Abstract
White pine blister rust (WPBR) is one of North America’s most damaging tree epidemics. Aggregating data from more than 80 independent studies across the western U.S. from 1995–2024, we estimate WPBR risk for high-elevation five-needle pine species (High-5) from 1980–2023 in the adaptive management tool RustMapper. WPBR risk is the probability of observing WPBR on the High-5. Stream density, topography, hardiness zone, precipitation, air temperature, vapor pressure deficit, and relative humidity were critical in estimating WPBR risk. WPBR risk increased with moisture and declined with temperature. Across the High-5 range, suitable conditions were found in areas where the disease had not yet invaded and throughout regions where the disease was well established. As a result, the mean risk for WPBR was much higher in the north (~0.6) compared to the southern portions of the High-5 range (~0.15). These findings indicate cautious optimism for disease mitigation success in regions where the disease is established and urgency for proactive management where WPBR occurrence is currently low.
Similar content being viewed by others
Background & Summary
White pine blister rust (WPBR), caused by the non-native pathogenic fungus Cronartium ribicola (J.C. Fischer ex Rabh.), is one of the most damaging tree epidemics in North America1,2. Since its introduction to western North America around 1910 into the Pacific Northwest via infected nursery stock imported from France3, C. ribicola continues to spread and infect new populations of five-needle white pines. The incidence of C. ribicola infection is not uniform on the landscape due in part to the distance from the point of introduction, the biology of the pathogen and hosts, and the environmental conditions required for successful spore production, transport, germination, and infection4.
Past efforts to contain the spread of C. ribicola failed5; the pathogen is now a permanent resident of North America. While its spread cannot be curtailed, management interventions can reduce the impacts of the disease on pine populations and ecosystem services. WPBR risk or hazard, host species adaptive capacity, and forest health condition of populations and ecosystems can affect the likelihood of treatment effectiveness6,7,8. Consequently, incorporation of these factors into management frameworks to prioritize limited resources has been recommended7,9,10. However, reliable geographic projections of current and future WPBR risk still need to be improved for efficient management planning10.
It has been hypothesized that warmer and drier conditions may reduce the prevalence of WPBR due to the pathogen’s requirement for cool, moist conditions to facilitate infection2,11,12,13. Consistent with this hypothesis, studies in the Rocky Mountains have found that trees in more arid habitats are less likely to be infected by C. ribicola14,15. However, for trees already infected in areas experiencing increasing aridity, the likelihood of developing severe symptoms and mortality rises15. In contrast, Dudney et al. (2021) proposed that increased aridity contributed to a shift in WPBR prevalence uphill in the Southern Sierra Nevada Mountains16. Additionally, warming temperatures may exacerbate WPBR in moist habitats by extending the growing season, allowing for a longer infectious season17 and greater canker expansion18. These findings suggest that understanding WPBR risk will be complex but provide support for customizing landscape restoration strategies by local climatic factors4,9,19.
To understand patterns in disease risk, we evaluate the conditions that have contributed to the presence of WPBR in high-elevation five-needle white pines (High-5, hereafter) to project risk across the High-5 distribution for improved analyses of large-scale patterns. With host groups co-occurring throughout the High-5 range and low genetic disease resistance in white pines, the major determinants of risk are the prevalence of C. ribicola inoculum20 and the conditions associated with spore production, dispersal, and host infection21. Past studies suggest that WPBR in trees may have reached the limits of the outbreak due to climatic constraints11,12,13, while others suggest suitable conditions exist beyond the current disease distribution22. In this study, we evaluate where conditions have been suitable for WPBR occurrence from 1980 to 2023 in a tool called RustMapper.
Methods
WPBR Pathosystem
The Cronartium ribicola life cycle is complex, involving two hosts and five spore stages. The primary infection of pine hosts occurs via fungal infection through needle stomata. Following an incubation period of one or more years, the rust fungus colonizes stem tissues23,24, initiating the formation of spermatia and, subsequently, aecia, which erupt through the bark during the spring and summer months, causing the characteristic canker. As the disease progresses, the canker elongates, spreading along the branch toward the main stem and, in many cases, may result in girdling lesions that lead to branch dieback or tree mortality25,26. Upon rupture of the aecia, airborne aeciospores are released, potentially infecting alternate hosts, most commonly Ribes species, provided environmental conditions are favorable. On the alternate hosts, C. ribicola produces uredinia, which may undergo multiple generations (ranging from one to seven), followed by the formation of telia. Basidiospores, released from the telia mid-summer through fall, are the only spore type capable of initiating new infections on pine hosts. Importantly, there is no direct transmission of C. ribicola between pine trees; the pathogen’s lifecycle is obligately heteroecious, requiring both pine and alternate hosts to complete its development. C. ribicola colonizes the deciduous leaves of alternate hosts, where it sporulates, but the pathogen does not cause significant impact to the plant, as the infected leaves are shed annually.
Airborne basidiospores can move relatively short distances from alternate hosts to pines27. However, new disjunct infection centers are likely initiated by long-distance aeciospore transport, infection of the alternate host, and subsequent basidiospore production and infection of the pines, conditions permitting. Aeciospores produced on the pine host can travel several hundred kilometers to infect a susceptible alternate host3. Analysis of atmospheric currents suggests that an isolated infection center on southwestern white pine (P. strobiformis Engelm.) in southern New Mexico originated from aeciospores transported in the upper atmosphere from central California28. Long-distance transport of spores may also explain the recent disjunct infection center in south Colorado on Rocky Mountain bristlecone pine (P. aristata Engelm.) and limber pine (P. flexilis James)29. These successful colonizations demonstrate that areas once thought too dry for the disease can be exposed to inoculum and be environmentally suitable for infections and disease development on pines.
In contrast, a failed founding event has been documented in Utah. Alternate hosts (Ribes inerme Rydb.) were found infected with C. ribicola in 2005 at a site 257 km south of the closest known diseased pines30, yet there is still no evidence of diseased five-needle pines in all of Utah9. These patterns of infection and disease indicate that areas at and beyond the current geographic extent of WPBR on pines have some level of exposure to inoculum. In plant pathology, inoculum refers to the part of the pathogen that can cause infection31. The risk of infection of the pine host and subsequent disease development is associated with environmental conditions and basidiospore inoculum abundance, which is a function of aeciospore availability20. Exposure of the alternate hosts to higher levels of aeciospores is more likely in areas with more regionally infected trees with active sporulating cankers. The occurrence of “wave years”, intense and widespread WPBR infection, is typically driven by episodic environmental conditions that favor abundant inoculum production and infection11,26,27.
Study system
The focal High-5 species included whitebark pine (P. albicaulis Engelm.; PIAL), limber pine (P. flexilis James; PIFL2), foxtail pine (P. balfouriana Grev. & Balf.; PIBA), Great Basin bristlecone pine (P. longaeva D.K. Bailey; PILO), southwestern white pine (P. strobiformis Engelm.; PIST), and Rocky Mountain bristlecone pine (P. aristata Engelm.; PIAR). We harmonized WPBR data for focal species from independent studies and reports (Table 1) from 1995-2020 across the Western U.S. (Fig. 1). Data sources included published and unpublished data from various federal agencies, universities, and the Hi5 Database (Table 1)32. Only data collected by forest health professionals with expertise in field disease detection were included due to the skill and experience required to identify WPBR infections on the High-5 species correctly. The majority of the data were from plot-based studies (73%). While the study designs varied, all plots used in this analysis recorded the location, assessment date, species present, and presence/absence of WPBR on High-5 species within plots. Inconsistencies in reporting methods for other disease metrics precluded including those measures in our analyses while retaining robust sampling across the western U.S. In addition to plot-based data, we also utilized trip reports from Forest Health Protection (FHP) professionals. Trip reports record insect and disease findings and relevant management options for recreation sites, administrative sites, certification stands, project areas, analysis areas, or other areas of interest. Trip reports were included for locations where more recent plot data were unavailable. We developed a point database compiling information on WPBR presence and absence throughout the High-5 species distributions in the U.S.9.
Sample points that recorded the presence or absence of WPBR within the landscape where the disease had been established for decades (black; n = 3926; Established dataset) and in areas with low inferred inoculum abundance at the time of measurement (blue; n = 2752; Invading dataset) based on Jacobi et al.4. The range of the susceptible High-5 species is shown in gray.
Our nearly 6678 WPBR sample points were categorized based on inferred inoculum abundance across the greater landscape and length of pathogen presence4. Therefore, areas with presumed low inoculum were included in the Invading dataset, and regions with presumed high inoculum were included in the Established dataset. The Invading dataset occurred mainly across the southern range of the High-5 (31°–44° latitude), where the disease had not been detected or present over the last 25 years. The one exception is the Sacramento Mountains of south-central New Mexico, where a disjunction WPBR outbreak, discovered in 1990, has become an area of high WPBR impacts28. WPBR data from this mountain range was included in the Established, not the Invading dataset. In contrast, the Established dataset was generally distributed in the more northern High-5 region closer to the point of disease introduction to western North America (Fig. 1), where the pathogen had been present for a long time (>25 years), and the disease was well established.
Modeling approach
We employed the machine learning algorithm Random Forests (RF) to evaluate the conditions associated with the presence or absence of WPBR on the High-5 species across the western U.S. The RF algorithm was introduced by Breiman 200133. A series of bootstrapped datasets are used to generate independent regression trees; at each node, a random sample of predictor variables was selected for use. We utilized the R package randomForest34 to create 500 regression trees using WPBR presence/absence datasets. The fit of each RF model was evaluated with the out-of-bag mean square error (OOB MSE), and variable importance was computed as the amount of prediction error increased when a particular predictor was permuted. Initially, 500 RF trees were generated using all possible predictor variables. The overall model fit was evaluated with the average of the 500 OOB MSEs from the final model, and variable importance was calculated as the average rank of each predictor variable for the 500 models.
Prior to fitting each model, the data was divided into training (70%) and testing (30%) for Established and Invading models. Climate, topographic, and presence/absence of WPBR were included as predictors in a preliminary step where we first estimated RF models with a series of subsets of the candidate independent variables using a modified backward elimination procedure with the package VSURF35. This process reduced the number of variables considered from >100 to less than 10. We explored bi-weekly to seasonal dynamic variables to identify the strongest predictors of WPBR presence. Variables with high importance were primary determinants of the conditions associated with WPBR risk. Variables with low importance or that were highly correlated with more important variables were omitted from the model. We used the mean decrease in accuracy and the mean decrease in Gini to measure variable importance. These essential values reveal which factors significantly impact the probability of conditions conducive to WPBR in the High-5 species. The VSURF approach was used to minimize the number of predictors required for both efficiency and to avoid overfitting36. The VSURF is a three-step variable selection procedure. The first step eliminates irrelevant variables from the dataset, the second step selects all variables related to the response, and the third step refines the selection by removing redundancy in the variables selected by the second step. Following the model fit with the chosen variables from VSURF, we used the testing dataset to measure model accuracy using a confusion matrix from the caret package37.
We used static and dynamic variables in models to define the conditions associated with WPBR on trees (Table 2). Static variables included stream density (STREAM_DEN) and topographic characteristics. We measured stream density across the western U.S. with the USGS National Hydrography Dataset (NHD)38. Stream density is the total length of streams (km) divided by the area of each cell and was estimated by creating a ~1 km grid across the western U.S. and measuring the total length of streams within each grid by taking the intersection of the NHD vector file and the grid using the sf and terra packages in R39,40. The total stream length was divided by the area to get stream density in km km−2. Stream density was used as a proxy for Ribes species commonly occurring in riparian corridors, including Ribes hudsonianum Richardson and R. inerme, both highly susceptible species in western states restricted to moist sites41. Topographic information was obtained with the package elevatr for each WPBR sampling location42. The elevatr package provides access to digital elevation models from the Shuttle Radar Topography Mission (SRTM), the Global Multi-resolution Terrain Elevation Data, and the ETOPO1 global relief model. An elevation raster layer was downloaded using the function get_elev_raster, and the topographic position index (TPI) and the terrain ruggedness index (TRI) were calculated using the spatialEco package. The TPI calculates topographic position using mean deviations within a specified window (s = 29), and the TRI is the topographic roughness index43. Topographic information was extracted for all WPBR sample points.
Dynamic variables were descriptive of the climate. We recorded the USDA plant hardiness zone (H_ZONE) for each sampling location and determined the Plant Hardiness Zone annually from 1980 to 2023 using the minimum winter temperature. Daily daymet climate data were obtained to calculate average climatic conditions44. Daymet provided gridded daily weather and climatology variables (daily minimum and maximum temperature, precipitation, vapor pressure deficit, shortwave radiation, snow water equivalent, and day length) by interpolating ground-based observations through statistical modeling techniques44. The daymet data products are produced on a 1 km × 1 km gridded surface over continental North America and Hawaii from 1980 through the end of the most recent full calendar year. We downloaded daymet data for locations using the daymetR package45. From daily data, we calculated seasonal conditions (April to November) based on the assessment date. We measured the conditions for each climate variable by season: spring (April - May), summer (June - August), and fall (September - October).
The spatial resolution of daymet data was advantageous for capturing high-resolution patterns in the climate at sample locations46. At a 1 km resolution, daymet offered fine-grained information that included small-scale climate features and processes that could affect the presence of WPBR46 and has been shown to capture crucial ecological processes in white pines15,47,48. The resolution of daymet was high enough to capture conditions that can affect exposure and susceptibility to WPBR48, and the resolution was also sufficient to bridge the gap between high-resolution data used to predict outcomes at more extensive spatial ranges. When developing models, high-resolution data generally leads to more accurate and detailed responses to conditions, particularly for capturing small-scale phenomena. Low-resolution data is more cost-effective for large-area monitoring and climate change projections, though they may lack detail. Studies have shown that information is lost with decreased resolution49,50.
Models were parameterized for climate data based on the assessment date and the seasonal means for 5, 10, and 20 years before the assessment date. This was done recognizing that the timeframe that influences disease susceptibility is complex and has yet to be defined. If the models behaved similarly with low standard deviations in the probability of WPBR occurrence (P(WPBR)), we used an ensemble of models developed with 5, 10, and 20-year climate summaries. Ensembles were the mean P(WPBR) across the 5, 10, and 20-year climate summaries. This approach would either integrate short-term and long-term climate effects or identify whether short or long-term effects were more relevant. We also parameterized models with data from the Invading dataset (P(WPBR)INV) or the Established dataset (P(WPBR)EST). We defined an Invading region to project the P(WPBR)INV for ecoregions with low inferred inoculum abundance4. We defined the Invading region, where WPBR was still invading the High-5 landscape, based on the analysis of Jacobi et al., 20184. The boundary of the Invading region was delineated by ecoregion boundaries that best approximated the WPBR establishment history at a landscape scale4. This included the southern portion of the High-5 range in the U.S.4 extending from 31°–44° latitude (Fig. 2). Parameterizing the model for the southern areas with the Invading dataset added resolution to near-term risk analysis for regions with low inferred inoculum abundance.
The Invading region (blue) for the High-5 range (blue and gray) in the U.S. The Invading region, 31°–44° latitude, as defined by the intersection of ecoregions and the High-5 range with low inferred inoculum abundance4.
Models were projected over the High-5 range, which was delineated by the union of tree species distribution maps from Little’s “Atlas of United States Trees” series available in R for the High-5 species51,52,53,54,55,56. The whitebark and limber pine species ranges were obtained from the Whitebark Pine Ecosystem Foundation (whitebarkfound.org), and the range for southwestern white pine was an adapted version from Shirk et al., (2018)54,57. The P(WPBR)EST was projected over the entire U.S. High-5 range (31°- 49° latitude) to assess the probability of conditions conducive to WPBR presence, assuming abundant inoculum availability. The Invading model was projected on the southern High-5 range in the Invading region to determine conditions conducive to WPBR presence under presumed low inoculum availability (Fig. 2). Consequently, this analysis addresses the risk of WPBR to extant forests.
We downloaded gridMET climate data using to project models in space for 1980–202358. While gridMET will not capture microclimates that arise at spatial scales finer than the native resolution (~4 km), this study is focused on understanding landscape-level risk. Using gridded climate data, we predicted annual probabilities of WPBR across the range of High-5 white pines. While there was a loss of spatial detail with the use of gridMET, this product provided a more seamless transition to climate change projections, a future area of research.
Data Records
RustMapper comprises the ensemble P(WPBR) risk layers archived in the Environmental Data Initiative (EDI) repository. The RustMapper consists of 1. annual (1980–2023) P(WPBR)INV based on 5, 10, and 20-year climate summaries; and 2. annual P(WPBR)EST based on 5, 10, and 20-year climate summaries59. The model values are the probability of rust occurrence, ranging from 0 to 1. Higher values denote a greater likelihood of WPBR. Spatial models were estimated with gridMET climate data at 4 × 4 km resolution in GeoTIFF format. Metadata with details of data architecture and attributes was also included.
Technical Validation
Data were distributed across the western U.S., where the High-5 occurred in regions that represented gradients in precipitation, temperature, vapor pressure deficit, and relative humidity (Fig. 1; Table 2). There were noticeable differences in climate for the Established and Invading datasets (Figs. 3–5). While warm and dry conditions were generally associated with the Invading dataset, conditions for the Established dataset were skewed towards lower temperatures and more precipitation in the spring and fall. Vapor pressure deficit was lower for the Established dataset than the Invading dataset in the fall. Similar temporal and regional patterns were found across the 5-year (Fig. 3), 10-year (Fig. 4), and 20-year climate summaries (Fig. 5). Differences in the conditions in the Invading and Established datasets prompted the development of separate models to estimate the probability of observing conditions associated with the occurrence of WPBR.
Climate summaries (5-Year) for (a–c) precipitation (mm), (d–f) minimum temperature (°C), (g–i) maximum temperature (°C), and (j–l) vapor pressure deficit (Pa) by season (spring, summer, and fall) indicate that the climate associated with sample locations in areas with presumed low inoculum (blue; Invading dataset) and where the disease was established (black; Established dataset) differ. Vertical lines denote the mean.
Climate summaries (10-Year) for (a–c) precipitation (mm), (d–f) minimum temperature (°C), (g–i) maximum temperature (°C), and (j–l) vapor pressure deficit (Pa) by season (spring, summer, and fall) indicate that the climate associated with sample locations in areas with presumed low inoculum (blue; Invading dataset) and where the disease was established (black; Established dataset) differ. Vertical lines denote the mean.
Climate summaries (20-Year) for (a–c) precipitation (mm), (d–f) minimum temperature (°C), (g–i) maximum temperature (°C), and (j–l) vapor pressure deficit (Pa) by season (spring, summer, and fall) indicate that the climate associated with sample locations in areas with presumed low inoculum (blue; Invading dataset) and where the disease was established (black; Established dataset) differ. Vertical lines denote the mean.
The Invading dataset comprised sample points distributed across the southern areas of the western U.S. (Fig. 1). Most of the sample points (83%) had no rust present, making the total WPBR occurrence ~0.17, though this varied in space, ranging from 0 to 1 with a mean of 0.17 (Fig. 6). Unlike the Invading data, most of the Established data had rust present (70%) (Fig. 6). WPBR occurrence for the Established data also varied in space and ranged from 0 to 1, with a mean of 0.70.
The observed WPBR occurrence was estimated from the Invading plot data (a,b) where inoculum was presumed low (WPBRINV) and (c,d) for the Established dataset (WPBREST). Occurrence estimates are only shown for areas where point data was available.
Across the western U.S., we estimate the probability of WPBR using presence/absence data (Table 3). Model performance measured by area under the curve (AUC) ranged between 0.67 to 0.96, and the accuracy of detecting WPBR where it was present ranged between 76–99% for the training data and 71–94% for the testing dataset. The error was lower for the P(WPBR)INV compared to the P(WPBR)EST, with higher accuracies compared to the P(WPBR)EST. The occurrence of WPBR was much lower for the Invading dataset than the Established dataset, and combined with the differences in sample sizes and climate patterns, separate models were appropriate.
Across models, both short-term (5 and 10-year) and long-term (20-year) climate summaries were important (Table 4). Stream density (STREAM_DEN), topography, hardiness zone (H_Zone), precipitation (PRCP), temperature (T), vapor pressure deficit (VPD), and relative humidity (RH) were included in both P(WPBR)INV and P(WPBR)EST models (Table 4 and Figs. 7, 8). Short-term effects on P(WPBR)INV were dominated by fall PRCP, while spring, summer, and fall climate was more influential for the 10 and 20-year risk. Summer PRCP, spring Tmin, and fall VPD were important in some P(WPBR)EST models. Altogether, model performance indicated that landscape attributes and climate summaries could be used to understand conditions that have supported WPBR at the landscape scale where the disease is invading and was well established.
Variable importance plots, the mean decrease in accuracy for the Invading (P(WPBR)INV) and Established (P(WPBR)EST) models, respectively, for (a,b) 5-year climate summaries, (c,d), 10-year climate summaries, and (e,f) 20-year climate summaries in the western U.S.
Variable importance plots, the mean decrease in the Gini for the Invading (P(WPBR)INV) and Established (P(WPBR)EST) models, respectively, for (a,b) 5-year climate summaries, (c,d), 10-year climate summaries, and (e,f) 20-year climate summaries in the western U.S.
The relationship between the P(WPBR) and climate variables generally indicated that probabilities increased with moisture and declined with temperature (Fig. 9). The P(WPBR) increased with stream density, precipitation, and vapor pressure deficit. Hardiness zone and temperature were negatively correlated with the P(WPBR) (Fig. 9). The highest values were at a VPD of approximately 700 Pa and a VPDmax of approximately 1400 Pa, and the P(WPBR) declined at lower and higher values in both models. The drivers of the P(WPBR)INV and the P(WPBR)EST were similar but exhibited significant differences in their magnitude and sensitivity to topography and climate. The magnitude of the P(WPBR)INV was half the P(WPBR)EST. Although the P(WPBR)INV declined with RH, the P(WPBR)EST increased with RH. The P(WPBR)INV was most sensitive to topography, minimum temperature, and vapor pressure deficit, and the P(WPBR)EST was most sensitive to stream density, precipitation, and relative humidity. The small differences in the primary drivers of P(WPBR)INV and P(WPBR)EST (Table 4, Figs. 7, 8) suggests that the WPBR dynamics may be slightly different in the northern and southern landscapes.
Sensitivity analysis for models of WPBR risk in areas with low inferred inoculum (blue) and where the disease was established (black). Landscape variation in (a) stream density (STREAM_DEN), (b) hardiness zone (H_ZONE), (c) topographic position index (TPI), (d) topographic roughness index (TRI), (e) precipitation (PRCP), (f) maximum (Tmax) and (g) minimum (Tmin) temperature, (h) vapor pressure deficit (VPD), (i) maximum vapor pressure deficit (VPDmax), and (j) relative humidity (RH) explained patterns in WPBR risk.
The magnitude of the probability of observing conditions suitable for WPBR across the High-5 landscape differed between the P(WPBR)INV and P(WPBR)EST models (Figs. 9, 10; Table 5). While the P(WPBR)INV in 1980–2023 was below 0.6 for all of the southern High-5 range, P(WPBR)EST was greater than 0.5 for the majority (97%) of the High-5 range. The standard deviation across models that were developed using 5, 10, and 20-year climate data was low (0–0.2) for both the P(WPBR)INV and the P(WPBR)EST. Given the consistent model performance across climate summaries (5, 10, and 20 years), combined with low variability across climate summaries (Fig. 9c,d), results are presented as ensembles of models developed with climate summaries of different lengths for both the Invading and Established models. Comparing the P(WPBR)INV to the P(WPBR)EST where the two models overlap, there was a significant positive correlation (R2 = 0.1; P < 0.001). The high (>0.5) P(WPBR)EST across the entire U.S. High-5 range suggests that if introduced, there are conditions conducive to WPBR across the whole range.
RustMapper ensemble estimates of the probability of WPBR from models parameterized from (a) areas with low inoculum (P(WPBR)INV) and (b) areas where WPBR is well established (P(WPBR)EST). The standard deviation in (c) P(WPBR)INV, and (d) P(WPBR)EST across the High-5 range was low in the western U.S. for 1980–2023.
Usage Notes
Management interventions can reduce the impacts of WPBR on High-5 populations and the ecosystem services they provide6. Current and projected forest conditions can affect the likelihood of success of some treatment actions7. Proactive management offers the opportunity to prepare healthy populations for resilience before extensive ecological impacts occur. The proactive strategy focuses on activities to promote pine regeneration to increase population size, genetic diversity, and adaptive capacity to maintain forest health. Alternatively, restoration strategies target impacted populations to re-establish viable pine populations, natural processes, and ecosystem services8,10,60. For both management approaches, interventions can include regenerating stands through planting or direct seeding; identifying, developing, and deploying WPBR-resistant seed sources; and implementing silvicultural treatments to increase resilience (e.g., generating a diverse mosaic of stand ages across a landscape to enhance forest adaptation).
Decision frameworks require tools that allow managers to prioritize limited resources. RustMapper integrates information on WPBR risk that can be used in mitigation planning for proactive and restoration strategies across boundaries. While the P(WPBR)INV is more appropriate for regions with low presumed inoculum, the P(WPBR)EST should be used in areas with presumed high inoculum. The P(WPBR)EST projections for the more southern U.S. landscapes provide a preview of possible risks if pathogen spread continues and inoculum levels increase. RustMapper can be used to understand current and future risks, identify regions with consistently low risk now and in the future (refugia), define the risk level of specific areas, and select treatment options based on risk and relative chances of success. RustMapper makes this information accessible to resource managers for focused applications and decision support.
Code availability
The code for RustMapper can be found here: Sparkle Malone. (2025). Malone-Disturbance-Ecology-Lab/RUSTMapper: RustMapper: White Pine Blister Rust Risk Across High Elevation Forests in the Western United States (RUSTMapper). Zenodo. https://doi.org/10.5281/zenodo.15212542.
References
Campbell, E. M. & Antos, J. A. Distribution and severity of white pine blister rust and mountain pine beetle on whitebark pine in British Columbia. Can. J. For. Res. 30, 1051–1059 (2000).
Geils, B. W., Hummer, K. E. & Hunt, R. S. White pines, Ribes, and blister rust: a review and synthesis: Review and synthesis. For. Pathol. 40, 147–185 (2010).
Mielke, J. L. White pine blister rust in western North America. Yale School of Forestry & Environmental Studies Bulletin Series (1943).
Jacobi, W. R., Kearns, H. S. J., Cleaver, C. M., Goodrich, B. A. & Burns, K. S. Epidemiology of white pine blister rust on limber pine in Colorado and Wyoming. For. Pathol. 48, e12465 (2018).
Maloy, O. C. White pine blister rust control in North America: a case history. Annu. Rev. Phytopathol. 35, 87–109 (1997).
Schoettle, A. W. & Sniezko, R. A. Proactive intervention to sustain high-elevation pine ecosystems threatened by white pine blister rust. J. Forest Res. 12, 327–336 (2007).
Schoettle, A. W., Jacobi, W. R., Waring, K. M. & Burns, K. S. Regeneration for resilience framework to support regeneration decisions for species with populations at risk of extirpation by white pine blister rust. New Forests 50, 89–114 (2019).
Tomback, D. F. et al. Tamm review: Current and recommended management practices for the restoration of whitebark pine (Pinus albicaulis Engelm.), an imperiled high-elevation western North American forest tree. For. Ecol. Manage. 522, 119929 (2022).
Schoettle, A. W. et al. Integrating forest health conditions and species adaptive capacities to infer future trajectories of the high elevation five-needle white pines. For. Ecol. Manage. 521, 120389 (2022).
Jenkins, M. B. et al. Restoring a forest keystone species: A plan for the restoration of whitebark pine (Pinus albicaulis Engelm.) in the Crown of the Continent ecosystem. For. Ecol. Manage. 522, 120282 (2022).
Kinloch, B. B. White pine blister rust in North America: past and prognosis. Phytopathology 93, 1044–1047 (2003).
Woods, A. J., Heppner, D., Kope, H. H., Burleigh, J. & Maclauchlan, L. Forest health and climate change: A British Columbia perspective. For. Chron. 86, 412–422 (2010).
Sturrock, R. N. et al. Climate change and forest diseases. Plant Pathol. 60, 133–149 (2011).
Thoma, D. P., Shanahan, E. K. & Irvine, K. M. Climatic correlates of white pine blister rust infection in whitebark pine in the greater Yellowstone ecosystem. Forests 10, 666 (2019).
Burns, K. S. et al. Interactions between white pine blister rust, bark beetles, and climate over time indicate vulnerabilities to limber pine health. Frontiers in Forests and Global Change https://doi.org/10.3389/ffgc.2023.1149456 (2023).
Dudney, J. et al. Nonlinear shifts in infectious rust disease due to climate change. Nat. Commun. 12, 5102 (2021).
Hunt, R. S. & Jensen, G. D. Long infection period for white pine blister rust in coastal British Columbia. in Horttechnology vol. 10 530–532 (American Society for Horticultural Science, 2000).
Kearns, H. S. J., Jacobi, W. R. & Geils, B. W. A method for estimating white pine blister rust canker age on limber pine in the central Rocky Mountains. For. Pathol. 39, 177–191 (2009).
Shepherd, B. et al. Climate dominates geographic influences on whitebark pine and limber pine trends and landscape patterns in Canada. Can. J. For. Res. 55, 1–19 (2025).
Zambino, P. J. Biology and pathology of Ribes and their implications for management of white pine blister rust. For. Pathol. 40, 264–291 (2010).
Van Arsdel, E. P. Environment in relation to white pine blister rust infection. in Biology of rust resistance in forest trees: Proceedings of a NATO-IUFRO Advanced Study Institute: Biology of rust resistance in forest trees (eds. Bingham, R. T., Hoffand, R. & McDonald, G. I.) Publication 112; 479–493 (USDA Forest Service, Washington, D.C., USA, 1972).
Kearns, H. S. J. et al. Risk of white pine blister rust to limber pine in Colorado and Wyoming. USA. For. Pathol. 44, 21–38 (2014).
Alfen, N. K. Fungal pathogens of plants. Encyclopedia of Life Sciences https://doi.org/10.1038/npg.els.0000362 (2001).
Kolmer, J. A., Ordonez, M. E. & Groth, J. V. The rust fungi. Encyclopedia of Life Sciences a0021264 https://doi.org/10.1002/9780470015902.a0021264 (2009).
Lachmund, H. G. Mode of entrance and periods in the life cycle of Cronartium ribicola on Finns montícola. J. Agric. Res. 47, 791–805 (1933).
McDonald, G. I. & Hoff, R. J. Blister rust: an introduced plague. in Whitebark pine communities: ecology and restoration (eds. Tomback, D. F., Arno, S. F. & Keane, R. E.) vol. 36 193–220 (Island Press, Washington, DC, 2001).
Van Arsdel, E. P., Riker, A. J. & Patton, R. The effects of temperature and moisture on the spread of white pine blister rust. Phytopathology 46, 307–318 (1956).
Frank, K. L., Geils, B. W., Kalkstein, L. S. & Thistle, H. W. Jr. Synoptic climatology of the long-distance dispersal of white pine blister rust II. Combination of surface and upper-level conditions. Int. J. Biometeorol. 52, 653–666 (2008).
Blodgett, J. T. & Sullivan, K. F. First report of white pine blister rust on Rocky Mountain bristlecone pine. Plant Dis. 88, 311–311 (2004).
Vogler, D. R., Geils, B. W. & Coats, K. First report of the white pine blister rust fungus, Cronartium ribicola, infecting Ribes inerme in north-central Utah. Plant Dis. 101, 386–386 (2017).
Tainter, F. H. & Baker, F. A. Principles of Forest Pathology. (John Wiley & Sons, Nashville, TN, 1996).
Malone, S. L. et al. White Pine Blister Rust (WPBR) Plot Data from the Western United States ver 2. Environmental Data Initiative https://doi.org/10.6073/pasta/70d7cd56799fc79668343e48e87bb9ef (2025).
Breiman, L. RandomForests. Mach. Learn. 45, 5–32 (2001).
Liaw, A. & Wiener, M. Classification and regression by randomForest. R news 2, 18–22 (2002).
Genuer, R., Poggi, J.-M. & Tuleau-Malot, C. VSURF: An R Package for Variable Selection Using Random Forests (2015).
Speiser, J. L., Miller, M. E., Tooze, J. & Ip, E. A comparison of randomForest variable selection methods for classification prediction modeling. Expert Syst. Appl. 134, 93–101 (2019).
Kuhn, M. et al. Package ‘caret’. R J. 223 (2020).
U.S. Geological Survey. National Hydrography Dataset (2019).
Hijmans, R. J. Spatial Data Analysis: R Package Terra Version 1.7. (Comprehensive R Archive Network (CRAN) 2023).
Hijmans, R. J. terra: Spatial data analysis (2023).
Van Arsdel, E. P. & Geils, B. W. The Ribes of Colorado and New Mexico and Their Rust Fungi. vols 4-13 https://research.fs.usda.gov/treesearch/download/66502.pdf (2004).
Hollister, J. W. elevatr: access elevation data from various APIs. R package version 0.4.2. (2022).
Riley, S. J., DeGloria, S. D. & Elliot, S. D. A Terrain Ruggedness Index that Quantifies Topographic Heterogeneity. Intermt. J. Sci. 5, 23–27 (1999).
Thornton, M. M., Thornton, P. E., Wei, Y., Vose, R. S. & Boyer, A. G. Daymet: Station-level inputs and model predicted values for North America, Version 3. ORNL DAAC (2017).
Hufkens, K., Basler, D., Milliman, T., Melaas, E. K. & Richardson, A. D. An integrated phenology modelling framework in R. Methods Ecol. Evol. 9, 1276–1285 (2018).
Thornton, P. E. et al. Gridded daily weather data for North America with comprehensive uncertainty quantification. Sci. Data 8, 190 (2021).
Cleaver, C. M., Jacobi, W. R., Burns, K. S. & Means, R. E. Limber pine regeneration and white pine blister rust in the central and southern Rocky Mountains. For. Sci. 63, 151–164 (2017).
Kovalenko, V., Bate, L. & Six, D. Can climate variability and landscape position predict white pine blister rust incidence, mortality due to the disease, and regeneration in whitebark pine? Can. J. For. Res. https://doi.org/10.1139/cjfr-2023-0270 (2024).
Markham, K., Frazier, A. E., Singh, K. K. & Madden, M. A review of methods for scaling remotely sensed data for spatial pattern analysis. Landsc. Ecol. 38, 619–635 (2023).
Frazier, A. E. Surface metrics: scaling relationships and downscaling behavior. Landsc. Ecol. 31, 351–363 (2016).
Crookston, N. L. Plant Species and Climate Profile Predictions. http://charcoal.cnre.vt.edu/climate/species/ (2012).
Monahan, W. B., Cook, T., Melton, F., Connor, J. & Bobowski, B. Forecasting distributional responses of limber pine to climate change at management-relevant scales in Rocky Mountain National Park. PLoS One 8, e83163 (2013).
Malone, S. L., Schoettle, A. W. & Coop, J. D. The future of subalpine forests in the southern Rocky Mountains: Trajectories for Pinus aristata genetic lineages. PLoS One 13, e0193481 (2018).
Shirk, A. J. et al. Southwestern white pine (Pinus strobiformis) species distribution models project a large range shift and contraction due to regional climatic changes. For. Ecol. Manage. 411, 176–186 (2018).
Fryer, J. L. Tree species distribution maps from Little's‘ Atlas of United States trees’ series. USDA Forest Service (2018).
Keane, R. E., Holsinger, L. M., Mahalovich, M. F. & Tomback, D. F. Restoring Whitebark Pine Ecosystems in the Face of Climate Change. https://research.fs.usda.gov/treesearch/download/54577.pdf (2017).
Whitebark Pine Ecosystem Foundation. Whitebark pine and limber pine range maps (2014).
Abatzoglou, J. T. Development of gridded surface meteorological data for ecological applications and modelling. Int. J. Climatol. 33, 121–131 (2013).
Malone, S. L. et al. RUST MAPPER: White Pine Blister Rust Risk in the Western United States ver 2. Environmental Data Initiative https://doi.org/10.6073/pasta/ae3ef7402b347a7b9ec92bbf8b314685 (2025).
Keane, R. E., Schoettle, A. W. & Tomback, D. F. Effective actions for managing resilient high elevation five-needle white pine forests in western North America at multiple scales under changing climates. For. Ecol. Manage. 505, 119939 (2022).
Burns, K. S. White Pine Blister Rust in the Sangre de Cristo and Wet Mountains of Southern Colorado (2006).
Burns, K. S. & Dudley, M. M. Health and Abundance of Five-Needle Pines in Southern Colorado and Northern New Mexico.
Burns, K.S., Schoettle, A.W. Evaluation of a White Pine Blister Rust Infestation in Rocky Mountain National Park and Recommendations for FY190 (2018).
Greater Yellowstone Coordinating Committee. Interagency whitebark pine monitoring protocol for the greater Yellowstone ecosystem (2011).
Dudney, J. C. et al. Compounding effects of white pine blister rust, mountain pine beetle, and fire threaten four white pine species. Ecosphere 11, e03263 (2020).
Conklin, D. A. Development of the white pine blister rust outbreak in New Mexico (2004).
Jackson, M. Protection Report 03-4 Update on Limber Pine Decline and Mortality on the Lewis and Clark National Forest, Montana. https://www.cabidigitallibrary.org/doi/full/10.5555/20033123859 (2003).
Jackson, M., Gannon, A., Kearns, H., Kendall, K. Current Status of Limber Pine in Montana (2010).
Jackson, J. I. et al. Whitebark pine in Crater Lake and Lassen Volcanic National Parks: Assessment of stand structure and condition in a management and conservation perspective. Forests 10, 834 (2019).
Jules, E. S., van Mantgem, P. J., Iberle, B. G., Nesmith, J. C. B. & Rochefort, R. M. Whitebark pine in the National Parks of the pacific states: An assessment of population vulnerability. Northwest Science 94, 129–141 (2020).
Schoettle, A. W. & Coop, J. D. Range-wide conservation of Pinus aristata: a genetic collection with ecological context for proactive management today and resources for tomorrow. New Forests 48, 181–199 (2017).
Keane, R. E., Morgan, P. & Menakis, J. P. Landscape assessment of the decline of whitebark pine (Pinus albicaulis) in the Bob Marshall Wilderness Complex, Montana, USA. Northwest Science 68, 213–229 (1994).
Kegley, S., Schwandt, J., Gibson, K. & Perkins, D. Health of whitebark pine forests after mountain pine beetle outbreaks 63, 85–93 (2011).
Kegley, S., Schwandt, J. W. & Gibson, K. E. Forest Health Assessment of Whitebark Pine on Pyramid Pass, Russell Mountain, and Burton Ridge in the Selkirk Mountains on the Idaho Panhandle National Forests. https://ir.library.oregonstate.edu/downloads/8w32r684q (2001).
Kendall, K. C. Internal plot data from northern Rocky Mountain Science Center, USGS, Glacier National Park, West Glacier, MT. (1997).
Klutsch, J., Goodrich, B. & Schoettle, A. Limber pine forests on the leading edge of white pine blister rust distribution in northern Colorado. in Proceedings of the High Five Symposium (eds. E. R., Tomback, D. F., Murray, M. P. & Smith, C. M.) 222–225 (US Department of Agriculture, Forest Service, Rocky Mountain Research Station, 2011).
Lockman, I. B., Denitto, G. A., Courter, A. & Koski, R. WLIS: The white bark limber pine information system and what it can do for you. in Proceedings of the conference whitebark pine: a Pacific Coast perspective 146–147 (Ashland, OR, 2007).
Nesmith, J. C. B., Wright, M., Jules, E. S. & McKinney, S. T. Whitebark and foxtail pine in Yosemite, Sequoia, and Kings Canyon National Parks: Initial assessment of stand structure and condition. Forests 10, 35 (2019).
Newcomb, M. White pine blister rust whitebark pine and Ribes species in the greater Yellowstone area. (The University of Montana, 2003).
Shanahan, E., Legg, K. & Daley, R. Status of Whitebark Pine in the Greater Yellowstone Ecosystem a Step-Trend Analysis with Comparisons from 2004 to 2015 (2017).
Shoal, R. Z. & Aubry, C. A. Assessment of Whitebark Pine Health on Eight National Forests in Oregon and Washington (2006).
Simons, R. & Cluck, D. Whitebark Pine Monitoring Plot Protocol for the Warner Mountains, Modoc National Forest (2010).
Smith, J. P., Hoffman, J. T. & Hoffmann, J. T. Site and stand characteristics related to white pine blister rust in high-elevation forests of southern Idaho and western Wyoming. West. N. Am. Nat. 61, 409–416 (2001).
Smith, J. P. & Hoffman, J. T. Status of white pine blister rust in the Intermountain West. West. N. Am. Nat. 60, 165–179 (2000).
Looney, C. E., Waring, K. & Fairweather, M. Chapter 12: Monitoring the health of Pinus strobiformis: Early impacts of White Pine blister rust invasion (project INT-EM-B-10-03). Forest health monitoring: national status 207, 167–175 (2019).
Looney, C. E. & Waring, K. M. Patterns of forest structure, competition and regeneration in southwestern white pine (Pinus strobiformis) forests. For. Ecol. Manage. 286, 159–170 (2012).
Goodrich, B. A., Waring, K. M., Auty, D. & Sánchez Meador, A. J. Interactions of management and white pine blister rust on Pinus strobiformis regeneration abundance in southwestern USA. Forestry 91, 492–505 (2018).
Zeglen, S. Whitebark pine and white pine blister rust in British Columbia, Canada. Can. J. For. Res. 32, 1265–1274 (2002).
Vogler, D. R., Maloney, P. E., Burt, T. & Snelling, J. First Report of the White Pine Blister Rust Fungus, Cronartium ribicola, Infecting Pinus flexilis on Pine Mountain, Humboldt National Forest, Elko County. Plant Dis (2017).
D F. Tomback, R E. Keane, Cyndi Smith Methods for Surveying and Monitoring Whitebark Pine for Blister Rust Infection and Damage. http://whitebarkfound.org/wp-content/uploads/2013/10/Methods-for-Surveying-and-Monitoring-Whitebark-Pine-for-Blister-Rustx.pdf (2024).
Oblinger, B. Incidence and Severity of Limber Pine Dwarf Mistletoe (Arceuthobium Cyanocarpum) on Whitebark Pine (Pinus Albicaulis) at Newberry Crater (2017).
Fairweather, M. L. & Geils, B. W. First report of the white pine blister rust pathogen, Cronartium ribicola, in Arizona. Plant Dis. 95, 494 (2011).
Cleaver, C. M., Jacobi, W. R., Burns, K. S. & Means, R. E. Limber pine in the central and southern Rocky Mountains: Stand conditions and interactions with blister rust, mistletoe, and bark beetles. For. Ecol. Manage. 358, 139–153 (2015).
Kearns, H. S. J. & Jacobi, W. R. The distribution and incidence of white pine blister rust in Central and Southeastern Wyoming and Northern Colorado. Can. J. For. Res. 37, 462–472 (2007).
Maloney, P. E. Incidence and distribution of white pine blister rust in the high-elevation forests of California. For. Pathol. 41, 308–316 (2011).
Maloney, P. E., Vogler, D. R., Jensen, C. E. & Delfino Mix, A. Ecology of whitebark pine populations in relation to white pine blister rust infection in subalpine forests of the Lake Tahoe Basin, USA: Implications for restoration. For. Ecol. Manage. 280, 166–175 (2012).
De Reu, J. et al. Application of the topographic position index to heterogeneous landscapes. Geomorphology 186, 39–49 (2013).
Acknowledgements
This work was supported by funding from the USDA Forest Service through a Western Wildland Environmental Threat Assessment Center (WWETAC) grant under agreement 19-JV-11221633-129. The contributions of U.S. Government employees to this research were partially supported by their respective agencies. The Yale School for the Environment and the Yale Center for Natural Carbon Capture provided additional support. The authors extend their gratitude to Anthony Culpepper of the Mountain Studies Institute for his assistance with GIS analysis and to Angel Chen for her efforts in publishing the final data products. The findings and conclusions in this report are those of the authors and should not be construed to represent any official USDA or U.S. Government determination or policy.
Author information
Authors and Affiliations
Contributions
S.L.M. and A.W.S. conceived the project idea. A.W.S. and K.S.B. harmonized data from various independent studies, while S.L.M. collected auxiliary data, developed models, and maintained the code. A.W.S. reviewed analyses and collaborated with S.L.M. on data visualization. Expert reviews were provided by K.S.B., H.S.J.K., J.E.S., M.N. and C.M.C. The manuscript was written, reviewed, and edited collectively by S.L.M., A.W.S., K.S.B., H.S.J.K., J.E.S., M.N. and C.M.C.
Corresponding author
Ethics declarations
Competing interests
The authors declare no competing interests.
Additional information
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.
About this article
Cite this article
Malone, S.L., Schoettle, A.W., Burns, K.S. et al. RustMapper: White Pine Blister Rust Risk Across High Elevation Forests in the Western United States. Sci Data 12, 1046 (2025). https://doi.org/10.1038/s41597-025-05382-1
Received:
Accepted:
Published:
Version of record:
DOI: https://doi.org/10.1038/s41597-025-05382-1
