Establishing Reference Models for Ecological Restoration—Case Study from Colorado National Monument, USA

Abstract

Restoration practitioners specify goals that describe how the focal ecosystem will look or function upon reaching recovery goals. Goals may be influenced by the level of degradation, surrounding landscape conditions, societal choice, and a changing climate regime. The Society for Ecological Restoration’s International Principles and Standards for the Practice of Ecological Restoration recommend that goals should be informed by reference models of site conditions, which include the biotic composition, the environmental setting, and dynamic processes—had anthropogenic degradation not occurred—while accounting for anticipated changes. The SER principles address many aspects of ecological restoration, and practical steps include conceptualizing the structure and function of the natural system, measuring ecological integrity, and assessing potential climate change effects and adaptations. Models optimally reflect a variety of information sources and are based, where possible, on multiple reference sites of similar native ecological conditions. Using a project site from the Colorado National Monument in the USA, we illustrate a stepwise process to address these principles and standards by compiling and synthesizing map, text, and tabular information from reference materials and sites. By addressing these principles and systematically utilizing existing frameworks and locally available data, practitioners can streamline the establishment of reference models for ecological restoration.

1. Introduction

The Kunming–Montreal Global Biodiversity Framework (KM-GBF) was adopted by the parties of the Convention on Biological Diversity (CBD) in December 2022. Among the framework’s key elements are 23 targets for 2030. Target 2 of the KM-GBF states the following: “Ensure that by 2030 at least 30 percent of areas of degraded terrestrial, inland water, and coastal and marine ecosystems are under effective restoration, in order to enhance biodiversity and ecosystem functions and services, ecological integrity and connectivity” [1]. Effective restoration can be defined as “standards-based restoration underpinned by agreed principles that results in appropriately balanced sustainable net gain that benefits and enhances biodiversity, ecosystem integrity and human well-being.”

In 2021, the United Nations adopted Principles for Ecosystem Restoration to Guide the United Nations Decade 2021–2030 [2]. These ten principles include the following: (#4) achieve the highest level of recovery of biodiversity and ecosystem integrity; (#5) address the indirect and direct causes of degradation; (#6) integrate knowledge from diverse sources; (#7) establish well-defined short-, medium-, and long-term objectives and goals; (#8) tailor plans to local conditions; (#9) enable ongoing monitoring and adaptive management; and (#10) support policies that foster replication and scaling up.

The Society for Ecological Restoration, in their International Principles and Standards for the Practice of Ecological Restoration (SER Standards) [3], further specifies that restoration goals can be developed from reference models of site conditions. That is, “the expected condition that the restoration site would have been in had it not been degraded, while accounting for anticipated change” [3]. Reference models optimally reflect a variety of information sources, include multiple reference sites of similar ecological conditions, and acknowledge the dynamic character of ecosystems.

However, ecological restoration is quite challenging. One recent meta-analysis indicated that many restoration projects fail to recover to reference levels of biodiversity [4]. Challenges to constructing reference models include limited knowledge of undegraded conditions and reference data sources that may be difficult to locate or access. Also, planners may lack appropriate frameworks to organize information and construct reference models. Today, relatively few projects have begun to rigorously factor in potential effects of climate change.

Reference data sources, especially for smaller restoration sites and certain agencies, are often identified from surrounding, less disturbed areas, driven by regulatory requirements for plant cover and soil loss. A lack of consistency in approaches to, or standards for, reference conditions and project restoration goal development has contributed to challenges with achieving ecosystem recovery.

Additionally, advances in the science supporting both biodiversity conservation and ecological restoration have led to a greater appreciation that ecosystems are dynamic across spatial and temporal dimensions, with or without climate change effects. Implications of this include the following:

• The potential for alternative condition states and recovery trajectories with compositional and structural characteristics [5].
• The likelihood that reference conditions will change during the process of recovery in the degraded system [6].
• Reference sites or the modeled condition should be placed in the context of the broader landscape conditions [7].
• Project goals should be based on a reference model using multiple data sources, allowing for a range of values [8] that support the variability of the ecosystem undergoing restoration across its range [9].

Recent work using different approaches to source reference models includes [10,11,12], with Oliver et al. [10] emphasizing the need to describe the reference in the context of an “acceptable” range of variation, thereby avoiding selection of excessively narrow restoration goals. They compared thematic approaches through classifications or parameters of classifications and state-and-transition models. They also evaluated selected indicators from restoration sites to estimate the number of existing sites that would need to be included to quantify a reasonable picture of the range of variation [10].

Constantz et al. [11] selected sites identified from earlier surveys that were “the best-preserved habitat in the region…” despite being impacted by several disturbances, including flow regulation, levees, and bank revetment, historic(al) logging, and the presence of numerous invasive exotic species.”

Sluis et al. [12] used species lists generated from a range of sites ranked by a state natural heritage program in the USA to help describe conditions as “High Quality.”

In this context, restoration practitioners need clearer structures to integrate stated principles and standards and to design effective projects, including measurable goals or targets. These should be based on both practical and scientifically sound approaches for managing native ecosystems, like (1) documenting knowledge of ecosystem attributes, (2) identifying and measuring indicators of ecological integrity, and (3) accounting for likely effects of climate-induced ecosystem stress.

Our objective is to demonstrate how to implement these global principles and standards using source information from ecological classifications and ecosystem inventories, along with frameworks to organize data for assessing ecological integrity and climate change vulnerability, when evaluating site conditions and developing a reference model of a given restoration site. We recommend a stepwise process to compile and synthesize common map, text, and tabular information from reference materials and sites to develop models with clear attributes to set and communicate restoration goals. We used the SER Five-star System and Ecological Recovery Wheel to concisely communicate both current and potential restored conditions from the reference model. By systematically utilizing commonly available data in the USA, we streamline the process of establishing these models.

This assessment was conducted as a “proof-of-concept” to support the development of a reference model, which begins with the analysis of existing, readily available information. It organizes available information to identify indicators and approximate thresholds that will form the focus for describing both current conditions and potentially achievable restoration goals. Local managers will collect additional site-specific information that was not available, or not at the appropriate scale, and fine-tune the model suggested here to refine a restoration goal (sensu the SER Principles [3]) before designing treatments.

While our case study addresses specific dryland ecosystems in the USA, analogous information elsewhere, including for wetland and aquatic ecosystem conditions, should be sought in those circumstances. We anticipate that restoration practitioners worldwide could replicate much of this process to develop models applicable to any given restoration project area.

2. Materials and Methods

2.1. Colorado National Monument—Case Study

Our case study is a degraded area designated for restoration (restoration site) within the Colorado National Monument (Monument), a unit of the United States National Park Service (NPS). The Monument is a protected area of nearly 8300 hectares along the Colorado River in western Colorado, USA. See Supplementary Materials for maps and photos of the Monument and restoration site. Ecologically, the Monument reflects the transition from montane forests down to cold desert tablelands of the Colorado Plateau where it receives 28 cm of precipitation per year. The Monument is dominated by pinyon-juniper woodlands (https://explorer.natureserve.org/Taxon/ELEMENT_GLOBAL.2.722905/Colorado_Plateau_Pinyon-Juniper_Woodland, accessed on 15 January 2025) on sandstone mesas and tablelands but also includes over 570 hectares of cold semi-arid desert shrublands and grasslands. These shrublands are often dominated by various species of sagebrush (Artemisia spp.) on piedmonts and flats that form the eastern wildland-urban interface (WUI) with the adjacent city of Grand Junction [13]. The restoration site is about 1450 m above sea level and encompasses roughly 350 hectares of the WUI, occurring as a 14 km long strip (~250 m wide) from just north of the Monument’s entrance and extending south. The Monument’s vegetation inventory indicates that the current restoration site is predominantly composed of three vegetation types dominated by four-winged saltbush (Atriplex canescens), black sagebrush (Artemisia nova), and invasive cheatgrass (Bromus tectorum), respectively [14].

This area had been heavily grazed by captive bison (Bison bison) that were brought into the Monument in the 1920s to support tourism [14]. The bison were removed in the 1980s after years of negatively impacting the area. The impacts from bison, plus other surface disturbances from a Civilian Conservation Corp camp established in the project area from the 1930s–1940s, severely degraded the area. The shrub and grass communities of the restoration site now reflect the severely altered composition and patchy dominance of invasive plant species in a condition that is worsening over time. The replacement of the native plant species by invasives has decreased native biodiversity [14].

It is also clear that climate change is occurring throughout the region and could have a substantial impact on restoration and broader management decisions at the Monument [15,16]. These decisions will clearly need to consider new approaches to account for these changing ecosystem stressors [17].

The NPS is mandated to conserve the natural and cultural resources under its jurisdiction. Intervention is directed “to restore natural ecosystem functioning that has been disrupted by past or ongoing human activities [18].” Temporal dimensions frame NPS restoration by considering past, current, and likely future conditions as a guide for developing restoration goals. Here we clarify sources to describe reference conditions with current and historical source materials but will also consider likely effects of intensifying nearby land use and climate stress emerging over upcoming decades.

2.2. Workflow, Analytical Frameworks, and Data Sources

We established a stepwise workflow (Figure 1) for integrating frameworks (ecosystem classification, ecological integrity assessments, and climate change vulnerability and adaptation) and data to construct reference models for site restoration. Steps 1–6 encompass key activities starting with site definition and ending with establishing climate-informed restoration objectives that can be measured and monitored with implementation.

Through this workflow, we integrate existing analytical frameworks and applicable data to identify potential reference models generated in part from existing reference sites. Monument managers identified priority sites within the larger degraded area (Step 1), and the SER Standards recognize the need for multiple reference models for use in distinct zones in similarly complex sites.

Reference models are described by key ecosystem attributes including the absence of threats, physical conditions, species composition, structural diversity, ecosystem function, and external exchanges (e.g., species dispersal, landscape dynamics, etc.). In turn, they can be translated to a five-star system developed to document the restoration project across a trajectory of recovery and to help managers communicate the progress of their work.

While three restoration zones were delineated within the Monument’s restoration site, here we focus on explaining reference model development for just one—the largest—of those zones. Lastly, we integrate our information sources to propose preliminary ecological restoration goals for the restoration zone. NOTE: See Supplementary Materials for substantial additional background and details of the methods summarized below.

2.3. Ecosystem Classifications (Step 2)

As noted above, the restoration site at the Monument includes three zones. We delineated these using map and field observations of geophysical substrates, related dynamic ecological processes, and site history. Criteria were needed to justify delineation, so multiple sources of information were sought as recommended by the SER Standards.

Ecosystem classifications are a practical starting point for documenting reference models suited to the distinct restoration zones within the project site. Classifications take several common forms around the world. Scientific traditions of phytosociology (e.g., detecting patterns in plant species assemblage in vegetation) [19] as well as those emphasizing geophysical drivers of ecosystems [20], both largely initiated in Europe in the early 20th century, exist today in most countries around the world. While no globally accepted standard for ecosystem classification exists, recent efforts, such as those in support of the global IUCN Red List of Ecosystems [21], are emerging with increasing utility and usage.

In practice, ecologists commonly reconcile local native community descriptions into common classifications developed for their country or regional jurisdiction.

In the USA, for terrestrial environments, classifications commonly focus on rooted plant assemblages or vegetation [22], while others emphasize geophysical components, like Natural Resource Conservation Service (NRCS) ecological sites, those affecting wetland hydrology or soil productivity [23], or some combination of the two [24]. Terrestrial ecological classifications are developed in part from the analysis of reference vegetation sample plots—selected from across the known natural range—to help define distinct but recurring types [22]. In aquatic environments, ecological classifications are much more limited, but they tend to emphasize geophysical attributes (hydrologic regime and water chemistry) in type definitions [25].

Three of these ecological classification products—from the U.S. National Vegetation Classification (USNVC) (https://usnvc.org/, accessed on 31 January 2025) [22], the NatureServe terrestrial ecological systems (https://explorer.natureserve.org/Search, accessed on 31 January 2025) [24] as used in LANDFIRE biophysical settings (BpSs), and Natural Resource Conservation Service (NRCS) Ecological site descriptions [23]—are complementary and widely available for application across the USA. We brought them together to form a foundation for the reference model for our restoration zones at the Monument.

2.4. Reference Data Describing the Types to Be Restored (Step 2)

Many have pointed to the importance of reference sites, associated data, and their central role in supporting ecological restoration [26]. Vegetation inventories document existing vegetation occurring within a given area of interest, typically through gathering plant species lists and documenting the relative density or areal coverage of each within established sample areas [27]. Globally, vegetation sample data are increasingly available to researchers [28].

We used both maps and vegetation sample data (i.e., plots measuring the percent cover of plant species and then labeled by classification type) from the Monument [29] to evaluate and describe vegetation types likely to occur in the restoration site, and the floristic composition of targeted vegetation as they occur locally was determined.

2.5. Modeling Landscape and Ecosystem Dynamics (Step 3)

Conceptual “State-and-Transition” models (STMs) describe key ecosystem components, their driving ecological processes, and their natural variation over time and space to typify a reference site [30]. Such defining characteristics may be viewed as the “Key Ecological Attributes” [sensu 8] of that resource. While not universally available, much guidance exists for the development of these models anywhere around the world [31].

We used existing STMs originating with LANDFIRE BpSs (https://landfire.gov/vegetation/bps-models, accessed on 12 December 2024) [32] and with NRCS ESD (https://www.nrcs.usda.gov/getting-assistance/technical-assistance/ecological-sciences/ecological-site-descriptions, 12 December 2024) classifications [23] as two primary sources. While ESDs are not available comprehensively in the USA, they do encompass a large proportion of the country, especially in western states. LANDFIRE BpS models exist for all natural types in the USA. They initially highlight geophysical constraints and dynamics one should anticipate in undegraded conditions. The ESDs, while not quantitative like LANDFIRE models, do describe effects of common ecological stressors or management practices.

2.6. Identify Indicators for Restoring Ecological Integrity (Step 4)

A primary goal for native ecosystem managers is the maintenance and restoration of ecological integrity [33]. A given ecosystem has integrity when its dominant ecological characteristics (e.g., elements of biotic composition, structure, function, and ecological processes) occur within their natural ranges of variation and can withstand and recover from most perturbations imposed by natural environmental dynamics or human disruptions [33]. Therefore, ecosystems with high levels of integrity have high levels of resistance and resilience to disturbances to which they are adapted.

Our use of the concept of ecological integrity was defined with related terminology by Parrish et al. [33], Unnasch et al. [8], and SERA [34], and related concepts of community assembly were found in Suding et al.’s [35] work, as well as the references within their publication. In the practice of ecological integrity assessments, one might need to assess a number of both biotic and abiotic ecosystem attributes. For example, one might need to address soil degradation, groundwater pumping, the wildfire regime, and the loss of nitrogen-fixing plant species, invasive species, and overall native vs. non-native species compositions—among other factors—to evaluate current conditions. It may also be that this would just be the starting point for evaluating potential effects of climate-induced stress that is projected for upcoming decades.

The Ecological Integrity Assessment Framework (EIAF), as described by Unnasch et al. [8] and used globally through the Conservation Measures Partnership (https://www.conservationmeasures.org/, accessed on 15 July 2025), provides a structured approach to identifying ecologically relevant measures for integrity assessments with direct application to restoration sites.

Relevant to reference model development, the EIAF describes identification of

(1)

Key Ecological Attributes (KEAs) driving natural states and transitions;

(2)

Measurable indicators of those attributes;

(3)

The expected/historic range of variation for each condition indicator;

(4)

Practical thresholds for indicator measurements suggesting categories for the current status and potential restoration milestones.

2.7. SER Five-Star System and Ecological Recovery Wheel (Step 4, 6)

We applied outputs of the data, frameworks, and models (above) to the SER Five-star System [3,36]. The system is a mechanism used to document the level of system recovery, from partial to full recovery. It works with the Ecological Recovery Wheel graphic that shows five condition levels for key ecosystem attributes (e.g., structural diversity) and sub-attributes (e.g., spatial mosaic) [sensu 3]. This was used here for documenting both baseline site conditions and for preliminary restoration goals.

2.8. Vulnerability and Adaptation to Climate Change (Step 5)

Climate change vulnerability is commonly defined as “the degree to which a system is susceptible to, or unable to cope with, adverse effects of climate change, including climate variability and extremes” [37]. Vulnerability assessments tend to include a series of measurements to quantify climate change exposure, sensitivity, and adaptive capacity. Exposure tends to be measured by modeled projections of climate and direct effects on the relevant environment (e.g., drought, sea surface temperature, etc.). Sensitivity measures tend to focus on likely interactions of exposure and the system characteristics, such as already degraded ecosystem integrity. Adaptive capacity measures tend to consider ecosystem attributes that lend resiliency to the ecosystem type, such as types that naturally support high diversity of functionally important species, like nitrogen-fixing plants. While under development only recently, applicable vulnerability assessments are increasingly available in many parts of the world [38].

We used outputs from NatureServe’s Habitat Climate Change Vulnerability Index [39], as they are applicable to our focal ecosystem type. These results directly inform adaptation strategies, such as selecting species for reintroduction or augmentation or restoring natural disturbance processes [40,41].

3. Results

Site definition was included above in our case study, so here we start by summarizing a stepwise workflow for organizing and applying available information. We then report on results using ecological classifications and reference site data. See Supplementary Materials for additional detail of results covered below.

3.1. Classification and Reference Sites (Step 2)

The NPS vegetation inventory indicates vegetation in distinct categories, including invasive cheatgrass, two-needle pinyon (Pinus edulis)—Utah juniper (Juniperus osteosperma) multi-shrub woodlands, four-winged saltbush, black sagebrush, and grasslands. Site visits clarified that black sagebrush was limited to a portion of the restoration area with alkaline bedrock outcroppings and related soil in a small portion of the overall site. Similarly, four-winged saltbush was found along with greasewood (Sarcobatus vermiculatus) in shallow washes from intermittent streams draining down from adjacent tablelands (Supplementary Materials).

Most of the restoration site falls within polygons mapped as two-needle pinyon–Utah juniper “multi-shrub” woodlands, and they defined the ecotonal transition from adjacent pinyon–juniper woodlands on steep slopes onto the gentle plain. These areas appear to have supported big sagebrush (Artemisia tridentata ssp. wyomingensis or ssp. tridentata) as a dense shrubland or open grassy steppe; but it is currently dominated by invasive cheatgrass. Local plant communities, as defined at the association level of the USNVC [22], that most closely match likely local conditions include Artemisia tridentata ssp. wyomingensis/Pleuraphis jamesii shrublands and Artemisia tridentata ssp. wyomingensis/Hesperostipa comata Colorado Plateau shrublands.

3.2. Reference Site Data Describing the Types to Be Restored (Step 2)

NPS vegetation inventory for the Monument included 239 samples labeled to USNVC Associations that were either dominated by Wyoming big sagebrush or by Pinyon–juniper woodlands that included sagebrush in the understory. These assist with characterizing local floristic composition in our reference model. Supporting Information includes a list of characteristic plant species in these data that include Wyoming big sagebrush and other shrubs along with common grass species in these sites. These were augmented with type descriptions from the USNVC Associations (Supplementary Materials).

The LANDFIRE BpS map indicates pinyon–juniper woodlands and big sagebrush types occurring throughout the Monument and in adjacent lands. Big sagebrush shrublands are mapped in small patches within the project area, but they are likely an artifact due to the site being a long narrow strip; most are mapped with the biophysical setting of the pinyon–juniper woodland occurring on adjacent slopes (Supplementary Materials). Therefore, we inferred that the restoration site would have been a big sagebrush shrubland and steppe had historical degradation not occurred. Using the NatureServe ecological systems’ classification, this would equate with Inter-Mountain Basins Big Sagebrush Shrubland (https://explorer.natureserve.org/Taxon/ELEMENT_GLOBAL.2.722895/Inter-Mountain_Basins_Big_Sagebrush_Shrubland, accessed on 15 May 2023) (Supplementary Materials). We will refer to this as our focal ecosystem, which is the focal unit for efficient analysis.

3.3. State-and-Transition Models (Step 3)

The description of the LANDFIRE biophysical setting of Inter-Mountain Basins Big Sagebrush Shrubland—Upland (BpS 10804) indicates that wildfire is the primary natural disturbance agent, characterized by replacement fires in all succession classes, although fire return intervals (FRIs) vary by class. Since this BpS can occupy vast areas, historic disturbances (fire) likely ranged from small (<4 hectares) to very large (>4000 hectares) depending on conditions, time since last ignition, and fuel loading (Supplementary Materials).

Five natural successional classes are described, including early, mid, and late stages, some with open and others with a closed shrub canopy. In each area, not all stages from A-E may occur. One pathway leads to late development with a closed canopy, while another retains open canopy conditions. Modeling that simulates a natural fire regime results in the following percentage (in parentheses) of each succession class one could expect across a given landscape supporting this biophysical setting. Percentages are approximate and should be applied assuming a 5% range of variation:

(A)

Early Development 1 All Structures (15% of this type in this stage)

(B)

Mid Development 1 Open (shrub-dominated—50% of this type in this stage)

(C)

Mid Development 1 Closed (shrub-dominated—25% of this type in this stage)

(D)

Late Development 1 Open (5% of this type in this stage)

(E)

Late Development 1 Closed (5% of this type in this stage)

That is, about 15% of a given land area would likely occur as early successional stage development, and 50% would occur at the mid stage but with an open shrub canopy. The remainder would occur as succession classes C–E. See Supplementary Materials for detailed description of the LANDFIRE model.

The NRCS Ecological Site Description most applicable here is R034BY306UT Upland Loam (Wyoming Big Sagebrush) (Figure 2). While less quantitative than the LANDFIRE model, this conceptual model describes a “Reference state” with three common natural conditions, including big sagebrush and varying densities of shrubs vs. grasses. It also describes a “Current Potential State” that could include effects of livestock and vegetation management. A related “Pinyon-Utah Juniper State” results from improper livestock grazing over time, and with surface disturbance and short fire return intervals, an “Invasive Annual Grass State” may dominate. With directed management or restoration actions, a “Seeded State” can be produced. Within each of these main states, internal dynamics are expressed in terms of direction change based on various natural factors, stressors, and types of vegetation management (Supplementary Materials).

Among the primary factors resulting in change among these states are succession (or “time without disturbance”), fire (presence, short interval, or long interval), insect herbivory, establishment of non-native species, improper livestock grazing, and surface disturbance. Vegetation management is expressed as either vegetation manipulation or prescribed grazing. A number of these dynamics could be affected by climate, in this case simply stated as with “drought” or “wet” conditions.

The current conditions of the restoration site could be mostly characterized as the “Invasive Annual State” in the NRCS model. It likely arrived at that state through surface disturbance, past improper grazing, and the introduction of non-native species. The model suggests that with continued time without fire and improper livestock grazing, a transition toward a pinyon–Utah juniper state would be likely. There is also the potential of vegetation manipulation shifting the site toward a “Seeded State.” Generally, this refers to areas that were seeded with a standard range seed mix, which could include non-native species, with an emphasis on soil stabilization and production for grazing or haying. However, the use of non-native species, and especially invasive species, is generally inconsistent with accepted practices for standards-based ecological restoration. These states as described could be stagnant in terms of recovery to a restoration goal within that recovery trajectory, or they could contribute to further degradation. The NRCS ESD provides general suggestions and implications (likelihood of success and changes in production) for types of species that could be seeded. Therefore, a restoration planner could use this section of the model to include desired species and restoration actions to recover ecosystem integrity.

3.4. Key Ecological Attributes and Indicators of Ecological Integrity (Step 4)

Following the EIAF process for the Monument restoration site, key ecological attributes for this big sagebrush shrubland restoration site can include the following:

• Landscape-scale processes requiring connectivity.
• Community composition contribution to characteristic ecological processes.
• Vegetation structural responses to disturbances.
• Wildfire regimes.
• Chemical and physical processes.

Both the condition and stressor-based indicators [8] have been identified to measure these attributes (

Table 1. Indicators used in big sagebrush shrubland restoration. This illustrates NatureServe’s ecological integrity criteria, which are structured into categories compatible with SER guidance. Stressor indicators are distinguished with italics. See Supplementary Materials for detail (Source: NatureServe). * Modeled Landscape Condition from Hak and Comer [42]).

Ecosystem Attribute	Sub-Attribute(s)	Indicator (Stressors in Italics)
Absence of threats	Invasive Species	Percent Cover of Invasive Annual Grass Species (Requires additional information on off-site sources)
Physical Conditions	Substrate physical/chemical	Soil Permeability and Aggregate Stability Percent Bare Ground
Species Composition	Desirable plants No undesirable species	Percent Cover of Native vs. Non-Native Plant Species Observed vs. Expected Vascular Plant Species Composition Observed vs. Expected Presence of Plant Associations Percent of Cover of Invasive Annual Grass Species
Structural Diversity	All vegetation strata Spatial mosaic	Invading Tree Density Observed vs. Expected Cover Biological Soil Crust Vegetation Departure Index
Ecosystem Function	Resilience/Recruitment	Vegetation Departure Index (including fire frequency and intensity)
External Exchanges	Landscape and Gene Flows, Habitat Links	Modeled Landscape Condition * (Composite Indicator)

). For example, the Percent Cover of Invasive Annual Grass Species is a primary “stressor”-based indicator of the community composition to be applied in the field at this restoration site. Importantly, “stressor”-based indicators would not have any expected natural range of variation, but instead, they occur along a continuum from lesser to greater degrees of stress based on the resource. That continuum is likely to be correlated with a range of variation among other “condition”-based indicators under the ecosystem attribute “Species Composition,” like the Percent Cover of Native Plant Species or the Observed vs. Expected Vascular Plant Species Composition.

An indicator for the key ecological attribute of the vegetation structural response to disturbance—wildfire regime—is constructed as an indicator of the “condition” through a model of natural wildfire dynamics, but it measures departure from those presumed natural dynamics (arguably as an expression of ecosystem stress).

We note below that “expert judgment” comes into play in our process. Here, expert judgment reflects the consensus of researchers and practitioners with extensive experience and expertise with the ecosystem or restoration practice. Much of this has already been incorporated into some of the data sources and frameworks presented here.

3.4.1. Expected/Historical Ranges of Variation for Indicators

The expected range of variation for the selected condition indicators for big sagebrush shrublands are ideally described through comparative quantitative analysis of many reference locations. However, expert judgment can be used to establish initial approximations of these ranges. For example, with the Percent Cover of Native vs. Non-native Plant Species, a range from 50% to 90% could adequately express common conditions for big sagebrush shrublands (Supplementary Materials Table S13).

The LANDFIRE Vegetation Departure Index (scored 0.0–1.0) addresses the effects of altered fire regimes on expected proportions of vegetation structural classes based on assumptions documented in LANDFIRE’s state-and-transition models (see above). Therefore, the model expresses the range of variation one would expect in structural classes given documented assumptions of ecosystem succession and disturbances from natural wildfires. These models developed for rangewide application to a given biophysical setting/vegetation type could be further refined to local site conditions with local knowledge.

3.4.2. Indicator Thresholds

While often hard to identify, one can presume that critical thresholds might exist within the range of potential variation for each indicator of each key ecological attribute. See Supplementary Materials and Table S13 for the most detailed breakdown of the indicator rating criteria applicable to this project. There are a variety of ways to express thresholds that define the expected ranges of variation in the indicators for each key ecological attribute of a focal native ecosystem. Some have used 3–5 generalized categories to characterize a range of “excellent” to “poor” conditions. One could equate “poor” conditions to a “threshold of imminent loss.” This is a hard threshold, suggesting some form of ecological collapse at the site [43]. Restoration may be initially focused on crossing this threshold from a current, apparently “collapsed” state to a state falling within the expected range of variation. Again, thresholds within the expected range of variation for the selected condition and stressor indicators for big sagebrush shrublands are ideally described through comparative quantitative analysis of many reference sites. However, once again, expert judgment can be used to establish initial approximations of threshold values and categories. For our example of the Percent Cover of Native vs. Non-Native Plant Species, a set of thresholds could adequately express a set of conditions for big sagebrush shrublands, i.e., “poor” (<50%), “fair” (50–80%), “good” (80–90%), and “excellent” (>90%).

The above measure does not specify or distinguish “non-native” from “invasive” plant species, but it applies more generally to the many “non-native” species that may be present. However, in this common dryland case study, the role of invasive annual grasses is well known and commonly addressed as an indicator of ecosystem stress. For this and other stressor-based indicators, expert judgment can be used to establish a parallel gradient, such as for the Percent Cover of Invasive Annual Grass Species; a set of thresholds could adequately express a set of conditions for big sagebrush shrublands, i.e., “poor” (>10%), “fair” (3–10%), “good” (1–3%), and “excellent” (<1%).

For the LANDFIRE Vegetation Departure Index, which was measured either through remote sensing and spatial modeling of vegetation structure or by field observation and estimation, the 0.0–1.0 index is tentatively thresholded into four categories: <0.3, suggesting a “poor” or severe fire regime departure; 0.31–0.6 for a “fair” or moderate to severe departure; 0.61–0.9 for a “good” or low to moderate departure; and 0.91 for an “excellent” or no departure.

3.4.3. Scorecard of Indicators and Ratings

We used this information to suggest baseline conditions of indicators as shown in

Table 2. Use of the ecological integrity assessment framework’s indicator thresholds to inform site baseline conditions (sensu Gann et al. [3]). This illustrates additional details from the SER guidance associated with these same categories.

Attribute Category	Star Level (1–5)	SER Ranking Guidance	Evidence for Recovery Level
Absence of Threats
Invasive species	1	Some invasive species drivers (e.g., planting or releasing invasive species, contaminated equipment or supplies) are absent, but others remain high in number and degree (e.g., >25% relative cover of reproductive invasive plants at the site).	Invasive species abundant (>10% absolute cover). Indicative of seedbank and offsite inputs.
Physical Conditions
Substrate physical/chemical	2	Physical properties of substrates (e.g., soil structure and layers, topography, erosion, compaction, and temperature) remain at low similarity levels relative to the reference, but they are capable of supporting some characteristic native biota.	Proportional area: 40–60% within the expected range. Bare soil areas are substantial, exacerbated by the loss of soil crusts and contributed to long-lasting impacts.
Species Composition
Desirable Plants	2	A small subset of characteristic native plant, fungi, and lichen species present (e.g., >25% richness and evenness of the reference) across the site. Very high levels of nonnative, invasive or other undesirable plants (e.g., >50% relative species richness, abundance, or cover) or non-native or undesirable animals.	Cover of native plants: <50%. Observed vs. expected vascular plant species composition: <50% similarity. Observed vs. expected presence of plant associations: <50% similarity.
No undesirable species	1		Invasive species abundant (>10% absolute cover).
Structural Diversity
All vegetation strata	2	Multiple strata of the reference are present and remain at low similarity levels relative to the reference, but they are capable of supporting some biota of the reference. Some similarity of spatial distribution of features (e.g., vegetation, animal populations, habitats) is seen relative to reference throughout most of the site.	Biological soil crust is present in protected areas and with a minor component elsewhere: 30–60% of the area with 0–5 trees expected per hectare. Average VCC Score = < 0.3 (severe departure).
Spatial Mosaic	2
Ecosystem Function
Productivity/Cycling	2	Low numbers and levels of physical and biological processes and functions (e.g., photosynthesis and growth, water and nutrient cycling), relative to the reference are present.	Biological soil crust is present in protected areas and with a minor component elsewhere.
Resilience/Recruitment	1	Processes and functions (e.g., water and nutrient cycling, habitat provision, appropriate disturbance regimes, and resilience) are at a foundational stage only, compared to the reference model. Resilience and recruitment are at a very foundational stage compared to the reference.	Average VCC Score = < 0.3 (severe departure).
External Exchanges
Landscape Flows/Habitat Links	3	Positive exchanges and flows between the site and the surrounding environment (e.g., the number of species, water, and fire) in place of intermediate levels of characteristic species and processes are seen (e.g., >50% of the reference).	Landscape condition model score: <0.5. Note: The landscape condition model creates a spatial surface with the lowest scores closest to the most intensive land uses. For the Monument site, it mainly indicated proximity to nearby urban development. Within the Monument, conditions are better, consistent with better connectivity. Average VCC score = < 0.3 (severe departure).

and as a Baseline Recovery Wheel in Figure 3. While the ecological integrity and SER five-star frameworks are not intended to translate exactly, we can set the highest levels of ecological integrity to the highest (five stars) level of recovery. Again, see Supplementary Materials Table S13 includes the parallel categories of “A = Excellent to D = “Poor”, roughly equating to the SER categories of “5” down to “1”.”

Lower levels of recovery require not only information from our analytical frameworks but also an understanding of site management, including a reduction in stressors, initial treatments, and related recovery. Our preliminary baseline scores are also based on secondary information, including a site visit and personal communications with Monument staff. Five-star categories contain two or more sub-attributes, requiring expert judgment to determine respective recovery levels. For example, EIAF analyses show a loss of integrity when the ratio of native to non-native plant cover is less than 50%. The five-star category for desirable plants has a different characterization of species presence, not cover, beginning with “some colonizing species present” at the one-star level of native species recovery. This site likely includes more than 10% of the preferred species, but we assign only a single star for species composition in Figure 2. To make this determination, we considered the high levels of undesirable, invasive species; the continued spread of these species; management focus on the areas of greatest degradation within the site; and the lack of a regeneration niche. Separate attempts to re-establish native species have yielded mixed results, including poor outcomes of seeding trials with connectivity modifiers in 2017 (PJC and GEE personal observation).

3.5. Accounting for Climate Change (Step 5)

Using outputs from NatureServe’s Habitat Climate Change Vulnerability Index [39], the climate change vulnerability assessment for our big sagebrush shrubland provides insights into likely climate change that affects the development of our reference model. Supplementary Materials includes maps of the projected change in suitability for this shrubland across its extensive range, and one can see the variation in this changing suitability for the area encompassing the Monument.

Looking out to the 2035–2065 timeframe, this shrubland type in this part of the Colorado Plateau scores in the moderate level of overall vulnerability. The mean annual temperature is projected to increase to 2.7–3.4 °C, with similar increases projected for mean temperatures of the coldest quarter and the minimum temperature of the coldest month. However, precipitation of the driest month is also projected to increase. While this increased precipitation would not fully mitigate increases in temperature, the projected climate change exposure appears to be moderate.

See Schmitz et al. [40] for additional guidance on steps for establishing climate-informed restoration goals. These outputs can be interpreted using emerging frameworks for climate change adaptation, such as the resist/accept/direct (RAD) [44] or resistance/resilience/transformation (RRT) frameworks [45]. In locations scoring relatively low in vulnerability, adaptive restoration can take a “resistance”-based approach. These are primarily preventive actions, such as avoiding landscape fragmentation or new introductions of invasive species. At the opposite extreme, very high vulnerability suggests serious consideration of “Direct” or “Transformation”-based strategies. These could include active measures to change the species composition with species suited to emerging climate conditions. However, for this case study, moderate vulnerability for upcoming decades suggests an acceptable or resilience-based stance in restoration. See Table S15 in Supplementary Materials for generalized strategies using these frameworks suitable for application to this big sagebrush shrubland of our restoration site.

This means that the reference model based on knowledge of historical and current conditions remains generally appropriate, at least for the upcoming decades, as the requirement to develop project goals and decision documents to implement restoration actions. With advancing climate change science, as well as actual climate-driven change in conditions on the ground, the model could require a change in the future, but for the current planning cycles extending into the 2030s, no significant change is indicated.

3.6. Preliminary Reference Goals by Ecosystem Attribute and Sub-Attribute (Step 6)

We next used our full model to suggest a progression of recovery goals for selected attributes. For example, many restoration site indicators might initially score within the “poor” EIAF category. For example, in Figure 3, indicators under “no undesirable species” and “resilience and recruitment” were displayed at the lowest #1 category on the SER Ecological Recovery Wheel. The “fair” category may be established as an interim measurable restoration goal. In this instance of the big sagebrush shrubland at the Monument, each of these indicators could be documented and, following the direction of the SER Standards, used to work towards “continuous improvement to the highest level of recovery attainable.” Subsequent measures could form a series of restoration goals within the “Good” category, falling within levels 3 and 4 for five-star recovery goals and reporting (

Table 3. Indicators (condition vs. stressor) used in big sagebrush shrubland restoration. Desired measurements of these indicators are quantitative expressions of a given indicator, presented here based on project analyses. See Supplementary Materials for detail.

Ecosystem Attributes	Sub-Attributes	Indicator (Stressors in Italics)	Measurements
Absence of threats	Invasive species	Percent Cover of Invasive Annual Grass Species	Invasive species prevalent (3–10% absolute cover)
Physical Conditions	Substrate physical/chemical	Soil permeability and Aggregate Stability Percent Bare Ground	Proportional area: 61–80% within an expected range
Species Composition	Desirable plants/no undesirable species (community composition contribution to characteristic ecological processes)	Percent Cover of Native vs. Non-Native Plant Species	Cover of native plants: 80–90%
Structural Diversity	All vegetation strata/ spatial mosaic vegetation structural response to disturbance	Invading Tree Density	61–90% of the area with 0–5 trees expected per hectare
Ecosystem Function	Productivity/cycling	Observed vs. Expected Cover Biological Soil Crust	Biological soil crust is present in protected areas and with a minor component elsewhere
	Resiliency/recruitment (wildfire regime)	Vegetation Departure Index	Average VCC Score = 0.61–0.9 (Low to Moderate Departure)
External Exchanges	Landscape flows/habitat links (landscape-scale processes requiring connectivity)	Modeled Landscape Condition	Landscape Condition Model Score: 0.80–0.5

and Figure 4). For example, in Figure 4, indicators under “no undesirable species” are now displayed at the higher #3 category, and “resilience and recruitment” are now displayed at the higher #4 category on the SER Ecological Recovery Wheel.

4. Discussion

4.1. Challenges in Ecological Restoration

Ecological restoration is challenging in nearly all circumstances. With accelerating environmental change due to land use and climate-driven stress, practitioners need ready access to analytical frameworks and data to organize information and establish practical and measurable restoration goals. We used this example from the Colorado National Monument, along with commonly available information in the USA, to illustrate the development of reference models and identify measurable goals to restore a degraded site. Our example is typical for semi-arid sites across the United States and beyond, where the current condition of the site can obscure likely pre-degradation conditions, data from applicable reference sites may be limited or challenging to locate, and climate-driven stress may add considerable uncertainty. Therefore, the integration of multiple common forms of information is key to establishing a reference model.

This paper illustrates a process that is directly replicable not only on sites with similar arid ecological conditions but also on all terrestrial ecosystems worldwide using comparable information. The specific reference sources available for this site are available for over 600 described and mapped terrestrial ecosystem types (https://explorer.natureserve.org/Search, accessed on 15 July 2025) (forests, shrublands, grasslands, etc.) across the entire USA. Therefore, the same type of analysis could be completed in about the same form that we presented for this one ecosystem type.

In this case study, source information from ecological classifications and vegetation inventories, along with frameworks and data for assessing ecological integrity, climate change vulnerability, and climate adaptation, was brought together to assist with evaluating site conditions and clarifying restoration goals. The SER Five-star System and the Ecological Recovery Wheel provided a concise form to communicate both current conditions and restoration goals. While additional effort will be required by local project managers to work with stakeholders and other management considerations, restoration specialists can use preliminary goals to design project workplans and use the identified indicators as the basis for monitoring work progress. The steps followed here can inform restoration projects elsewhere around the world, assisting in the implementation of global principles and standards laid out by SER, the UN, and other international bodies [1,2].

As noted above, the SER Standards define a reference model as the presentation of “the expected condition that the restoration site would have been in had it not been degraded, while accounting for anticipated change” [3]. Here, this includes the text, tabular, and spatial information expressing the condition of the native ecosystem had degradation not occurred. The model specifies the key ecological attributes, indicators, and thresholds among those indicators for measuring ecological integrity. Those indicators provide the mechanism to describe relative degradation in baseline conditions at the site. They also provide potential milestones for evaluating progress along the trajectory of recovery. Additional analysis of climate change vulnerability for our focal native ecosystem provided additional insights for consideration of the potential magnitude and rate of vegetation change induced by climate stress. The relative level of likely climate-induced stress had direct implications for setting restoration goals for the site.

4.2. Coping with Uncertainty

Each step of our outlined process may involve considerable uncertainty to restoration practitioners. This is to be expected regardless of whether the restoration project is in “data rich” or “data poor” circumstances. However, there are several approaches that we recommend to others for coping with this uncertainty.

First, it is worth acknowledging that US agencies like the National Park Service have legal mandates requiring the use of the “best available” information in decision-making. This generally exerts pressure to both invest in and use appropriate data sets. They also encourage the use of “adaptive management” principles [46], where, practically speaking, every decision reflects an experiment where uncertainty is acknowledged, and both scenario planning and monitoring are deployed to reduce uncertainty and enable ongoing improvement.

Second, there are always gaps in available data. It is therefore important to document those data gaps as feasible to enable their prioritization (i.e., which data sets should reduce acknowledged uncertainty), and then fill those data gaps as feasible.

Third, we recommend utilizing expert knowledge as it can take many forms, from initial sketches of reference models to estimating parameters within state-and-transition models and indicator thresholds and documenting and prioritizing data gaps. In most cases, this can be essential to advancing reference models and support effective restoration planning.

4.3. Coping with Climate-Induced Stress When Developing Restoration Goals

Our case study provided an opportunity to explore the implications of climate-induced stress on restoration planning and goals. For our case study site, the overall indication of climate change vulnerability appears to be moderate—at least up through the mid-21st century timeframe; one can more clearly see how existing frameworks for climate change vulnerability and adaptation could interact with those used here for developing reference models and measurable restoration goals across a range of circumstances. As noted above, a ‘low-to-moderate’ climate change vulnerability assessment can suggest proceeding with outputs of the EIAF and SER frameworks and other “resilience”-based strategies (sensu Schmitz et al. [40]).

In the increasingly common case where climate change assessments suggest high to very high vulnerability [39], practitioners can first refer to a decision tree (Figure 3 provided in the SER Standards) [3]. This tool addresses a range of scenarios for long-term changes to underlying environmental conditions and guides readers through options for identifying appropriate reference conditions and types of restorative actions. Other tools provide more specific strategies to incorporate climate-related strategies for restoration, where some form of ecosystem transformation [41] is either in progress or foreseen over upcoming years or decades.

While the climate change assessment provides sufficient detail, both thematically and spatially, for components of climate change exposure, sensitivity, and adaptive capacity for the degraded ecosystem being considered for restoration, alternative reference models may be established. These alternative models can be communicated in terms of indicators where the restoration of ecosystem conditions that take climate change into account may no longer be achievable (e.g., historical species composition or structure). In these cases, entirely new goals may be needed (e.g., new climate-induced natural disturbance regimes) [3,35,40]. Documented species compositions and the dynamics of naturally adjacent ecosystems can be one ready source for adaptive responses. In all cases or scenarios, an understanding of the role and status of the ecological attributes provides direction for managers and restoration practitioners as they evaluate adaptation options.

5. Conclusions

This case study served to illustrate a structured approach to compile information to define reference conditions for a restoration site to subsequently identify restoration goals. Given both existing and emerging challenges, the use of practical analytical frameworks and all applicable data will be of increasing importance to restoration practices. We also agree with an emerging call for placing ecological restoration projects within a broader socio-ecological context to advance “long-lasting benefits to people and nature across time and place” [47]. The outputs of this analysis feed directly into subsequent sanctions by Monument staff as they explore management alternatives with stakeholders. The outputs are also being incorporated into arid land restoration training modules for use beyond the Monument.