Quantifying regional trends in large live tree and snag availability in support of forest management

doi:10.1016/j.foreco.2020.118554

Forest Ecology and Management

Volume 479, 1 January 2021, 118554

https://doi.org/10.1016/j.foreco.2020.118554 Get rights and content

Highlights

•
Satellite-based maps support monitoring of large tree and snag distributions.
•
Oregon and Washington have lost many forests supporting large trees and snags.
•
Recent increases have occurred primarily on public lands.

Abstract

In the Pacific Northwest of the United States, 20th century timber harvesting resulted in major declines in area of forests supporting large live and dead trees (i.e., snags), that are not only key habitat elements for many wildlife species but also critical components of ecosystem function. Regional forest management guidance, such as the Northwest Forest Plan (1994) and Eastside Screens (1995), may aim to conserve and foster the development of late-successional old-growth forests, characterized by large live and dead trees. Satellite remote sensing supports regional monitoring efforts of these habitat characteristics, but managers may require additional guidance in order to leverage these data for large landscape or regional assessments. In this study, our objectives were to assess long-term (historical vs. contemporary) and short-term (1993–2017) changes in lands supporting large live trees and snags across 10 wildlife habitat types (WHTs) – disparate vegetation conditions that support significantly different wildlife communities – in Oregon and Washington, USA. We generated 30-m, annual maps of large live trees (>50 cm, >75 cm, and >100 cm diameter) and snags (>25 cm and > 50 cm diameter) based on the gradient nearest neighbor (GNN) imputation method. GNN integrates Landsat satellite imagery, geospatial climatic and topographic data, and USDA Forest Service Forest Inventory and Analysis data to predict forest attributes for all forested lands in the study area. GNN classification accuracy was poor to good for large live trees (Cohen’s kappa = 0.2 – 0.6) and poor to fair for snags (Cohen’s kappa = 0.1 – 0.3) in most WHTs, though performance was substantially lower in drier WHTs where large live trees and snags were rare. Our results highlighted long-term reductions in forest supporting large live trees and snags from historical to contemporary times, especially in wetter, more productive WHTs. In contrast, we observed short-term (1993–2017) increases in areas supporting large live trees and snags. Federal forests were both more similar to reference conditions and exhibited greater recent increases in areas supporting large live trees compared to nonfederal lands. Thus, Oregon and Washington have lost a substantial proportion of forests containing large live trees and snags and recent recruitment of these trees at regional scales is a slow process primarily occurring on federal lands. However, detecting such changes through current Landsat satellite mapping technologies remains challenging, highlighting the need for new mapping methods to aid in future management.

Introduction

The presence of large live trees and standing dead trees, or snags, is a defining characteristic of old-growth forest ecosystems in western North America (Franklin et al., 1981, Franklin et al., 2002, Kaufmann et al., 2007, Lindenmayer et al., 2012, Reilly and Spies, 2015). They provide structural elements supporting high quality habitat for many wildlife species (Hunter and Bond 2001). Since the mid-20th century, anthropogenic stressors, such as timber harvesting, land conversion, and wildfire, have greatly reduced the extent of old-growth forests in Oregon and Washington (Bolsinger and Waddell 1993). Many wildlife species in this region rely entirely or in part on the availability of large live and dead trees (Ohmann et al. 1994) and monitoring their availability over short (i.e., decades) and long (i.e., centuries) time-scales may highlight trends in the capacity of landscapes to support them. Both the changes in late successional and old-growth abundance and their importance in supporting ecosystem function motivated regional management plans contributing to the conservation of these forests, such as the Northwest Forest Plan implemented in 1994 (Spies et al. 2019) and the Eastside Screens implemented in 1995 (Steen-Adams et al. 2017). Therefore, tools that integrate reference conditions with contemporary vegetation patterns and trends increasingly form the basis for regional and sub-regional monitoring programs (Davis et al. 2016) and support decision making associated with, for example, the management of forests for wildlife and biodiversity (Marcot et al. 2010).

Short- and long-term landscape changes are not expected to be spatially uniform, with variation within the region in the divergence from historical conditions and/or differing rates of change during recent decades. There can be substantial regional variation in the abundance of large live trees and snags, with environmental conditions contributing to these patterns. Local climatic and topographic factors constrain, for example, site productivity (Latta et al. 2010) and maximum tree height (Ryan and Yoder, 1997, Koch et al., 2004, Swetnam et al., 2015). Tree mortality rates are constrained by environmental gradients, with higher mortality rates in high productivity forests (Franklin et al., 1987, Reilly and Spies, 2016), resulting in greater abundance of snags. Broad-scale patterns in the frequency and severity of disturbance regimes impacts distributions of large live trees and snags by modifying successional trajectories (Reilly and Spies 2015) and contributing structural legacies to regrown stands.

Succession and management also affect snag densities, with a greater frequency in older, unmanaged stands or for several years following moderate or high severity wildfire. Recent (i.e., following the Northwest Forest Plan [1994] and Eastside Screens [1995]) management of federal forests in Oregon and Washington have emphasized promotion of late succession and old-forest attributes, including large snags (e.g.; Davis et al. 2015). Other management goals can conflict with the aim to increase snag densities, such as thinning to accelerate old forest conditions in moister forests (northwestern OR, western WA), and fuel reduction treatments in drier forests (southwestern OR, eastern OR and WA).

Integration of remote sensing, forest inventory, and spatially-distributed environmental data through multivariate analysis provides a method for assessing temporal trends in large live tree and snag abundance at relevant spatial scales (i.e., landscape/watershed scales). In Oregon and Washington, gradient nearest neighbor (GNN) imputation mapping – a method for integrating Landsat multispectral imagery with forest inventory data to produce wall-to-wall 30-m maps of forest attributes (Ohmann and Gregory 2002) – has become a standard tool for quantifying landscape pattern and change in forest ecosystems, especially as it relates to Northwest Forest Plan monitoring for both old-growth forest structure (Davis et al. 2015) and wildlife habitat (e.g., Davis et al. 2016). Uncertainty in pixel-level predictions of snag densities generated using GNN can be high and vary considerably across forest regions (Bell et al. 2015). Regional assessments of area within differing vegetation categories appear to be comparable to sample-based estimates for the same areas (Pierce et al., 2009, Ohmann et al., 2014). However, explicit assessment of bias and precision in estimates have been lacking. Consequently, map uncertainty impacts the degree to which managers can leverage remotely-sensed vegetation data for landscape to regional assessments of wildlife habitat.

Our main objective was to explore trends in the forestland area supporting large live trees and snags in the various wildlife habitat types distributed across Oregon and Washington, USA. Wildlife habitat types (WHTs; Table 1) represent macrohabitat groupings of vegetation communities by geographic distribution, physical setting, landscape setting, structure, and composition in Oregon and Washington shown to have similar patterns of use by breeding species of wildlife (O’Neil and Johnson 2001). Some wildlife habitat types are further divided into geographic subregions to take into account variation in dead wood amounts (Marcot et al. 2010) and wildlife responses to dead wood habitat features (J. Ohmann, personal communication; Mellen-McLean et al. 2017). As opposed to other classification schemes that rely solely on abiotic factors, such as physiographic region, WHTs were used to summarize our results because they account for current forest structure and composition and are thus more relevant to wildlife managers in the study region. We assessed long-term changes in large live tree and snag availability by comparing predicted historical reference conditions (i.e., properties of undisturbed ecosystems available for direct evaluation of natural ecosystem structure, composition, and function, similar to those found prior to Euro-American settlement) (Kaufmann et al., 1994, Kaufmann et al., 2007) vs. contemporary forest area supporting large live trees and snags. We assessed short-term trends in large live tree and snag availability by examining interannual changes in mapped forest area supporting large live trees and snags following the initiation of the Northwest Forest Plan (1993) and Eastside Screens (1995).

Section snippets

Study region

Our study area included all forestlands, regardless of ownership, in Oregon and Washington (Fig. 1). This region exhibits substantial variation in climate, topography, and ownerships (public and private), as well as various histories of disturbance, forest conditions, and departure from historical conditions (Spies et al. 2018). Forests range from wet temperate rainforests that rarely burned historically, but tended to be larger with mixed severity, to semi-arid woodlands that burned at high

Results

Empirical density functions for the relative frequency of lands with differing densities of large live trees and snags varied by WHT (Fig. 2). As the diameter threshold for large live trees and snags increased, the frequency of reference plots with no such trees increased and the maximum densities decreased, reflecting increasing scarcity of large live trees and snags. Both the relative frequency with no large trees increased and maximum densities decreased as one shifted from wetter western

Discussion

Contemporary patterns of large live tree and snag distributions implied an important role for regional patterns of 19th and 20th century disturbance legacies, including widespread timber harvesting and historic wildfires, as well as contemporary land management. Forests in the wetter and warmer portion of the study region (WLCH_OCA, WLCH_WCA, WLCH_OCO, WLCH_OCO, and SWOMC; Table 1, Fig. 1), generally contained 20% to 60% of the forested area supporting large live trees and snags compared to

Conclusions

In part, the maintenance of forest biodiversity depends on the retention of existing and restoration of late-successional and older forest habitat for wildlife species to within natural or desired ranges. Our results indicate that the forests of Oregon and Washington have lost a substantial proportion of forests supporting large live trees and snags compared to reference conditions and that recruitment of these structural elements of wildlife habitat at regional scales is a slow process. Slow

CRediT authorship contribution statement

David M. Bell: Conceptualization, Methodology, Formal analysis, Writing - original draft, Writing - review & editing, Visualization, Supervision, Project administration. Steven A. Acker: Conceptualization, Methodology, Writing - review & editing. Matthew J. Gregory: Conceptualization, Methodology, Formal analysis, Writing - review & editing, Visualization. Raymond J. Davis: Conceptualization, Writing - review & editing. Barbara A. Garcia: Conceptualization, Writing - review & editing.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgements

This research would not be possible without data collection by the USDA Forest Service Forest Inventory and Analysis program. This research was supported by Oregon Department of Forestry, California Natural Resources Agency (CCCA4-CNRA-2018-014), and the USDA Forest Service Pacific Northwest Research Station (PNW-JV-11261959-040).

References (55)

A.M. Barry et al.
Long-term dynamics and characteristics of snags created for wildlife habitat
For. Ecol. Manage.
(2017)
D.M. Bell et al.
Imputed forest structure uncertainty varies across elevational and longitudinal gradients in the western Cascade Mountains, Oregon, USA
For. Ecol. Manage.
(2015)
W.B. Cohen et al.
A LandTrendr multispectral ensemble for forest disturbance detection
Remote Sens. Environ.
(2018)
W.B. Cohen et al.
Detecting trends in forest disturbance and recovery using yearly Landsat time series: 2. TimeSync — Tools for calibration and validation
Remote Sens. Environ.
(2010)
J.F. Franklin et al.
Disturbances and structural development of natural forest ecosystems with silvicultural implications, using Douglas-fir forests as an example
For. Ecol. Manage.
(2002)
R. Haugo et al.
A new approach to evaluate forest stucture restoration needs across Oregon and Washington, USA
For. Ecol. Manage.
(2015)
G. Latta et al.
Analysis of potential impacts of climate change on forests of the United States Pacific Northwest
For. Ecol. Manage.
(2010)
J.L. Ohmann et al.
Scale considerations for integrating forest inventory plot data and satellite image data for regional forest mapping
Remote Sens. Environ.
(2014)
J.L. Ohmann et al.
Mapping change of older forest with nearest-neighbor imputation and Landsat time-series
For. Ecol. Manage.
(2012)
P. Olofsson et al.
Good practices for estimating area and assessing accuracy of land change
Remote Sens. Environ.
(2014)

P. Olofsson et al.

Making better use of accuracy data in land change studies: Estimating accuracy and area and quantifying uncertainty using stratified estimation

Remote Sens. Environ.

(2013)

M.J. Reilly et al.

Disturbance, tree mortality, and implications for contemporary regional forest change in the Pacific Northwest

For. Ecol. Manage.

(2016)

S.V. Stehman

Estimating area from an accuracy assessment error matrix

Remote Sens. Environ.

(2013)

S.V. Stehman et al.

Pixels, blocks of pixels, and polygons: choosing a spatial unit for thematic accuracy assessment

Remote Sens. Environ.

(2011)

H.S.J. Zald et al.

Influence of lidar, Landsat imagery, disturbance history, plot location accuracy, and plot size on accuracy of imputation maps of forest composition and structure

Remote Sens. Environ.

(2014)

Battles, J.J., Bell, D.M., Kennedy, R.E., Saah, D.S., Collins, B.M., R.A. York, Sanders, J.E., Lopez-Ornelas, F., 2018....

Bechtold, W.A., Patterson, P.L., 2005. The Enhanced Forest Inventory and Analysis Program — National Sampling Design...

Bolsinger, C. L., and K. L. Waddell. 1993. Area of old-growth forests in California, Oregon, and Washington. Resource...

C.J.F. Ter Braak

Canonical correspondence analysis: A new eigenvector technique for multivariate direct gradient analysis

Ecology

(1986)

L. Breiman

Random forests

Mach. Learn.

(2001)

E.P. Crist et al.

A physically-based transformation of Thematic Mapper data –- The TM Tasseled Cap

IEEE Trans. Geosci. Remote Sens.

(1984)

C. Daly et al.

Physiographically sensitive mapping of climatological temperature and precipitation across the conterminous United States

Int. J. Climatol.

(2008)

Davis, R.J., Hollen, B., Hobson, J., Gower, J.E., Keenum, D., 2016. Status and trends of northern spotted owl habitats....

Davis, R.J., Ohmann, J.L., Kennedy, R.E., Cohen, W.B., Gregory, M.J., Yang, Z., Roberts, H.M., Gray, A.N., Spies, T.A.,...

M.S. Delheimer et al.

If a tree falls in a forest: implications of forest structure persistence for the Pacific marten (Martes caurina)

Ecosphere

(2019)

T. DeMeo et al.

Expanding Our Understanding of Forest Structural Restoration Needs in the Pacific Northwest

Northwest Sci.

(2018)

A.B. Edworthy et al.

Tree cavity occupancy by nesting vertebrates across cavity age

J. Wildl. Manage.

(2018)

Cited by (21)

Prioritizing new conservation areas during forest plan updates
2024, Forest Ecology and Management
This paper presents a novel spatial analysis process designed to prioritize new conservation areas in anticipation of upcoming updates to the Northwest Forest Plan (NWFP) for national forests in the western United States. Illustrated through a case study of the Gifford Pinchot National Forest, the approach seeks to identify locations for conversion from matrix to late-successional reserve (LSR) land use allocation, with the goal of enhancing climate resilience, improving carbon storage, and safeguarding mature and old-growth forests. Matrix lands are areas in national forests where timber production and harvest are primary management objectives. Contrary to common assumption, there are many mature and old-growth stands currently located within matrix designation. To prioritize locations for consideration for a matrix to LSR transfer, I used connectivity model outputs (depicting densely clustered mature forests and corridors), forest age data (identifying forests estimated to be over 100 years old), and land management allocation data to rank locations as primary and secondary areas for protection. Working from areas with dense aggregations of these ranks, I then included additional variables—including carbon storage metrics, a secondary age layer identifying forests over 200 years old, locations of recent or current timber harvest projects, and a westside-eastside categorization—to delineate final areas for conservation. This method offers a replicable framework for other national forests and provides a science-based process to guide conservation efforts, inform policy recommendations, and ensure the long-term well-being of mature and old-growth forests and their associated ecosystems.
Timber harvest and wildfires drive long-term habitat dynamics for an arboreal rodent
2023, Biological Conservation
Animals dependent on old, closed-canopy forests may be particularly sensitive to forest disturbances, such as timber harvest and high-severity fires, that may take centuries to recover from. Tracking habitat trends can provide broader context to conservation decisions, including the role of forest disturbance type. We assessed habitat dynamics across the range of the red tree vole (Arborimus longicaudus), a canopy-obligate rodent with low-mobility that typically occurs in old forests. We developed and applied a red tree vole habitat model to imagery data each year across 37 years (1986–2022). We quantified changes in habitat across four regions that differed by land ownership and dominant disturbance type. Overall, habitat declined 18 % since implementation of a major conservation initiative (1994–2022). Habitat change was highest in the northern coast (65 %) where timber harvest was the most common disturbance. Within interior regions, several large wildfires in the previous two decades correlated with a 15 % decline of habitat on federal lands, particularly at the northern range periphery where tree voles were already scarce due to historic disturbances. Recruitment of old forest and subsequent recolonization by red tree voles and other low mobility species after high-severity disturbance can take centuries whereas old forest loss can happen rapidly. Given their strong association with old forest, scarcity within historic disturbance footprints, and relatively limited geographic range, we anticipate red tree voles will remain a species of conservation concern.
Trends in forest structure restoration need over three decades with increasing wildfire activity in the interior Pacific Northwest US
2023, Forest Ecology and Management
Citation Excerpt :
Within the study area, there were 10 biophysical settings (Table 2) and 365 strata (i.e., unique combinations of biophysical setting and landscape level). Current forest structure conditions were mapped annually from 1986 to 2017 using gradient nearest neighbor (GNN) datasets from the Landscape Ecology, Modeling, Mapping and Analysis (LEMMA) lab at Oregon State University (https://lemma.forestry.oregonstate.edu/data) (Bell et al., 2021). Three GNN-derived forest structure attributes (canopy cover, trees per acre, and quadratic mean diameter per diameter class; Table A.1) were used to develop rule-based criteria with which each 30-meter resolution pixel was classified as one of 5 structural/successional classes—early development, mid-development closed canopy, mid-development open canopy, late-development open canopy, and late-development closed canopy.
Wildfire is a keystone ecological process in many forests worldwide, but fire exclusion and suppression have driven profound shifts in forest structure (e.g., increased density, canopy cover, biomass) that have contributed to increases in large, high-severity fire in many seasonally dry forests and woodlands of the western United States. Comparisons between contemporary and historic range of variability (HRV) in forest structure can quantify the amount and types of restoration that shift landscapes toward structural conditions that have historically fostered resilience to fire. However, landscapes are dynamic over time and conditions reflect the net effects of planned actions (e.g., fuel reduction treatments) and unplanned actions (e.g., wildfire). How wildfire activity may shift landscapes toward or away from the HRV and correspondingly affect the need for restoration, has not been widely tested. Here, we quantify long-term (1986–2017), and continuous (annual resolution) trends of forest restoration need and ask how wildfire activity during this period has affected restoration need trends at three nested spatial extents: across the eastern Washington (USA) ecoregion, among fire regimes within this ecoregion, and between watersheds of a frequent/low-severity fire regime that experienced contrasting amounts of wildfire. At the broadest scale, restoration need did not change substantially during the study period, with approximately 35 % of forest area in need of disturbance restoration—despite 16.6 % of the total forested area (593,000 ha) burning from 1986 to 2016. At intermediate spatial extents (among fire regimes), forests characterized by historically frequent/low-severity fire experienced the greatest decrease in disturbance restoration need following recent fire activity. Although > 50 % of forests within this fire-regime remained in need of disturbance restoration at the end of the study period, we found a strong correspondence between forested area burned and decreased restoration need in this fire regime; relationships were equivocal or non-existent in other fire regimes. At the finest spatial scale (watersheds dominated by historically frequent/low-severity fire-regime forests), we found sharp contrasts between areas that experienced high fire activity in recent years and those that did not. At this scale, recent large fires have decreased disturbance restoration need by > 25 %. Our findings suggest that recent large wildfires have reduced the amount of forest in need of restoration, but have done so modestly and primarily at local or sub-regional extents. Overall, our approach can be applied to understanding how wildfires or other disturbances contribute to affecting forest structure and management targets in other ecosystems through time and space.
Uncovering forest dynamics using historical forest inventory data and Landsat time series
2022, Forest Ecology and Management
Citation Excerpt :
GNN is a well-known algorithm in the family of nearest neighbors techniques that relies on canonical correspondence analysis (CCA) (ter Braak, 1986). It has been widely used across the Pacific Northwest (Bell et al., 2015, 2021; Kennedy, Ohmann, et al., 2018) and in other parts of the United States (Wilson et al., 2012) to simultaneously map distributions of various forest attributes. For the GEE implementation, we used a new library developed by the authors (GEE-KNN) which automates many of the steps in nearest neighbor techniques and allows comparative evaluation of different algorithms.
While many countries regularly produce timely and accurate estimates for a range of forest attributes through national forest inventory programs, many have access to no such inventory or their inventory is discontinuous in space and time, such as a sample-based approach in Ukraine limited to two regional (Ivano-Frankivsk and Sumy) inventories conducted in 2008–2015 at a total area of 37,700 ha. This study addresses the extent to which limited historical forest inventory data may facilitate forest mapping efforts using dense Landsat time series (LTS) and mapping techniques (i.e., classification and imputation) to predict multiyear forest dynamics. We used the Continuous Change Detection and Classification (CCDC) segmentation approach to extract inter- and intra-annual trends in LTS and utilized these fitted trends as covariates in further modeling. We developed random forest (RF) classification models for forest mapping based on field data collection and LTS, and then generated yearly forest maps for 1990–2020. The RF model accuracies were high for both producer’s and user’s accuracy (>0.93 ± 0.04). Based on yearly forest dynamics, we detected an increasing trend in forest loss during 2000–2010 and 2010–2020 in our regions. We used the gradient nearest neighbor (GNN) method to generate species presence (basal area (BA) ≥ 1.0 m² ha⁻¹) and growing stock volume (GSV) maps at the 30-m pixel level. Our nearest neighbor imputation models achieved better accuracies (Cohen’s kappa > 0.4) for species that are prevalent within regions, occupy distinct geographical areas, and have higher plot BA abundance. To obtain estimates of species abundance with higher R², we used the GNN model with k = 3 nearest neighbors. Aggregating species based on similar environmental niche resulted in greater R² values of BA that varied within two regions from low (0.129 and 0.416) for a group of hardwood deciduous species to high (0.531 and 0.712) for coniferous species. The accuracy was systematically better at 5-km hexagonal level, thus we recommend using the maps at coarser aggregations. The study showed that forests accumulated more GSV in 1990–2000, albeit higher rates of GSV loss were observed during the two last decades. We attribute the spatial and temporal character of forest change to forestry activities in the past, the current age structure of forests, and afforestation of abandoned farmlands. We are aware that forecasting performance of the nearest neighbor imputation approach merits more detailed consideration as new forest inventory data are available.
An ecological perspective on living with fire in ponderosa pine forests of Oregon and Washington: Resistance, gone but not forgotten
2021, Trees, Forests and People
Wildland fires (WLF) have become more frequent, larger, and severe with greater impacts to society and ecosystems and dramatic increases in firefighting costs. Forests throughout the range of ponderosa pine in Oregon and Washington are jeopardized by the interaction of anomalously dense forest structure, a warming and drying climate, and an expanding human population. These forests evolved with frequent interacting disturbances including low-severity surface fires, droughts, and biological disturbance agents (BDAs). Chronic low-severity disturbances were, and still are, critical to maintaining disturbance resistance, the property of an ecosystem to withstand disturbance while maintaining its structure and ecological function. Restoration of that historical resistance offers multiple social and ecological benefits.
Moving forward, we need a shared understanding of the ecology of ponderosa pine forests to appreciate how restoring resistance can reduce the impacts of disturbances. Given contemporary forest conditions, a warming climate, and growing human populations, we predict continued elevation of tree mortality from drought, BDAs, and the large high-severity WLFs that threaten lives and property as well as ecosystem functions and services. We recommend more comprehensive planning to promote greater use of prescribed fire and management of reported fires for ecological benefits, plus increased responsibility and preparedness of local agencies, communities and individual homeowners for WLF and smoke events. Ultimately, by more effectively preparing for fire in the wildland urban interface, and by increasing the resistance of ponderosa pine forests, we can greatly enhance our ability to live with fire and other disturbances.
Classification and Assessment of Riparian Ecosystems in Northwest Oregon for Restoration Planning
2023, Northwest Science

View all citing articles on Scopus

View full text

Published by Elsevier B.V.