) for which we estimated aboveground biomass using shrub species-specific allometric equations (McGinnis et al., 2010; Wakimoto, 1978) or a generalized shrub-herb biomass equation (Lutes et al., 2006, see Schrader-Patton and Underwood, 2021 for details). We assigned each species in the FIA plots to one of three lifeform categories: shrub, tree or herb (forbs and grasses). Shrub species were further categorized into one of three post-fire regeneration strategies: obligate seeder (OS), facultative seeder (FS), or obligate resprouter (OR) using descriptions of regeneration strategy and life history reported in primary literature and public databases (Gordon and White, 1994; Borchert et al., 2004; CNPS, 2021; FEIS, 2021). For each plot, we calculated the proportion of aboveground live biomass for each of these five plant groups: OS, FS, OR, tree, and herb, by dividing the estimated biomass of each by the summed total biomass across all groups. To predict the distribution of OS, FS, OR, trees and herbs, we used a multinomial regression model (mnet package and function multinom in R software, R Core Team, 2016; Ripley and Venables, 2021) using: average annual solar radiation, actual evapotranspiration (AET), climate water deficit (CWD), average annual precipitation, the normalized difference vegetation index (NDVI, using the maximum composite value from July to August each year), modeled aboveground biomass (a proxy for productivity), eastness (a measure of continentality and dryness), slope, flow accumulation, soil bulk density, soil clay content, and soil percent carbon (see Table S1). Finally, we included vegetation type from the CWHR classification system from the FVEG vegetation data (FRAP, 2015) and time since last fire from the Fire Return Interval Departure geodatabase (USDA, 2015). We omitted any variables that were strongly correlated (r > 0.55). To select the best model, we started with a full model and removed predictors sequentially using Akaike Information Criteria (AIC) to evaluate model fit. Predictor variables remained in the final model if they improved model fit by a minimum ΔAIC of –2 (Anderson and Burnham, 2004). For the final model, p-values were generated with a Wald-z test. Obligate seeder shrubs were selected as the baseline variable for the multinomial model (as we expected OS to differ most from FS and OR) against which the other four plant groups were calculated. To evaluate model performance, we performed leave-k-out cross-validation with k = 8 (Hastie et al., 2009) and examined the distribution of cross-validation errors to evaluate predictive accuracy: mean, standard deviation, kurtosis, skew, and interquartile range. We then used the raster surfaces corresponding to each predictor as model inputs into the 'predict' function in R software. We created a raster spatial layer (30 m resolution) with the proportion of biomass for each of the five plant groups by applying the model predictions of the proportion of each plant group to each pixel in our study area (see Underwood et al. 2023 for details). Figure 1. The boundary of the study area in southern California within which we identify the distribution of obligate resprouter, obligate seeder, and facultative seeder species. CAVEATS In using these spatial data, users should note that the USFS FIA program that provided the input plots for the multinomial model is designed to measure forest conditions across the US, so FIA sampling is biased to upland interior, moister sites which contain trees. Consequently, OS species in coastal and desert scrub communities may not be well represented. In addition, studies (e.g., Keeley 2023) have shown some shrub species, such as Ceanothus leucodermis, vary temporarily and spatially in their post-fire regeneration strategy, changing from OR to OS with longer fire-fire intervals. REFERENCES Anderson, D., and Burnham, K. (2004). Model Selection and Multi-Model Inference (2nd edition). New York, NY: Springer-Verlag. Barbour, M. G., Keeler-Wolf, T., and Schoenherr, A. A. (2007). Terrestrial Vegetation of California. Berkeley, CA: University of California Press. Borchert, M., Lopez, A., Bauer, C., and Knowd, T. (2004). Field guide to coastal sage scrub and chaparral alliances of Los Padres National Forest. USDA Forest Service Region 5, Ecological Field Guide. Burkman, B. (2005). Forest inventory and analysis sampling and plot design; FIA fact sheet Series. USDA Forest Service Forest Inventory and Analysis National Program. Washington DC, USA: USDA Forest Servicehttps://www.fia.fs.usda.gov/library/fact-sheets/data-collections/Sampling%20and%20Plot%20Design.pdf CNPS (California Native Plant Society). (2021). A Manual of California Vegetation, online edition. Sacramento, CA: California Native Plant Society http://www.cnps.org/cnps/vegetation/ FEIS (Fire Effects Information System). https://www.feis-crs.org/feis/ FRAP (Fire and Resource Assessment Program). Data from: Fveg15_1 vegetation data. California Department of Forestry and Fire Protection's CALFIRE Fire and Resource Assessment Program (FRAP). (2015) http://frap.fire.ca.gov/data/frapgisdata-sw-fveg_download Gordon, H., and White, T. C. (1994). Ecological guide to southern California chaparral plant series. Transverse and Peninsular Ranges: Angeles, Cleveland, and San Bernardino National Forests. Report R5-ECOL-TF-005. Albany WA: USDA Forest Service Pacific Southwest Region. Hastie, T., Tibshirani, R., and Friedman, J. (2009). The elements of statistical learning: data mining, inference, and prediction (2nd ed.). Springer. Lutes, D., Keane, R., Caratti, J., Key, C., Benson, N., Sutherland, S., and Gangi, L. (2006). FIREMON: Fire Effects Monitoring and Inventory System. Fort Collins, CO: Rocky Mountain Research Station. McGinnis, T. W., Shook, C. D., and Keeley, J. E. (2010). Estimating aboveground biomass for broadleaf woody plants and young conifers in Sierra Nevada, California, forests. West J App For 25, 203–209. Meentemeyer, R. K., Moody, A. and Franklin, J. (2001). Landscape-scale patterns of shrub-species abundance in California chaparral. Plant Ecol 156, 19–41. Omernik, J. M., and Griffith, G. E. (2014). Ecoregions of the conterminous United States: evolution of a hierarchical spatial framework. Environ Manage 54, 1249–1266. https://doi.org/10.1007/s00267-014-0364-1. https://www.epa.gov/eco-research/ecoregions R Core Team. (2016). R: A Language and Environment for Statistical Computing. Vienna, Austria: R Foundation for Statistical Computing. Ripley. B., and Venables, W. (2021). nnet: feed-forward neural networks and multinomial log-linear models. R Package. Schrader-Patton, C. C., and Underwood, E. C. (2021). New biomass estimates for chaparral-dominated Southern California Landscapes. Remote Sens 13(8), 1581. https://doi.org/10.3390/rs13081581. Schrader-Patton, C. C., Underwood E. C., and Sorenson, Q. M. (2023). Annual biomass spatial data for southern California (2001–2021): Above- and belowground, standing dead, and litter. Ecology e4031. Underwood, E. C., Sorenson, Q. M., Schrader-Patton, C. C., Molinari N. A., and Safford, H. D. (2023). Assessing spatial and temporal variation in obligate resprouting, obligate seeding, and facultative seeding shrub species in California's Mediterranean-type climate region. Front Ecol Evol https://doi.org/10.3389/fevo.2023.1158265 USDA [US Department of Agriculture]. Data from: Fire-Return Interval Departure (FRID) Geodatabase. (2015) http://www.fs.usda.gov/detail/r5/landmanagement/gis/?cid=STELPRDB5327836 Wakimoto, R. H. (1978). Responses of Southern California brushland vegetation to fuel manipulation [dissertation]. [Berkeley (CA)]: University of California