*Sensitivity—Landscape Condition*

Since human land uses, such as built infrastructure for transportation, urban development, industry, agriculture and other vegetation alterations, are depicted in maps that are periodically updated, they can be used in spatial models to make inferences about the status and trends in human-induced stress and ecological condition of ecosystems and landscapes at regional to global scales [54–57]. The spatial model of landscape condition used here [58] built on a growing body of published methods and software tools for ecological effects assessment and spatial modeling; all of which aim to characterize relative ecological condition of landscapes [59–61]. The model uses regionally available spatial data to transparently express user knowledge regarding the relative effects of land uses on natural ecosystems.

Values close to 1.0 indicate almost no measurable ecological impact from the land use at a given pixel. As described in [57], model parameters were calibrated, and subsequently validated using tens of thousands of field observations indicating relative ecological condition. The result is a map surface that provides relative index scores per pixel between 0.0 and 1.0. Calibration of this model against over 50,000 field occurrences ranked as A = excellent, B = good, C = fair, and D = poor condition was used to identify thresholds in the 0.0–1.0 scale for applications. In this instance, we used one standard deviation above the mean of the index value for the D occurrences to determine the C. vs. D threshold. The overall threshold value breaks are as follows; A-Rank ≥ 0.36, B-Rank ≥ 0.30, C-Rank ≥ 0.25, D-Rank < 0.25. Per pixel scores were summarized to average values per vegetation type per 100 km<sup>2</sup> hexagon for display.

#### *Sensitivity—Invasive Plant Species*

Among desert shrubland and steppe, the effects of invasive species on ecosystem integrity is well known and there is considerable concern for their interactions with climate change [61]. Spatial models depicting likely presence and abundance of invasive annual grasses provide an important indication of vegetation condition, and therefore, relative sensitivity under the HCCVI framework.

See [23] and [62] for further explanation of spatial models used here. Using the master database of over 20,000 invasive plant locality records with satellite imagery and a suite of environmental variables, inductive modeling was completed using Random Forests [41]. The resultant independently evaluated map surfaces represent invasive annual grass presence in five categories of expected absolute cover (<5%, 5–15%, 16–25%, 26–45%, and >45%). The five models were then combined onto one surface with higher predicted invasive cover classes taking precedence over lower cover classes on a per pixel basis. These absolute cover values were translated to index scores to reflect "1.0 = most favorable" to "0.0 = least favorable" index values as follows: <5% = 1.0, 5–15% = 0.80, 16–25% = 0.6, 26–45% = 0.4, >45% = 0.2. These per pixel scores were then summarized to average values per vegetation type per 100 km<sup>2</sup> hexagon. This measure applied to desert shrubland and grassland vegetation types where invasive annual grasses have substantial impact.

#### *Sensitivity—Fire Regime Departure*

Using estimates of fire frequency and successional rates, fire regime models predict the relative proportion of natural successional stages one might expect to encounter for a community type across a given landscape. They are therefore useful for indicating ecosystem degradation due to wildfire suppression or other human-caused alteration [63]. The US Interagency LANDFIRE program provides both quantitative reference models of vegetation states (i.e., successional stages) and transitions, as well as spatial models of wildfire regime departure (measured in percent of departure) that compare observed vs. predicted aerial extent of each successional stage [33]. For each vegetation type treated in this project, these percent departure values (in 10% increments) were translated to index scores to reflect "1.0 = most favorable" to "0.0 = least favorable" index values as follows: FRCC 1 = 1.0, FRCC 2 = 0.5, and FRCC 3 = 0.15. These per pixel scores were then summarized to average values per vegetation type per 100 km<sup>2</sup> hexagon.

#### *Sensitivity—Forest Insect and Disease Risk*

Forest insect and disease impacts on Western US forests and woodlands are becoming pronounced, especially with increasing frequency of relatively mild winters [64]. With increasing rates of overwintering survival of both native and introduced insects, as well as compounded effects of drought [65] there is increasing potential for substantial disruption in forest stand structure, composition, and interacting effects with other natural disturbance processes [66]. The National Insect and Disease Risk Map defines forest areas where, "*the expectation that, without remediation, at least 25% of standing live basal area greater than one inches in diameter will die over a 15-year timeframe (2013–2027) due to insects and diseases"* [67]. The resultant 240 m pixel resolution map represents insect and disease risk along a 0.0–1.0 ramp depicting low to high severity of predicted biomass loss (e.g., 0.05 = 5%, 0.25 = 25%, 0.35 = 35%, etc.). These index values were flipped in order to reflect our "1.0 = most favorable" to "0.0 = least favorable" index values. These per pixel scores were then summarized to average values per vegetation type per 100 km<sup>2</sup> hexagon. This index was applied only to forest and woodland types where forest insects and diseases have substantial impact.

#### *Resilience—Ecosystem Adaptive Capacity*

Below are described several measures of adaptive capacity, including diversity within functional species groups, climate change vulnerability of "keystone species," and topo-climate variability.

#### *Diversity within Characteristic Functional Species Groups*

Natural communities may include several functional groups, or groups of organisms that pollinate, graze, disperse seeds, fix nitrogen, decompose organic matter, depredate smaller organisms, or perform other functions [68,69]. Functional species groups (FSGs) form a link between key ecosystem processes and structures and ecological resilience. Experimental evidence gathered over recent decades supports the theoretical prediction that communities with functional groups made up of increasingly diverse

members tend to be more resilient to perturbations [27]. Therefore, the more diverse the FSG (as measured by taxonomic richness), the greater the likelihood that at least one taxon will have characteristics that allow it to continue to perform its function in the community as climate changes.

Approaches to identifying FSGs for natural communities center on analysis of specific traits in response to environmental constraints [70]. In this e ffort, environmental settings, dynamics processes, species responses to those settings and processes, and key biotic interactions required for maintenance of each vegetation type, were evaluated to identify the most critical ecological processes and their related FSG. FSGs applicable to the vegetation types in this project included nitrogen fixers, biotic pollinators, biotic seed dispersers, biological soil crusts, perennial cool season grasses, perennial warm season graminoids, halophytes, xerophytes, and pyrophytes. Within each identified group, a listing of characteristic species is included based on existing documentation for the type. Lists of functional species groups used are described in the supporting information (Supplementary Materials).

In each instance, available literature was reviewed to document each group, and for each vegetation type, score them along a 0.0 to 1.0 scale. Due to limited knowledge of variation in FSG composition across the Western US, the same score was applied consistently across the entire distribution of each vegetation type, and we scored FSG diversity in three categories of Low, Medium, or High. While ranges varied by FSG, generally those groups with 1–5 species were scored as low (=0.15, 6–15 species as medium (0.5), and >15 species as high (1.0). Where several FSGs were identified for a given vegetation type, the lowest scoring FSG was applied to the type overall, as it would likely have a strong controlling e ffect on resilience.

#### *CC Vulnerability of Keystone Species*

To assess community resilience, it is important to consider the relative climate change vulnerability of species that play particularly important functional roles. We use the term "keystone species" here to refer to any species that, due to their key functional role in the community, if extirpated or reduced in abundance, could cause disproportionate e ffects on the populations of other species that characterize the community. Determining the species that can be considered keystone requires an understanding of the natural history of many species in the community being assessed. Although there are quantitative means of identifying keystone species via food web analysis [71], these methods can be time and data intensive. However, identification of potential keystone species may follow directly from the above process "diversity within functional species groups". That is, if an important ecosystem function is represented by just one species, that species is likely providing some "keystone" function for the purposed of this analysis. We reviewed all lists of species across our FSGs and identified a set of keystone species for each vegetation type.

We assessed keystone species vulnerability using the NatureServe Climate Change Vulnerability Index (CCVI), a trait-based tool that allows relatively rapid assessment of suites of species and is applicable to all terrestrial and aquatic plant and animal species [24]. The CCVI places species on a categorical scale from extremely vulnerable to those likely to benefit from climate change. For this e ffort, the CCVI was applied to the distribution of each species within each of the primary ecoregions that make up the distribution of the vegetation type. The CCVI categories were translated to a numerical scale (0.0–1.0 scale) for combination with other adaptive capacity measures.
