- Email: [email protected]
Using object-based image analysis to conduct high- resolution conifer extraction at regional spatial scales
Int J Appl Earth Obs Geoinformation 73 (2018) 148–155
Contents lists available at ScienceDirect
Int J Appl Earth Obs Geoinformation journal homepage: www.elsevier.com/locate/jag
Using object-based image analysis to conduct high- resolution conifer extraction at regional spatial scales
T
⁎
K. Benjamin Gustafson, Peter S. Coates , Cali L. Roth, Michael P. Chenaille, Mark A. Ricca, Erika Sanchez-Chopitea, Michael L. Casazza Western Ecological Research Center, U.S. Geological Survey, Dixon, CA, United States
A R T I C LE I N FO
A B S T R A C T
Keywords: Object-based image analysis Pinyon-juniper Sage-grouse Sagebrush Feature extraction Image processing
Distributional expansion and infill of pinyon (Pinus monophylla) and juniper (Juniperus osteosperma, J. occidentalis) trees (hereinafter, "pinyon-juniper") into sagebrush ecosystems alters the ecological function and economic viability of these ecosystems and represents a major contemporary challenge facing land and wildlife managers. Therefore, accurate and high-resolution maps of pinyon-juniper distribution and abundance across broad geographic extents would facilitate science that quantifies ecological effects of pinyon-juniper expansion and help guide land management decisions that better target areas for pinyon-juniper treatment projects. We mapped conifers at a high (1- m2; i.e., 1 × 1-m) resolution across the majority of Nevada and northeastern California. We used digital orthophoto quad tiles from National Agriculture Imagery Program (USDA, 2013) to classify conifers using automated feature extraction (AFE) with the program Feature Analyst™ (Overwatch, 2013). Overall accuracy was > 86% across all mapped areas for ground referencing methods. We provide five sets of full-extent maps for land managers: (1) a shapefile representing accuracy results linked to mapping subunits; (2) binary rasters representing conifer presence or absence at a 1-m2 resolution; (3) a 900-m2 resolution raster representing percentages of conifer canopy cover within each cell; (4) 1-m2 resolution canopy cover classification rasters derived from a 50-m radius moving window analysis; and (5) an example map derived from our canopy cover product that prioritizes pinyon-juniper treatment by significance to sage-grouse habitat improvement. Importantly, the canopy cover maps were developed to allow user-specified flexibility based on their own objectives (i.e., develop phases of expansion). These products improve upon or complement existing conifer maps for the Western United States and will help facilitate habitat management and sagebrush ecosystem restoration through an accurate understanding of conifer distribution and abundance at multiple spatial scales.
1. Introduction 1.1. Ecological background The iconic "sagebrush sea" that characterizes the Great Basin of the Western United States provides habitat for several at-risk sagebrushobligate species (Homer et al., 2009; Knick et al., 2013) and supports outdoor recreation valued at $1 billion annually (BLM and USFS, 2015) and cattle grazing, valued as high as $60 billion in a single year (USDA NASS, 2014). It is also one of the most endangered ecosystems in the United States (Noss et al., 1995), as the range of sagebrush has contracted by more than 50% since European settlement (Schroeder et al., 2004). The remaining range is subjected to continual fragmentation, largely from anthropogenic-related disturbances (Schroeder et al., 2004). Accordingly, sagebrush ecosystems are at the center of national
⁎
conservation strategy (U.S. Department of the Interior, 2015). Conifers such as pinyon (Pinus monophylla) and juniper (Juniperus sp.; hereinafter, "pinyon-juniper") are a native component of the sagebrush ecosystems within the Great Basin. However, the distribution and abundance of pinyon-juniper has greatly expanded following European settlement (Miller and Tausch, 2001) owing to a variety of factors, including changes in climate (Soulé et al., 2004; Romme et al., 2009), land use (Romme et al., 2009), and fire regimes (Miller et al., 2000; Soulé et al., 2004). Accordingly, expansion of pinyon-juniper is a major factor contributing to loss and fragmentation of sagebrush ecosystems (Davies et al., 2011; Miller et al., 2011; Knick et al., 2013). Dominance of sagebrush and perennial grasses, which contribute strongly to sagebrush ecosystem resilience to disturbance and resistance to invasion (Chambers et al., 2014), decreases as cover of pinyon-juniper increases (Miller et al., 2005). This relationship can be categorized into three
Corresponding author. E-mail address: [email protected] (P.S. Coates).
https://doi.org/10.1016/j.jag.2018.06.002 Received 14 February 2018; Received in revised form 4 June 2018; Accepted 5 June 2018 0303-2434/ Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/BY-NC-ND/4.0/).
Int J Appl Earth Obs Geoinformation 73 (2018) 148–155
K.B. Gustafson et al.
Fig. 1. Schematic showing conceptual model of conifer classification framework using automated feature extraction methods within greater sage-grouse habitats in Nevada and California. Input is represented as orange diamonds and green hexagons identify products. Blue and gray boxes represent steps conducted in Feature Analyst™ and a GIS, respectively. Starred items (*) denote iterated geoprocessing steps.
distinct phases of woodland development, where sagebrush is dominant over pinyon-juniper in phase 1, sagebrush and pinyon-juniper are codominant in phase 2, and pinyon-juniper is dominant and has replaced sagebrush in phase 3 (Miller et al., 2005, 2008). The loss of herbaceous understory diminishes forage and cover for constituent wildlife (Miller et al., 2000, 2011) and reduces available forage for cattle (Soulé and Knapp, 1999; Twidwell et al., 2013). Pinyon-juniper expansion also reduces streamflow (Kormos et al., 2017), depletes soil water availability (Roundy et al., 2014; Kormos et al., 2017), and increases bare earth contiguity that promotes runoff and subsequent soil erosion (Davenport et al., 1998; Pierson et al., 2010). Perhaps most importantly, the spread and infill of pinyon-juniper into low-elevation sagebrush ecosystems introduces woody biomass that fuels more intense wildfires (Bradley and Fleishman, 2008; Romme et al., 2009; Weiner et al., 2016; Strand et al., 2013). Burned stands are often replaced by invasive annual grasses that have a positive feedback cycle with wildfire, thereby increasing fire extent and frequency and spreading fire into sagebrush that would otherwise be much less likely to burn (Tausch et al., 2009; Romme et al., 2009; Davies et al., 2011). Because of the multitude of economic and ecological effects from sagebrush loss, pinyon-juniper expansion is now a primary challenge facing land managers in the Western United States (Davies et al., 2011).
regional GIS products derived from satellite imagery generally do not have the spatial resolution necessary to delineate early stages of pinyon-juniper expansion (e.g., as low as 10% canopy cover) that often constitute the best candidate areas for treatment (Miller et al., 2008; Baruch-Mordo et al., 2013; Coates et al., 2017). Additionally, mapping land cover at a 900-m2 resolution can overestimate pinyon-juniper density because the cell size of the imagery often exceeds the diameter of individual trees on the landscape. Feature recognition technologies such as object-based image analysis (OBIA) can be applied to very high spatial resolution (VHSR) imagery (< 4-m2) to identify individual trees across regional extents (Davies et al., 2010; Falkowski and Evans, 2012; Mishra and Crews, 2014; Falkowski et al., 2017). OBIA exploits VHSR imagery by segmenting cells into image-objects based on their spectral, spatial, and structural properties, and then classifies these image-objects to extract features of interest (Burnett and Blaschke, 2003; Hay et al., 2003; Hay and Castilla, 2006). Segmentation facilitates accurate tree delineation from VHSR imagery as conifer canopies encompass multiple cells with variable properties (Falkowski et al., 2017). Conifer classification using OBIA on VHSR imagery has been shown to reliably identify forest canopy in relatively small test scenarios (Davies et al., 2010; Falkowski and Evans, 2012; Hulet et al., 2014; Poznanovic et al., 2014; Roundy, 2015).
1.2. Need for high-resolution remote sensing
1.3. Objectives
Pinyon-juniper expansion has regional consequences, requiring cooperative efforts by Federal and State agencies that total millions of dollars (Sanford et al., 2017) and form a principal component of current wildlife conservation plans (USFWS, 2013; Miller et al., 2014; USDA, 2014; Miller et al., 2017; Severson et al., 2017). As agencies implement restoration efforts over thousands of hectares (Miller et al., 2017), there is an immediate need for high-resolution pinyon-juniper data over regional spatial extents to optimize targeted treatment efforts (Connelly et al., 2004; Homer et al., 2009; Falkowski et al., 2017). Customary
We used OBIA to create 1-m2 resolution binary conifer rasters across our study extent. We assessed mapping accuracy by analyzing errors of omission and commission using both reference imagery and ground referencing. We then scaled the 1-m2 resolution data into unsmoothed (900-m2) and 50-m radius moving-window smoothed (1-m2) estimates of percent canopy cover. The 900-m2 product allows for seamless integration into current geospatial applications that use standard 900-m2 resolution products, while the 1-m2 product allows users to estimate conifer cover into more customizable bins than currently available. To 149
Int J Appl Earth Obs Geoinformation 73 (2018) 148–155
K.B. Gustafson et al.
grouse avoid canopy cover as low as 4% (Baruch-Mordo et al., 2013; Coates et al., 2017), therefore we maintained individual intervals for each percentage < 10%. For conservation purposes, we developed a map to help guide decisions related to pinyon-juniper treatment, specifically related to sage-grouse conservation within our study extent. The map consisted of phase I expansion (< 10%) intersected with documented sage-grouse habitat management categories (Coates et al., 2016c). These categories were a result of combining spatially explicit habitat quality maps based on seasonal resource selection function with sage-grouse abundance and space use data (Coates et al., 2016c). We defined categories of pinyon-juniper management for sage-grouse habitat restoration as phase I pinyon-juniper within: 1) Priority – high quality seasonal habitats that are often occupied by sage-grouse; 2) High – high quality seasonal habitats that are less likely to be occupied by sage-grouse, as well as transitional areas comprised of non-habitat that are temporarily occupied (between seasons); 3) Moderate –moderate to low quality seasonal habitats that are not typically occupied; and 4) Low – non-habitat and non-occupied.
illustrate the utility of our products for sagebrush ecosystem management efforts, we derived a map that prioritizes areas for pinyon-juniper treatment aimed at restoring sage-grouse habitat. 2. Methods 2.1. Feature extraction We conducted conifer mapping using Automated Feature Extraction (AFE) in the program Feature AnalystTM (Overwatch, 2013) for all 61 Nevada Department of Wildlife sage-grouse Population Management Units (PMU), which corresponded to sage-grouse habitat mapping efforts (Coates et al., 2016b,c). We focus on PMUs because sage-grouse are an important indicator-species for the health of the sagebrush ecosystems and a major impetus for sagebrush restoration in the Great Basin, as the species has been considered for listing multiple times under the Endangered Species Act of 1973 (USFWS, 2010, 2015). Our study extent included 6230 DOQQs from Nevada, California, Oregon, Idaho and Utah (USDA, 2013). Fig. 1 illustrates our mapping process steps, based on the methodology provided in Roth et al. (2018). We performed traditional accuracy assessment with an estimated accuracy coefficient (kappa) analysis (Roth et al., 2018), where the kappa coefficient (Khat ) represents the percent accuracy adjusted for correct classification due to random chance (Congalton and Green, 2009). Khat > 60% indicate substantial agreement between classification and truth, and those > 80% are almost perfect (Landis and Koch, 1977).
3. Results 3.1. Reference imagery accuracy assessment We summarized accuracy results across all PMUs (Table 1); more detailed reports of accuracy assessments for each PMU are provided in Roth et al. (2018). On average, conifer classification had higher incidence of errors of commission reported as conifer user’s accuracy (76%; SD = 9%; Table 1) than non-conifer user's accuracy (97%; SD = 3%; Table 1). Accordingly, errors of omission were more common for non-conifer classification (producer’s accuracy = 81%; SD = 6%; Table 1) compared to conifer classification (97%; SD = 4%; Table 1). The mean overall accuracy was 86% (SD = 4%; Table 1). Mean adjusted accuracy showed that there was substantial agreement between classes and reference imagery (Khat = 73%; SD = 9%; Table 1). The average distance between the extracted conifer feature and the nearest vegetation > 3-m in height within a 30-m radius was 3.18-m (SD = ± 8.80 m). On average, 87% of the points we classified as conifers were confirmed by ground referencing, yielding a 13% error of commission (Table 2). Error of omission was 17%, with 83% of the conifers selected in the field as omission points (n = 450) correctly classified by our 1-m2 conifer map (Table 2). Table 2 compares the ground-referenced accuracies to the user’s and producer’s accuracies of the PMU that contains the respective site.
2.2. Ground referencing We ground referenced our results to estimate OBIA mapped conifer accuracy for on-the-ground pinyon-juniper management. Because land managers will use the conifer map to target conifers for treatment, we selectively evaluated the accuracy of the conifer class. We generated commission points from our 1-m conifer map to be ground referenced for correct classification in 8 sage-grouse study sites (Coates et al., 2016c) within the project extent. We randomly selected a maximum (up to 100) conifer locations that met three criteria: (1) in phase 1 or 2 woodland development (potential targets for treatment), (2) within 100-m of roads and (3) at slopes ≤ 30% rise for accessibility. Commission points were at least 180-m apart to avoid overlap. We recorded the coordinates and species of the nearest tree-like vegetation > 3-m in height within a 30-m search radius to account for possible compounded location error in NAIP imagery and GPS units. To measure errors of omission, we navigated to the nearest conifer that was between 60 and 90-m away from the commission point and recorded location and species. We specified a window of 60–90-m to ensure independence between commission and omission points. Error estimates were generated for commission and omission points by identifying the presence of a ground referenced or extracted conifer, respectively, within a 10-m buffer. We used a 10-m buffer, permitting for 5-m reported accuracy in both NAIP and handheld GPS units.
3.2. Conifer extraction products We generated distinct conifer mapping products for the full extent of our study area (10.5066/F7348HVC), which are: (1) confusion matrix results linked to PMU or zone; (2) binary rasters identifying conifer presence or absence at a 1-m2 resolution, by PMU; (3) a 900-m2 resolution raster representing percentages of conifer canopy cover within each cell (Fig. 1b); a (4) 50-m radius moving window canopy cover class rasters at a 1-m2 resolution, in quadrants (Fig. 1c); and (5) A map prioritizing pinyon-juniper management for sage-grouse habitat restoration efforts (Fig. 3). Importantly, the percent canopy cover products can be reclassified in a GIS into user-specified bins to meet specific objectives by land and wildlife managers (i.e., approximations for phases of pinyon-juniper expansion).
2.3. Map applications After validation, we used the final mosaics to calculate high resolution estimates of percent canopy cover for the study area in two distinct ways (Fig. 2). In the first method, we calculated percent canopy cover per 900-m2 cell by summing the number of 1-m conifer cells within a 900-m2 cell and dividing by its area, converting to whole numbers of percent canopy cover (Fig. 1f). We also calculated percent canopy cover by summing all cell values within a 50-m radius neighborhood (equivalent to 7845 m2) and dividing by the total number of cells (Fig. 1g). To facilitate use by land managers, we reclassified the cells according to multiple intervals of percent canopy cover (Supplemental Material, Table S1) based on known biological significance to sage-grouse (Baruch-Mordo et al., 2013; Coates et al., 2017). Sage-
4. Discussion 4.1. Applications We provide highly accurate maps of conifer distribution and canopy cover derived from conifer extraction from NAIP imagery at a 1-m2 resolution across the full extent of mapped sage-grouse habitat in 150
Int J Appl Earth Obs Geoinformation 73 (2018) 148–155
K.B. Gustafson et al.
Fig. 2. Products produced using automated feature extraction methods across greater sage-grouse habitat in Nevada and California depicting: a) conifer presence or absence at a 1-m2 resolution; b) continuous conifer canopy cover at 900-m2 resolution raster, and c) example of using canopy cover bins at a 1-m2 resolution produced from a 50-m radius moving window to depict progressive phases of conifer expansion.
Nevada and northeastern California, which will help facilitate conservation and restoration efforts across large spatial extents, such as sage-grouse habitat restoration within the Great Basin. We achieved high accuracy by implementing AFE in Feature Analyst™, a highly precise OBIA classification method that heretofore has had limited application at large spatial extents largely because of computational and time demands (Hay and Castilla, 2006; Bruce, 2008; Tsai et al., 2011) but can be more accurate than traditional supervised classification methods normally conducted at such extents (e.g., O’Brien, 2003; Bruce, 2008; Opitz and Blundell, 2008; Blaschke, 2010; Tsai et al., 2011). To our knowledge, we present the only AFE-based products across a significant part of the geographic distribution of sagebrush and sage-grouse, and our products complement other OBIA-based conifer maps for sagebrush ecosystems in the Western United States (e.g., Falkowski et al., 2017). Moreover, our outputs more accurately (87%) and precisely (within ∼4-m) reflect ground conditions than 900-m2 resolution Landsat- derived conifer products (Fig. 4) in areas of low to moderate canopy cover. This allows confidence for landscape-level management of sagebrush ecosystems through traditional site-level treatments, whereby managers can perform detailed, remote evaluations of treatment efforts, down to the cutting of individual trees. Therefore the combination of landscape- and site-level accuracies can help land managers compare candidate sites and optimize conifer treatment throughout the entire state (Falkowski and Evans, 2012; Falkowski et al., 2017). Our high resolution conifer maps can offer a decision support tool for land managers and also help to inform further ecological studies. For example, such maps can help refine models that predict distribution and abundance for a variety of wildlife species that respond to pinyonjuniper in the environment, including sage-grouse populations within Nevada and California (Coates et al., 2014; Coates et al., 2016a,b,c). Additionally, high resolution conifer maps are necessary to understand
Table 2 User’s and producer’s accuracies for ground-referencing sites compared to the accuracy of their respective PMU determined from reference imagery. Site
PMUs
Conifer User’s Accuracy (%)
Conifer Producer’s Accuracy (%)
Site
PMU
Site
PMU
Desatoya
Desatoya, Reese River, Toiyabe, Clan Alpline
93
86
81
98
McGinness
Toyiabe
92
73
89
99
Midway
Butte/Buck/White Pine, Diamond
94
82
84
92
Monitor
Monitor (Zone 2)
88
62
86
96
North SWIP
Schell/Antelope, Butte/Buck/White Pine
100
78
100
94
South SWIP
Butte/Buck/White Pine
80
83
91
92
Susanville
Buffalo-Skedaddle
86
71
83
92
Virginia Mountains
Virginia-Pahrah
73
78
50
98
87
Mean
83
the underlying mechanisms (Coates et al., 2017; Prochazka et al., 2017) of how conifers adversely influence sage-grouse population dynamics (Baruch-Mordo et al., 2013). Moreover, management is typically conducted within categorized bins rather than continuous scales, so managers require maps that allow for demarcation of canopy-cover classes based on the most recent quantified effects of pinyon-juniper on factors such as sage-grouse
Table 1 Summarized results of mapping conifers at the 1-m2 resolution across all population management units in Nevada and California using feature extraction methods within greater sage-grouse habitat of Nevada and California. Means are shown with + 1 standard deviation. Khat is the estimated accuracy coefficient.
Mean Minimum Maximum
Conifer user’s accuracy (%)
Non-conifer user's accuracy (%)
Conifer producer's accuracy (%)
Non-conifer producer’s accuracy (%)
Overall accuracy (%)
Khat (%)
76 ± 9 60 95
97 ± 3 90 100
97 ± 3.59 88 100
81 ± 6 71 95
86 ± 4 79 97
73 ± 9 58 94
151
Int J Appl Earth Obs Geoinformation 73 (2018) 148–155
K.B. Gustafson et al.
Fig. 3. Phase I expansion pinyon-juniper prioritized by sage-grouse management categories.
2008; Boswell et al., 2017). At large spatial extents, these high-resolution products can help land managers remotely evaluate and compare candidate sites for conifer treatment across large spatial extents (Falkowski and Evans, 2012; Falkowski et al., 2017). In regard to sage-grouse, high resolution conifer maps could be used to calculate ecological benefits of removing pinyon-juniper trees to sage-grouse populations by simulating pinyon-juniper treatment and integrating results into sage-grouse habitat and demographic models, allowing for quantification of direct improvement to habitat suitability and
population dynamics (Baruch-Mordo et al., 2013; Coates et al., 2017) and sagebrush ecosystem process (Miller et al., 2000, 2005; Tausch et al., 2009; Falkowski and Evans, 2012). Our percent canopy cover outputs (smoothed and unsmoothed) have built-in flexibility that allow users to select the cutoffs that are appropriate for their purposes. Land managers can further economically value their sites by developing scenarios for assessing how well they meet the specific goals of each project, which can then help streamline allocation of time and resources currently necessary for project appraisals (Opitz and Blundell, 152
Int J Appl Earth Obs Geoinformation 73 (2018) 148–155
K.B. Gustafson et al.
Fig. 4. Example conifer classification output overlay (in yellow; left) and percent canopy cover calculation overlay (right) from conifers mapped in the same area of Nevada using automated feature extraction.
land cover covariates used to spatially explicitly model sage-grouse habitat suitability (Coates et al., 2016b). Thus, this overarching framework supports the development of "living layers" that reflect existing sage-grouse habitat conditions and provide highly accurate, timely information to land managers responding to changing landscapes. Our results can also be used to determine accuracy of other products that cover larger extents, which might provide an indication of the reliability of products developed for management efforts beyond our study extent.
demographic rates. This process can be valuable for calculating conservation credits as ecological currency for large extent programs. Additional use of the cover class settings can refine target treatment areas to those within transitional phase I, identifying areas where trees have the greatest adverse effects on sage-grouse survival (Coates et al., 2017) and, thus, allocate funds to those sites that maximize benefits to sagegrouse (Miller et al., 2008; Baruch-Mordo et al., 2013; Ricca et al., 2018). Our map of pinyon-juniper management categories can be used to guide managers when prioritizing sites for pinyon-juniper treatment with the goal of sage-grouse habitat restoration. We combine accurate identification of phase I expansion with knowledge of composite habitat quality and space use to provide suggested areas to target for pinyonjuniper treatment. While site-level evaluation of woodland development such as historic state and woodland age cannot be addressed with our map, we provide spatially explicit direction for future selection of candidate treatment sites. These mapping products also enable non-wildlife related landscapelevel management through traditional site-level treatments and facilitates managers to prioritize and efficiently implement restoration projects, which include forage for cattle (Soulé and Knapp, 1999; Twidwell et al., 2013), changes in soil erosion patterns (Davenport et al., 1998; Pierson et al., 2010), or reductions of wildfire fuel loads, extent, and severity (Hulet et al., 2014). Similar to our development of pinyon-juniper management categories for restoring sage-grouse habitat, our products can be used to develop new maps and datasets for other management goals, such as combining canopy cover with topographic and bare earth information for soil erosion studies or determining crown density or proximity to fine fuel build up wildfire prevention. Finally, a major advantage of our framework is its replicability. The methods described in Roth et al. (2018) can be applied as is to other imagery in other parts of the western United States and the flexibility of Feature Analyst™ allows the specific parameters of the classification algorithm to be easily adapted for other land cover classes in different regions, permitting image availability. We designed our framework with the ability to include improved reference imagery as it becomes available. For example, areas of low accuracy caused by poor image quality can be readily reclassified and updated as higher resolution NAIP imagery becomes available (typically every 2–3 years). In the process, on-the-ground changes in conifer distribution and canopy cover following extensive conifer treatment (Falkowski et al., 2017) or catastrophic wildfires (Coates et al., 2016a) can be accurately quantified. Updated conifer outputs can then be integrated as contemporary
4.2. Spectral incongruence The accuracy of our conifer classification as measured against reference imagery and ground referencing inspires confidence in its utility for pinyon-juniper treatment efforts. Accuracies for the PMUs are relatively high (Landis and Koch, 1977), but we calculated wide variability among scores driven largely by fluctuation in user’s accuracies for conifer classification across the PMUs. Conifer class commission errors are likely a result of the algorithms’ incorrect assignment of shadows and non-target vegetation to the conifer class, which was prevalent in dense canopy cover and riparian areas, respectively. Inclusion of shadows was most likely exaggerated in regions of topographic shading, causing potential overestimation of canopy cover. The variation in user’s accuracies among PMU mosaics was caused by inconsistent lighting, spectral values, parallax, and processing artifacts among tiles. We attempted to address these inconsistencies by performing object training on a tile-by-tile basis in order to achieve the highest accuracy for each processed area possible. Unfortunately seamlines were still apparent, which likely stemmed from variable radiometric properties inherent in raw NAIP imagery owing to the uncalibrated digital sensors used to obtain that imagery (Falkowski et al., 2017). With the continuous improvement in resolution and the increasing availability of multi- and hyper-spectral imagery due to advancing sensor technology, we expect vast improvement in the ability to effectively extract conifers by AFE methods. Fusions of VHSR imagery with ancillary data sources such as lidar or canopy height models could also enhance the conifer classification process, which could help eliminate commission of shadows or low-growth vegetation by discerning individual trees in merged canopies from peaks in canopy structure (Maier et al., 2008; Sankey and Glenn, 2011; Jakubowski et al., 2013), which would benefit site-level management planning by estimating the number of trees to be treated. 153
Int J Appl Earth Obs Geoinformation 73 (2018) 148–155
K.B. Gustafson et al.
Conflicts of interest
http://dx.doi.org/10.1016/j.rama.2016.09.001. Congalton, R.G., Green, K., 2009. Assessing the Accuracy of Remotely Sensed Data: Principles and Practices, 2nd edition. Lewis Publishers, Boca Raton. Connelly, J.W., Knick, S.T., Schroeder, M.A., Stiver, S.J., 2004. Conservation assessment of greater sage-grouse and sagebrush habitats. Western Association of Fish and Wildlife Agencies. pp. 1–591 Unpublished Report. Davenport, D.W., Breshears, D.D., Wilcox, B.P., Allen, C.D., 1998. Viewpoint: sustainability of piñon-juniper ecosystems a unifying perspective of soil erosion thresholds. J. Range Manage. 51, 231–240. http://dx.doi.org/10.2307/4003212. Davies, K.W., Petersen, S.L., Johnson, D.D., Davis, D.B., Madsen, M.D., Zvirzdin, D.L., Bates, J.D., 2010. Estimating juniper cover from national agriculture imagery program (NAIP) imagery and evaluating relationships between potential cover and environmental variables. Rangel. Ecol. Manage. 63, 630–637. http://dx.doi.org/10. 2111/REM-D-09-00129.1. Davies, K.W., Boyd, C.S., Beck, J.L., Bates, J.D., Svejcar, T.J., Gregg, M.A., 2011. Saving the sagebrush sea: an ecosystem conservation plan for big sagebrush plant communities. Biol. Conserv. 144, 2573–2584. Falkowski, M.J., Evans, J.S., 2012. Mapping conifer trees in sage-grouse habitat using spatial wavelet analysis: a tool for prioritizing treatment of conifer encroachment. Natural Resources Conservation Service Completion Report 68-7482-10-525 and University of Montana Agreement M6480. pp. 1–33. Falkowski, M.J., Evans, J.S., Naugle, D.E., Hagen, C.A., Carleton, S.A., Maestas, J.D., Khalyani, A.H., Poznanoic, A.J., Lawrence, A.J., 2017. Mapping tree canopy cover in support of proactive prairie grouse conservation in western North America. Rangel. Ecol. Manage. 70, 15–24. http://dx.doi.org/10.1016/j.rama.2016.08.002. Hay, G.J., Castilla, G., 2006. Object-based image analysis: strengths, weaknesses, opportunities and threats (SWOT). The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. University of Calgary. Hay, G.J., Blaschke, T., Marceau, D.J., Bouchard, A., 2003. A comparison of three imageobject methods for the multiscale analysis of landscape structure. ISPRS J. Photogramm. Remote Sens. 57, 327–345. http://dx.doi.org/10.1016/S09242716(02)00162-4. Homer, C.G., Aldridge, C.L., Meyer, D.K., Coan, M.J., Bowen, Z.H., 2009. Multiscale sagebrush rangelandhabitat modelling in southwest Wyoming. USGS Open-File Report 2008-1027. pp. 1–14. Hulet, A., Roundy, B.A., Petersen, S.L., Jensen, R.R., Bunting, S.C., 2014. An object-based image analysis of pinyon and juniper woodlands treated to reduce fuels. Environ. Manage. 53, 660–671. http://dx.doi.org/10.1007/s00267-013-0227-1. Jakubowski, M.K., Li, W., Guo, Q., Kelly, M., 2013. Delineating individual trees from lidar data: a comparison of vector- and raster-based segmentation approaches. Remote Sens. 5, 4163–4186. http://dx.doi.org/10.3390/rs5094163. Knick, S.T., Hanser, S.E., Preston, K.L., 2013. Modeling ecological minimum requirements for distribution of greater sage-grouse leks: implications for population connectivity across their western range, U.S.A. Ecol. Evol. 3, 1539–1551. http://dx.doi.org/10. 1002/ece3.557. Kormos, P.R., Marks, D., Pierson, F.B., Williams, C.J., Hardegree, S.P., Havens, S., Hedrick, A., Dates, J.D., Svejcar, T.J., 2017. Ecosystem water availability in juniper versus sagebrush snow-dominated rangelands. Rangel. Ecol. Manage. 70, 116–128. http://dx.doi.org/10.1016/j.rama.2016.05.003. Landis, J.R., Koch, G.G., 1977. The measurement of observer agreement for categorical data. Biometrics 33, 159–174. http://dx.doi.org/10.2307/2529310. Maier, B., Tiede, D., Dorren, L., 2008. Characterizing Mountain Forest Structure Using Landscape Metrics on LiDAR-based Canopy Surface Models. Object-based Image Analysis. Springer, Berlin, Germany, pp. 625–643 doi:10.1007%2F978-3-540-770589_34. Miller, R.F., Tausch, R.J., 2001. The role of fire in juniper and pinyon woodlands: a descriptive analysis. In: Fire Conference. Proceedings of the Invasive Species Workshop: the Role of Fire in the Control and Spread of Invasive Species 27. pp. 15–30. Miller, R.F., Svejcar, T.J., Rose, J.A., 2000. Impacts of western juniper on plant community composition and structure. J. Range Manage. 53, 574–585. http://dx.doi.org/ 10.2307/4003150. Miller, R.F., Bates, J.D., Svejcar, T.J., Pierson, F.B., Eddleman, L.E., 2005. Biology, Ecology, and Management of Western Juniper 152. Oregon State University Technical Bulletin, pp. 1–77. Miller, R.F., Tausch, R.J., McArthur, E.D., Johnson, D.D., Sanderson, S.C., 2008. Age Structure and Expansion of Piñon-juniper Woodlands: a Regional Perspective in the Intermountain West. Rocky Mountain Research Station, pp. 1–15. http://dx.doi.org/ 10.2737/RMRS-RP-69. Report RMRS-RP-69. Miller, R.F., Knick, S.T., Pyke, D.A., Meinke, C.W., Hanser, S.E., Wisdom, M.J., Hild, A.L., 2011. Characteristics and dynamics of greater sage-grouse populations. In: Connelly, John W., Knick, Steven T. (Eds.), Greater Sage-grouse—Ecology and Conservation of a Landscape Species and its Habitats. University of California Press, Berkeley, CA, pp. 145–184. Miller, R.F., Chambers, J.C., Pellant, M., 2014. A Field Guide for Selecting the Most Appropriate Treatment in Sagebrush and Piñon-juniper Ecosystems in the Great Basin—evaluating Resilience to Disturbance and Resistance to Invasive Annual Grasses, and Predicting Vegetation Response. General Technical Report RMRS-GTR322. U.S. Department of Agriculture, Forest Service, Rocky Mountain Research Station, pp. 1–66. http://dx.doi.org/10.2737/RMRS-GTR-322. Miller, R.F., Naugle, D.E., Maestas, J.D., Hagen, C.A., Hall, G., 2017. Special issue: targeted woodland removal to recover at-risk grouse and their sagebrush-steppe and prairie ecosystems. Rangel. Ecol. Manage. 70, 1–8. http://dx.doi.org/10.1016/j. rama.2016.10.004. Mishra, N.B., Crews, K.A., 2014. Mapping vegetation morphology types in a dry savanna ecosystem: integrating hierarchical object-based image analysis with random forest. Int. J. Remote Sens. 35, 1175–1198. http://dx.doi.org/10.1080/01431161.2013.
None. Acknowledgements This research was funded by the Nevada Sagebrush Ecosystem Program with support from the Nevada Department of Wildlife and U.S. Bureau of Land Management. Additional logistical coordination was provided by the U.S. Forest Service and California Department of Fish and Wildlife. We extend gratitude to personnel with U.S. Geological Survey, particularly W. Perry for guidance and expertise, K. Mauch and T. Kroger for assistance in developing feature extraction workflows, and R. Kelble and B. Prochazka for coordinating conifer ground referencing conducted by numerous field technicians. We also thank the Great Basin Bird Observatory for field data collection. Use of trade or product names does not imply endorsement by the U.S. Government. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at https://doi.org/10.1016/j.jag.2018.06.002. References Baruch-Mordo, S., Evans, J.S., Severson, J.P., Naugle, D.E., Maestas, J.D., Kiesecker, J.M., Falkowski, M.J., Hagen, C.A., Reese, K.P., 2013. Saving sage-grouse from the trees: a proactive solution to reducing a key threat to a candidate species. Biol. Conserv. 167, 233–241. http://dx.doi.org/10.1016/j.biocon.2013.08.017. Blaschke, T., 2010. Object based image analysis for remote sensing. ISPRS J. Photogramm. Remote Sens. 65, 2–16. http://dx.doi.org/10.1016/j.isprsjprs.2009.06. 004. BLM, USFS, 2015. Fact Sheet: BLM, USFS Greater Sage-grouse Conservation Effort. Accessed April 2017. Available online. https://www.fs.fed.us/sites/default/files/ fact-sheet-greater-sage-grouse.pdf. Boswell, A., Petersen, S., Roundy, B., Jensen, R., Summers, D., Hulet, A., 2017. Rangeland monitoring using remote sensing: comparison of cover estimates from field measurements and image analysis. AIMS Environmental Science 4, 1–16. http://dx.doi. org/10.3934/environsci.2017.1.1. Bradley, B.A., Fleishman, E., 2008. Relationships between expanding pinyon-juniper cover and topography in the central Great Basin, Nevada. J. Biogeogr. 35, 951–964. http://dx.doi.org/10.1111/j.1365-2699.2007.01847.x. Bruce, D.A., 2008. Object oriented classification: case studies using different image types with different spatial resolutions. Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci. 37, 515–520. Burnett, C., Blaschke, T., 2003. A multi-scale segmentation/object relationship modelling methodology for landscape analysis. Ecol. Model. 168, 233–249. http://dx.doi.org/ 10.1016/S0304-3800(03)00139-X. Chambers, J.C., Bradley, B.A., Brown, C.S., D’Antonio, C., Germino, M.J., Grace, J.B., Hardegree, S.P., Miller, R.F., Pyke, D.A., 2014. Resilience to stress and disturbance, and resistance to Bromus tectorum L. invasion in cold desert shrublands of western North America. Ecosystems 17 (2), 360–375. http://dx.doi.org/10.1016/j.rama. 2016.08.005. Coates, P.S., Casazza, M.L., Ricca, M.A., Brussee, B.E., Blomberg, E.J., Gustafson, K.B., Overton, C.T., Davis, D., Neill, L., Espinosa, S.P., Gardner, S.C., Delehanty, D.J., 2014. Spatially explicit modeling of greater sage-grouse (Centrocercus urophasianus) habitat in Nevada and Northeastern California—a decision-support tool for management. U.S. Geological Survey Open-File Report 2014-1163. http://dx.doi.org/10. 3133/ofr20141163. 84 p. Coates, P.S., Ricca, M.A., Prochazka, B.G., Brooks, M.L., Doherty, K.E., Kroger, T., Blomberg, E.J., Hagen, C.A., Casazza, M.L., 2016a. Wildfire, climate, and invasive grass interactions negatively impact an indicator species by reshaping sagebrush ecosystems. Proc. Natl. Acad. Sci. 113, 12745–12750. http://dx.doi.org/10.1073/ pnas.1606898113. Coates, P.S., Casazza, M.L., Ricca, M.A., Brussee, B.E., Blomberg, E.J., Gustafson, K.B., Overton, C.T., Davis, D.M., Niell, L.E., Espinosa, S.P., Gardner, S.C., Delehanty, D.J., 2016b. Integrating spatially explicit indices of abundance and habitat quality: an applied example for greater sage-grouse management. J. Appl. Ecol. 53, 83–95. http://dx.doi.org/10.1111/1365-2664.12558. Coates, P.S., Casazza, M.L., Brussee, B.E., Ricca, M.A., Gustafson, K.B., Sanchez-Chopitea, E., Mauch, K., Niell, L., Gardner, S., Espinosa, S., Delehanty, D.J., 2016c. Spatially explicit modeling of annual and seasonal habitat for greater sage-grouse (Centrocercus urophasianus) in Nevada and Northeastern California – an updated decision-support tool for management. U.S. Geological Survey Open-File Report 2016-1080. pp. 1–160. Coates, P.S., Prochazka, B.G., Ricca, M.A., Gustafson, K.B., Ziegler, P., Casazza, M.L., 2017. Pinyon and juniper encroachment into sagebrush ecosystems impacts distribution and survival of greater sage-grouse. Rangel. Ecol. Manage. 70, 25–38.
154
Int J Appl Earth Obs Geoinformation 73 (2018) 148–155
K.B. Gustafson et al.
Basin. Rangel. Ecol. Manage. 70, 50–58. http://dx.doi.org/10.1016/j.rama.2016.07. 011. Soulé, P.T., Knapp, P.A., 1999. Western juniper expansion on adjacent disturbed and near-relict sites. J. Range Manage. 52, 525–533. Soulé, P.T., Knapp, P.A., Grissino-Mayer, H.D., 2004. Human agency, environmental drivers, and western juniper establishment during the late holocene. Ecol. Appl. 14, 96–112. http://dx.doi.org/10.1890/02-5300. Strand, E.K., Bunting, S.C., Keefe, R., 2013. Influence of wildland fire along a successional gradient in sagebrush steppe and western juniper woodlands. Rangel. Ecol. Manage. 66 (6), 667–679. Tausch, R.J., Miller, R.F., Roundy, B.A., Chamber, J.C., 2009. Piñon and juniper field guide: asking the right questions to select appropriate management actions. USGS Circ. 1335, 1–95. Tsai, Y.H., Stow, D., Weeks, J., 2011. Comparison of object-based image analysis approaches to mapping new buildings in Accra, Ghana using multi-temporal QuickBird satellite imagery. Int. J. Remote Sens. 3, 2707–2726. Twidwell, D., Allred, B.W., Fuhlendorf, S.D., 2013. National-scale assessment of ecological content in the world’s largest land management framework. Ecosphere 4, 1–27. http://dx.doi.org/10.1890/ES13-00124.1. U.S. Department of the Interior, 2015. Secretarial Order No. 3336 - Rangeland Fire Prevention, Management, and Restoration. Accessed August 2015. Available online. http://www.fws.gov/greatersagegrouse/documents/Threats/20150106DOI-FireOrder.pdf. USDA, 2013. National Agriculture Imagery Program (NAIP) Digital Orthophoto Quarter Quads. Accessed November 2013. Available online. https://gdg.sc.egov.usda.gov/. USDA, 2014. “Targeted conifer Removal: a Proactive Solution to Conserving Sagegrouse.” Conservation Insight Conservation Effects Assessment Project, 1-4. Accessed April 2017. Available online. https://www.nrcs.usda.gov/Internet/FSE_ DOCUMENTS/stelprdb1261131.pdf. USDA NASS, 2014. 2012 Census of Agriculture United States Summary and State Data. 1, 1-695. Accessed April 2017. Available online. . https://www.agcensus.usda.gov/ Publications/2012/Full_Report/Volume_1,_Chapter_1_US/usv1.pdf. USFWS, 2010. Endangered and Threatened Wildlife and Plants; 12-Month Finding on a Petition to List Greater Sage-grouse (Centrocercus Urophasianus) as Threatened or Endangered. Proposed Rule: Federal Register, 75, 13910-14014. Accessed April 2017. Available online. https://www.federalregister.gov/documents/2010/03/23/ 2010-5132/endangered-and-threatened-wildlife-and-plants-12-month-findings-forpetitions-to-list-the-greater. USFWS, 2013. Greater Sage-grouse (Centrocercus Urophasianus) Conservation Objectives: Final Report, 1-92. Accessed April 2017. Available online. . https:// www.fws.gov/greatersagegrouse/documents/COT-Report-with-Dear-InterestedReader-Letter.pdf. USFWS, 2015. Endangered and Threatened Wildlife and Plants; 12-Month Finding on a Petition to List Greater Sage-grouse (Centrocercus Urophasianus) as an Endangered Species. Accessed April 2017. Available online. USFWS, Department of Interior, 80 FR 59857. https://www.federalregister.gov/documents/2015/10/02/2015-24292/ endangered-and-threatened-wildlife-and-plants-12-month-finding-on-a-petition-tolist-greater. Weiner, N.I., Strand, E.K., Bunting, S.C., Smith, A.M.S., 2016. Duff distribution influences fire severity in post-fire vegetation recovery in sagebrush steppe. Ecosystems 19, 1196–1209. http://dx.doi.org/10.1007/s10021-016-9994-x.
876120. Noss, R.F., LaRoe III, E.T., Scott, J.M., 1995. Endangered Ecosystems of the United States: a Preliminary Assessment of Loss and Degradation 28. U.S. Fish and Wildlife Service, pp. 1–95 Biological Report. O’Brien, M.A., 2003. Feature extraction with the VLS feature analyst system. Proc. the Am. Soc. Photogramm. Remote Sens. 1–9. Opitz, D., Blundell, S., 2008. Object Recognition and Image Segmentation: the Feature Analyst Approach. Object-based Image Analysis, Ch. 2.3. Springer, Berlin, Germany, pp. 153–167. http://dx.doi.org/10.1007/978-3-540-77058-9_8. Overwatch, 2013. Feature Analyst Version 5.1.2.0 for ArcGIS. Overwatch Systems Ltd. Pierson, F.B., Williams, C.J., Kormos, P.R., Hardegree, S.P., Clark, P.E., Rau, B.M., 2010. Hydrologic vulnerability of sagebrush steppe following pinyon and juniper encroachment. Rangel. Ecol. Manage. 63, 614–629. http://dx.doi.org/10.2111/REM-D09-00148.1. Poznanovic, A.J., Falkowski, M.J., Maclean, A.L., Smith, A.M.S., Evans, J.S., 2014. An accuracy assessment of tree detection algorithms in juniper woodlands. Photogramm. Eng. Remote Sens. 80, 627–637. http://dx.doi.org/10.14358/PERS.80.7.627. Prochazka, B.G., Coates, P.S., Ricca, M.A., Casazza, M.L., Gustafson, K.B., Hull, J.M., 2017. Encounters with pinyon-juniper influence riskier movements in greater sagegrouse across the great basin. Rangel. Ecol. Manage. 70, 39–49. http://dx.doi.org/10. 1016/j.rama.2016.07.004. Ricca, M.A., Coates, P.S., Gustafson, K.B., Brussee, B.E., Chambers, J.C., Espinosa, S.E., Gardner, S.C., Lisius, S., Ziegler, P., Delehanty, D.J., Casazza, M.L., 2018. A conservation planning tool for greater sage-grouse using indices of species distribution, resilience, and resistance. Ecol. Appl. 28, 878–896. Romme, W.H., Allen, C.D., Bailey, J.D., Baker, W.L., Bestelmeyer, B.T., Brown, P.M., Eisenhart, K.S., Floyd, M.L., Huffman, D.W., Jacobs, B.F., Miller, R.F., Muldavin, E.H., Swetnam, T.W., Tausch, R.J., Weisberg, P.J., 2009. Historical and modern disturbance regimes, stand structures, and landscape dynamics of piñon-juniper vegetation of the western United States. Rangel. Ecol. Manage. 62, 203–222. http://dx. doi.org/10.2111/08-188R1.1. Roth, C.L., Coates, P.S., Gustafson, K.B., Chenaille, M.P., Ricca, M.A., Sanchez-Chopitea, E., and Casazza, M.L. (In Review). A customized framework for regional classification of conifers using automated feature extraction. Manuscript submitted for publication. Roundy, D.B., 2015. Estimating pinyon and juniper cover across Utah using NAIP imagery. All Theses and Dissertations, Paper 5575. pp. 24. http://dx.doi.org/10.3934/ environsci.2016.4.765. Roundy, B.A., Young, K., Cline, N., Hulet, A., Miller, R.F., Tausch, R.J., Chambers, J.C., Rau, B., 2014. Piñon-juniper reduction increases soil water availability of the resource growth pool. Rangel. Ecol. Manage. 67, 495–505. http://dx.doi.org/10.2111/ REM-D-13-00022.1. Sankey, T., Glenn, N., 2011. Landsat-5 TM and lidar fusion for sub-pixel juniper tree cover estimates in a western rangeland. Photogramm. Eng. Remote Sens. 77, 1–8. http:// dx.doi.org/10.14358/PERS.77.12.1241. Schroeder, M.A., Aldridge, C.L., Apa, A.D., Bohne, J.R., Braun, C.E., Bunnell, S.D., Connelly, J.W., Deibert, P.A., Gardner, S.C., Hilliard, M.A., Kobriger, G.D., McAdam, S.M., McCarthy, C.W., McCarthy, J.J., Mitchell, D.L., Rickerson, E.V., Stiver, S.J., 2004. Distribution of sage-grouse in North America. Condor 106, 363–376. http://dx. doi.org/10.1650/7425. Severson, J.P., Hagen, C.A., Maestas, J.D., Naugle, D.E., Forbes, J.T., Reese, K.P., 2017. Short-term response of sage-grouse nesting to conifer removal in the northern Great
155
Contents lists available at ScienceDirect
Int J Appl Earth Obs Geoinformation journal homepage: www.elsevier.com/locate/jag
Using object-based image analysis to conduct high- resolution conifer extraction at regional spatial scales
T
⁎
K. Benjamin Gustafson, Peter S. Coates , Cali L. Roth, Michael P. Chenaille, Mark A. Ricca, Erika Sanchez-Chopitea, Michael L. Casazza Western Ecological Research Center, U.S. Geological Survey, Dixon, CA, United States
A R T I C LE I N FO
A B S T R A C T
Keywords: Object-based image analysis Pinyon-juniper Sage-grouse Sagebrush Feature extraction Image processing
Distributional expansion and infill of pinyon (Pinus monophylla) and juniper (Juniperus osteosperma, J. occidentalis) trees (hereinafter, "pinyon-juniper") into sagebrush ecosystems alters the ecological function and economic viability of these ecosystems and represents a major contemporary challenge facing land and wildlife managers. Therefore, accurate and high-resolution maps of pinyon-juniper distribution and abundance across broad geographic extents would facilitate science that quantifies ecological effects of pinyon-juniper expansion and help guide land management decisions that better target areas for pinyon-juniper treatment projects. We mapped conifers at a high (1- m2; i.e., 1 × 1-m) resolution across the majority of Nevada and northeastern California. We used digital orthophoto quad tiles from National Agriculture Imagery Program (USDA, 2013) to classify conifers using automated feature extraction (AFE) with the program Feature Analyst™ (Overwatch, 2013). Overall accuracy was > 86% across all mapped areas for ground referencing methods. We provide five sets of full-extent maps for land managers: (1) a shapefile representing accuracy results linked to mapping subunits; (2) binary rasters representing conifer presence or absence at a 1-m2 resolution; (3) a 900-m2 resolution raster representing percentages of conifer canopy cover within each cell; (4) 1-m2 resolution canopy cover classification rasters derived from a 50-m radius moving window analysis; and (5) an example map derived from our canopy cover product that prioritizes pinyon-juniper treatment by significance to sage-grouse habitat improvement. Importantly, the canopy cover maps were developed to allow user-specified flexibility based on their own objectives (i.e., develop phases of expansion). These products improve upon or complement existing conifer maps for the Western United States and will help facilitate habitat management and sagebrush ecosystem restoration through an accurate understanding of conifer distribution and abundance at multiple spatial scales.
1. Introduction 1.1. Ecological background The iconic "sagebrush sea" that characterizes the Great Basin of the Western United States provides habitat for several at-risk sagebrushobligate species (Homer et al., 2009; Knick et al., 2013) and supports outdoor recreation valued at $1 billion annually (BLM and USFS, 2015) and cattle grazing, valued as high as $60 billion in a single year (USDA NASS, 2014). It is also one of the most endangered ecosystems in the United States (Noss et al., 1995), as the range of sagebrush has contracted by more than 50% since European settlement (Schroeder et al., 2004). The remaining range is subjected to continual fragmentation, largely from anthropogenic-related disturbances (Schroeder et al., 2004). Accordingly, sagebrush ecosystems are at the center of national
⁎
conservation strategy (U.S. Department of the Interior, 2015). Conifers such as pinyon (Pinus monophylla) and juniper (Juniperus sp.; hereinafter, "pinyon-juniper") are a native component of the sagebrush ecosystems within the Great Basin. However, the distribution and abundance of pinyon-juniper has greatly expanded following European settlement (Miller and Tausch, 2001) owing to a variety of factors, including changes in climate (Soulé et al., 2004; Romme et al., 2009), land use (Romme et al., 2009), and fire regimes (Miller et al., 2000; Soulé et al., 2004). Accordingly, expansion of pinyon-juniper is a major factor contributing to loss and fragmentation of sagebrush ecosystems (Davies et al., 2011; Miller et al., 2011; Knick et al., 2013). Dominance of sagebrush and perennial grasses, which contribute strongly to sagebrush ecosystem resilience to disturbance and resistance to invasion (Chambers et al., 2014), decreases as cover of pinyon-juniper increases (Miller et al., 2005). This relationship can be categorized into three
Corresponding author. E-mail address: [email protected] (P.S. Coates).
https://doi.org/10.1016/j.jag.2018.06.002 Received 14 February 2018; Received in revised form 4 June 2018; Accepted 5 June 2018 0303-2434/ Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/BY-NC-ND/4.0/).
Int J Appl Earth Obs Geoinformation 73 (2018) 148–155
K.B. Gustafson et al.
Fig. 1. Schematic showing conceptual model of conifer classification framework using automated feature extraction methods within greater sage-grouse habitats in Nevada and California. Input is represented as orange diamonds and green hexagons identify products. Blue and gray boxes represent steps conducted in Feature Analyst™ and a GIS, respectively. Starred items (*) denote iterated geoprocessing steps.
distinct phases of woodland development, where sagebrush is dominant over pinyon-juniper in phase 1, sagebrush and pinyon-juniper are codominant in phase 2, and pinyon-juniper is dominant and has replaced sagebrush in phase 3 (Miller et al., 2005, 2008). The loss of herbaceous understory diminishes forage and cover for constituent wildlife (Miller et al., 2000, 2011) and reduces available forage for cattle (Soulé and Knapp, 1999; Twidwell et al., 2013). Pinyon-juniper expansion also reduces streamflow (Kormos et al., 2017), depletes soil water availability (Roundy et al., 2014; Kormos et al., 2017), and increases bare earth contiguity that promotes runoff and subsequent soil erosion (Davenport et al., 1998; Pierson et al., 2010). Perhaps most importantly, the spread and infill of pinyon-juniper into low-elevation sagebrush ecosystems introduces woody biomass that fuels more intense wildfires (Bradley and Fleishman, 2008; Romme et al., 2009; Weiner et al., 2016; Strand et al., 2013). Burned stands are often replaced by invasive annual grasses that have a positive feedback cycle with wildfire, thereby increasing fire extent and frequency and spreading fire into sagebrush that would otherwise be much less likely to burn (Tausch et al., 2009; Romme et al., 2009; Davies et al., 2011). Because of the multitude of economic and ecological effects from sagebrush loss, pinyon-juniper expansion is now a primary challenge facing land managers in the Western United States (Davies et al., 2011).
regional GIS products derived from satellite imagery generally do not have the spatial resolution necessary to delineate early stages of pinyon-juniper expansion (e.g., as low as 10% canopy cover) that often constitute the best candidate areas for treatment (Miller et al., 2008; Baruch-Mordo et al., 2013; Coates et al., 2017). Additionally, mapping land cover at a 900-m2 resolution can overestimate pinyon-juniper density because the cell size of the imagery often exceeds the diameter of individual trees on the landscape. Feature recognition technologies such as object-based image analysis (OBIA) can be applied to very high spatial resolution (VHSR) imagery (< 4-m2) to identify individual trees across regional extents (Davies et al., 2010; Falkowski and Evans, 2012; Mishra and Crews, 2014; Falkowski et al., 2017). OBIA exploits VHSR imagery by segmenting cells into image-objects based on their spectral, spatial, and structural properties, and then classifies these image-objects to extract features of interest (Burnett and Blaschke, 2003; Hay et al., 2003; Hay and Castilla, 2006). Segmentation facilitates accurate tree delineation from VHSR imagery as conifer canopies encompass multiple cells with variable properties (Falkowski et al., 2017). Conifer classification using OBIA on VHSR imagery has been shown to reliably identify forest canopy in relatively small test scenarios (Davies et al., 2010; Falkowski and Evans, 2012; Hulet et al., 2014; Poznanovic et al., 2014; Roundy, 2015).
1.2. Need for high-resolution remote sensing
1.3. Objectives
Pinyon-juniper expansion has regional consequences, requiring cooperative efforts by Federal and State agencies that total millions of dollars (Sanford et al., 2017) and form a principal component of current wildlife conservation plans (USFWS, 2013; Miller et al., 2014; USDA, 2014; Miller et al., 2017; Severson et al., 2017). As agencies implement restoration efforts over thousands of hectares (Miller et al., 2017), there is an immediate need for high-resolution pinyon-juniper data over regional spatial extents to optimize targeted treatment efforts (Connelly et al., 2004; Homer et al., 2009; Falkowski et al., 2017). Customary
We used OBIA to create 1-m2 resolution binary conifer rasters across our study extent. We assessed mapping accuracy by analyzing errors of omission and commission using both reference imagery and ground referencing. We then scaled the 1-m2 resolution data into unsmoothed (900-m2) and 50-m radius moving-window smoothed (1-m2) estimates of percent canopy cover. The 900-m2 product allows for seamless integration into current geospatial applications that use standard 900-m2 resolution products, while the 1-m2 product allows users to estimate conifer cover into more customizable bins than currently available. To 149
Int J Appl Earth Obs Geoinformation 73 (2018) 148–155
K.B. Gustafson et al.
grouse avoid canopy cover as low as 4% (Baruch-Mordo et al., 2013; Coates et al., 2017), therefore we maintained individual intervals for each percentage < 10%. For conservation purposes, we developed a map to help guide decisions related to pinyon-juniper treatment, specifically related to sage-grouse conservation within our study extent. The map consisted of phase I expansion (< 10%) intersected with documented sage-grouse habitat management categories (Coates et al., 2016c). These categories were a result of combining spatially explicit habitat quality maps based on seasonal resource selection function with sage-grouse abundance and space use data (Coates et al., 2016c). We defined categories of pinyon-juniper management for sage-grouse habitat restoration as phase I pinyon-juniper within: 1) Priority – high quality seasonal habitats that are often occupied by sage-grouse; 2) High – high quality seasonal habitats that are less likely to be occupied by sage-grouse, as well as transitional areas comprised of non-habitat that are temporarily occupied (between seasons); 3) Moderate –moderate to low quality seasonal habitats that are not typically occupied; and 4) Low – non-habitat and non-occupied.
illustrate the utility of our products for sagebrush ecosystem management efforts, we derived a map that prioritizes areas for pinyon-juniper treatment aimed at restoring sage-grouse habitat. 2. Methods 2.1. Feature extraction We conducted conifer mapping using Automated Feature Extraction (AFE) in the program Feature AnalystTM (Overwatch, 2013) for all 61 Nevada Department of Wildlife sage-grouse Population Management Units (PMU), which corresponded to sage-grouse habitat mapping efforts (Coates et al., 2016b,c). We focus on PMUs because sage-grouse are an important indicator-species for the health of the sagebrush ecosystems and a major impetus for sagebrush restoration in the Great Basin, as the species has been considered for listing multiple times under the Endangered Species Act of 1973 (USFWS, 2010, 2015). Our study extent included 6230 DOQQs from Nevada, California, Oregon, Idaho and Utah (USDA, 2013). Fig. 1 illustrates our mapping process steps, based on the methodology provided in Roth et al. (2018). We performed traditional accuracy assessment with an estimated accuracy coefficient (kappa) analysis (Roth et al., 2018), where the kappa coefficient (Khat ) represents the percent accuracy adjusted for correct classification due to random chance (Congalton and Green, 2009). Khat > 60% indicate substantial agreement between classification and truth, and those > 80% are almost perfect (Landis and Koch, 1977).
3. Results 3.1. Reference imagery accuracy assessment We summarized accuracy results across all PMUs (Table 1); more detailed reports of accuracy assessments for each PMU are provided in Roth et al. (2018). On average, conifer classification had higher incidence of errors of commission reported as conifer user’s accuracy (76%; SD = 9%; Table 1) than non-conifer user's accuracy (97%; SD = 3%; Table 1). Accordingly, errors of omission were more common for non-conifer classification (producer’s accuracy = 81%; SD = 6%; Table 1) compared to conifer classification (97%; SD = 4%; Table 1). The mean overall accuracy was 86% (SD = 4%; Table 1). Mean adjusted accuracy showed that there was substantial agreement between classes and reference imagery (Khat = 73%; SD = 9%; Table 1). The average distance between the extracted conifer feature and the nearest vegetation > 3-m in height within a 30-m radius was 3.18-m (SD = ± 8.80 m). On average, 87% of the points we classified as conifers were confirmed by ground referencing, yielding a 13% error of commission (Table 2). Error of omission was 17%, with 83% of the conifers selected in the field as omission points (n = 450) correctly classified by our 1-m2 conifer map (Table 2). Table 2 compares the ground-referenced accuracies to the user’s and producer’s accuracies of the PMU that contains the respective site.
2.2. Ground referencing We ground referenced our results to estimate OBIA mapped conifer accuracy for on-the-ground pinyon-juniper management. Because land managers will use the conifer map to target conifers for treatment, we selectively evaluated the accuracy of the conifer class. We generated commission points from our 1-m conifer map to be ground referenced for correct classification in 8 sage-grouse study sites (Coates et al., 2016c) within the project extent. We randomly selected a maximum (up to 100) conifer locations that met three criteria: (1) in phase 1 or 2 woodland development (potential targets for treatment), (2) within 100-m of roads and (3) at slopes ≤ 30% rise for accessibility. Commission points were at least 180-m apart to avoid overlap. We recorded the coordinates and species of the nearest tree-like vegetation > 3-m in height within a 30-m search radius to account for possible compounded location error in NAIP imagery and GPS units. To measure errors of omission, we navigated to the nearest conifer that was between 60 and 90-m away from the commission point and recorded location and species. We specified a window of 60–90-m to ensure independence between commission and omission points. Error estimates were generated for commission and omission points by identifying the presence of a ground referenced or extracted conifer, respectively, within a 10-m buffer. We used a 10-m buffer, permitting for 5-m reported accuracy in both NAIP and handheld GPS units.
3.2. Conifer extraction products We generated distinct conifer mapping products for the full extent of our study area (10.5066/F7348HVC), which are: (1) confusion matrix results linked to PMU or zone; (2) binary rasters identifying conifer presence or absence at a 1-m2 resolution, by PMU; (3) a 900-m2 resolution raster representing percentages of conifer canopy cover within each cell (Fig. 1b); a (4) 50-m radius moving window canopy cover class rasters at a 1-m2 resolution, in quadrants (Fig. 1c); and (5) A map prioritizing pinyon-juniper management for sage-grouse habitat restoration efforts (Fig. 3). Importantly, the percent canopy cover products can be reclassified in a GIS into user-specified bins to meet specific objectives by land and wildlife managers (i.e., approximations for phases of pinyon-juniper expansion).
2.3. Map applications After validation, we used the final mosaics to calculate high resolution estimates of percent canopy cover for the study area in two distinct ways (Fig. 2). In the first method, we calculated percent canopy cover per 900-m2 cell by summing the number of 1-m conifer cells within a 900-m2 cell and dividing by its area, converting to whole numbers of percent canopy cover (Fig. 1f). We also calculated percent canopy cover by summing all cell values within a 50-m radius neighborhood (equivalent to 7845 m2) and dividing by the total number of cells (Fig. 1g). To facilitate use by land managers, we reclassified the cells according to multiple intervals of percent canopy cover (Supplemental Material, Table S1) based on known biological significance to sage-grouse (Baruch-Mordo et al., 2013; Coates et al., 2017). Sage-
4. Discussion 4.1. Applications We provide highly accurate maps of conifer distribution and canopy cover derived from conifer extraction from NAIP imagery at a 1-m2 resolution across the full extent of mapped sage-grouse habitat in 150
Int J Appl Earth Obs Geoinformation 73 (2018) 148–155
K.B. Gustafson et al.
Fig. 2. Products produced using automated feature extraction methods across greater sage-grouse habitat in Nevada and California depicting: a) conifer presence or absence at a 1-m2 resolution; b) continuous conifer canopy cover at 900-m2 resolution raster, and c) example of using canopy cover bins at a 1-m2 resolution produced from a 50-m radius moving window to depict progressive phases of conifer expansion.
Nevada and northeastern California, which will help facilitate conservation and restoration efforts across large spatial extents, such as sage-grouse habitat restoration within the Great Basin. We achieved high accuracy by implementing AFE in Feature Analyst™, a highly precise OBIA classification method that heretofore has had limited application at large spatial extents largely because of computational and time demands (Hay and Castilla, 2006; Bruce, 2008; Tsai et al., 2011) but can be more accurate than traditional supervised classification methods normally conducted at such extents (e.g., O’Brien, 2003; Bruce, 2008; Opitz and Blundell, 2008; Blaschke, 2010; Tsai et al., 2011). To our knowledge, we present the only AFE-based products across a significant part of the geographic distribution of sagebrush and sage-grouse, and our products complement other OBIA-based conifer maps for sagebrush ecosystems in the Western United States (e.g., Falkowski et al., 2017). Moreover, our outputs more accurately (87%) and precisely (within ∼4-m) reflect ground conditions than 900-m2 resolution Landsat- derived conifer products (Fig. 4) in areas of low to moderate canopy cover. This allows confidence for landscape-level management of sagebrush ecosystems through traditional site-level treatments, whereby managers can perform detailed, remote evaluations of treatment efforts, down to the cutting of individual trees. Therefore the combination of landscape- and site-level accuracies can help land managers compare candidate sites and optimize conifer treatment throughout the entire state (Falkowski and Evans, 2012; Falkowski et al., 2017). Our high resolution conifer maps can offer a decision support tool for land managers and also help to inform further ecological studies. For example, such maps can help refine models that predict distribution and abundance for a variety of wildlife species that respond to pinyonjuniper in the environment, including sage-grouse populations within Nevada and California (Coates et al., 2014; Coates et al., 2016a,b,c). Additionally, high resolution conifer maps are necessary to understand
Table 2 User’s and producer’s accuracies for ground-referencing sites compared to the accuracy of their respective PMU determined from reference imagery. Site
PMUs
Conifer User’s Accuracy (%)
Conifer Producer’s Accuracy (%)
Site
PMU
Site
PMU
Desatoya
Desatoya, Reese River, Toiyabe, Clan Alpline
93
86
81
98
McGinness
Toyiabe
92
73
89
99
Midway
Butte/Buck/White Pine, Diamond
94
82
84
92
Monitor
Monitor (Zone 2)
88
62
86
96
North SWIP
Schell/Antelope, Butte/Buck/White Pine
100
78
100
94
South SWIP
Butte/Buck/White Pine
80
83
91
92
Susanville
Buffalo-Skedaddle
86
71
83
92
Virginia Mountains
Virginia-Pahrah
73
78
50
98
87
Mean
83
the underlying mechanisms (Coates et al., 2017; Prochazka et al., 2017) of how conifers adversely influence sage-grouse population dynamics (Baruch-Mordo et al., 2013). Moreover, management is typically conducted within categorized bins rather than continuous scales, so managers require maps that allow for demarcation of canopy-cover classes based on the most recent quantified effects of pinyon-juniper on factors such as sage-grouse
Table 1 Summarized results of mapping conifers at the 1-m2 resolution across all population management units in Nevada and California using feature extraction methods within greater sage-grouse habitat of Nevada and California. Means are shown with + 1 standard deviation. Khat is the estimated accuracy coefficient.
Mean Minimum Maximum
Conifer user’s accuracy (%)
Non-conifer user's accuracy (%)
Conifer producer's accuracy (%)
Non-conifer producer’s accuracy (%)
Overall accuracy (%)
Khat (%)
76 ± 9 60 95
97 ± 3 90 100
97 ± 3.59 88 100
81 ± 6 71 95
86 ± 4 79 97
73 ± 9 58 94
151
Int J Appl Earth Obs Geoinformation 73 (2018) 148–155
K.B. Gustafson et al.
Fig. 3. Phase I expansion pinyon-juniper prioritized by sage-grouse management categories.
2008; Boswell et al., 2017). At large spatial extents, these high-resolution products can help land managers remotely evaluate and compare candidate sites for conifer treatment across large spatial extents (Falkowski and Evans, 2012; Falkowski et al., 2017). In regard to sage-grouse, high resolution conifer maps could be used to calculate ecological benefits of removing pinyon-juniper trees to sage-grouse populations by simulating pinyon-juniper treatment and integrating results into sage-grouse habitat and demographic models, allowing for quantification of direct improvement to habitat suitability and
population dynamics (Baruch-Mordo et al., 2013; Coates et al., 2017) and sagebrush ecosystem process (Miller et al., 2000, 2005; Tausch et al., 2009; Falkowski and Evans, 2012). Our percent canopy cover outputs (smoothed and unsmoothed) have built-in flexibility that allow users to select the cutoffs that are appropriate for their purposes. Land managers can further economically value their sites by developing scenarios for assessing how well they meet the specific goals of each project, which can then help streamline allocation of time and resources currently necessary for project appraisals (Opitz and Blundell, 152
Int J Appl Earth Obs Geoinformation 73 (2018) 148–155
K.B. Gustafson et al.
Fig. 4. Example conifer classification output overlay (in yellow; left) and percent canopy cover calculation overlay (right) from conifers mapped in the same area of Nevada using automated feature extraction.
land cover covariates used to spatially explicitly model sage-grouse habitat suitability (Coates et al., 2016b). Thus, this overarching framework supports the development of "living layers" that reflect existing sage-grouse habitat conditions and provide highly accurate, timely information to land managers responding to changing landscapes. Our results can also be used to determine accuracy of other products that cover larger extents, which might provide an indication of the reliability of products developed for management efforts beyond our study extent.
demographic rates. This process can be valuable for calculating conservation credits as ecological currency for large extent programs. Additional use of the cover class settings can refine target treatment areas to those within transitional phase I, identifying areas where trees have the greatest adverse effects on sage-grouse survival (Coates et al., 2017) and, thus, allocate funds to those sites that maximize benefits to sagegrouse (Miller et al., 2008; Baruch-Mordo et al., 2013; Ricca et al., 2018). Our map of pinyon-juniper management categories can be used to guide managers when prioritizing sites for pinyon-juniper treatment with the goal of sage-grouse habitat restoration. We combine accurate identification of phase I expansion with knowledge of composite habitat quality and space use to provide suggested areas to target for pinyonjuniper treatment. While site-level evaluation of woodland development such as historic state and woodland age cannot be addressed with our map, we provide spatially explicit direction for future selection of candidate treatment sites. These mapping products also enable non-wildlife related landscapelevel management through traditional site-level treatments and facilitates managers to prioritize and efficiently implement restoration projects, which include forage for cattle (Soulé and Knapp, 1999; Twidwell et al., 2013), changes in soil erosion patterns (Davenport et al., 1998; Pierson et al., 2010), or reductions of wildfire fuel loads, extent, and severity (Hulet et al., 2014). Similar to our development of pinyon-juniper management categories for restoring sage-grouse habitat, our products can be used to develop new maps and datasets for other management goals, such as combining canopy cover with topographic and bare earth information for soil erosion studies or determining crown density or proximity to fine fuel build up wildfire prevention. Finally, a major advantage of our framework is its replicability. The methods described in Roth et al. (2018) can be applied as is to other imagery in other parts of the western United States and the flexibility of Feature Analyst™ allows the specific parameters of the classification algorithm to be easily adapted for other land cover classes in different regions, permitting image availability. We designed our framework with the ability to include improved reference imagery as it becomes available. For example, areas of low accuracy caused by poor image quality can be readily reclassified and updated as higher resolution NAIP imagery becomes available (typically every 2–3 years). In the process, on-the-ground changes in conifer distribution and canopy cover following extensive conifer treatment (Falkowski et al., 2017) or catastrophic wildfires (Coates et al., 2016a) can be accurately quantified. Updated conifer outputs can then be integrated as contemporary
4.2. Spectral incongruence The accuracy of our conifer classification as measured against reference imagery and ground referencing inspires confidence in its utility for pinyon-juniper treatment efforts. Accuracies for the PMUs are relatively high (Landis and Koch, 1977), but we calculated wide variability among scores driven largely by fluctuation in user’s accuracies for conifer classification across the PMUs. Conifer class commission errors are likely a result of the algorithms’ incorrect assignment of shadows and non-target vegetation to the conifer class, which was prevalent in dense canopy cover and riparian areas, respectively. Inclusion of shadows was most likely exaggerated in regions of topographic shading, causing potential overestimation of canopy cover. The variation in user’s accuracies among PMU mosaics was caused by inconsistent lighting, spectral values, parallax, and processing artifacts among tiles. We attempted to address these inconsistencies by performing object training on a tile-by-tile basis in order to achieve the highest accuracy for each processed area possible. Unfortunately seamlines were still apparent, which likely stemmed from variable radiometric properties inherent in raw NAIP imagery owing to the uncalibrated digital sensors used to obtain that imagery (Falkowski et al., 2017). With the continuous improvement in resolution and the increasing availability of multi- and hyper-spectral imagery due to advancing sensor technology, we expect vast improvement in the ability to effectively extract conifers by AFE methods. Fusions of VHSR imagery with ancillary data sources such as lidar or canopy height models could also enhance the conifer classification process, which could help eliminate commission of shadows or low-growth vegetation by discerning individual trees in merged canopies from peaks in canopy structure (Maier et al., 2008; Sankey and Glenn, 2011; Jakubowski et al., 2013), which would benefit site-level management planning by estimating the number of trees to be treated. 153
Int J Appl Earth Obs Geoinformation 73 (2018) 148–155
K.B. Gustafson et al.
Conflicts of interest
http://dx.doi.org/10.1016/j.rama.2016.09.001. Congalton, R.G., Green, K., 2009. Assessing the Accuracy of Remotely Sensed Data: Principles and Practices, 2nd edition. Lewis Publishers, Boca Raton. Connelly, J.W., Knick, S.T., Schroeder, M.A., Stiver, S.J., 2004. Conservation assessment of greater sage-grouse and sagebrush habitats. Western Association of Fish and Wildlife Agencies. pp. 1–591 Unpublished Report. Davenport, D.W., Breshears, D.D., Wilcox, B.P., Allen, C.D., 1998. Viewpoint: sustainability of piñon-juniper ecosystems a unifying perspective of soil erosion thresholds. J. Range Manage. 51, 231–240. http://dx.doi.org/10.2307/4003212. Davies, K.W., Petersen, S.L., Johnson, D.D., Davis, D.B., Madsen, M.D., Zvirzdin, D.L., Bates, J.D., 2010. Estimating juniper cover from national agriculture imagery program (NAIP) imagery and evaluating relationships between potential cover and environmental variables. Rangel. Ecol. Manage. 63, 630–637. http://dx.doi.org/10. 2111/REM-D-09-00129.1. Davies, K.W., Boyd, C.S., Beck, J.L., Bates, J.D., Svejcar, T.J., Gregg, M.A., 2011. Saving the sagebrush sea: an ecosystem conservation plan for big sagebrush plant communities. Biol. Conserv. 144, 2573–2584. Falkowski, M.J., Evans, J.S., 2012. Mapping conifer trees in sage-grouse habitat using spatial wavelet analysis: a tool for prioritizing treatment of conifer encroachment. Natural Resources Conservation Service Completion Report 68-7482-10-525 and University of Montana Agreement M6480. pp. 1–33. Falkowski, M.J., Evans, J.S., Naugle, D.E., Hagen, C.A., Carleton, S.A., Maestas, J.D., Khalyani, A.H., Poznanoic, A.J., Lawrence, A.J., 2017. Mapping tree canopy cover in support of proactive prairie grouse conservation in western North America. Rangel. Ecol. Manage. 70, 15–24. http://dx.doi.org/10.1016/j.rama.2016.08.002. Hay, G.J., Castilla, G., 2006. Object-based image analysis: strengths, weaknesses, opportunities and threats (SWOT). The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. University of Calgary. Hay, G.J., Blaschke, T., Marceau, D.J., Bouchard, A., 2003. A comparison of three imageobject methods for the multiscale analysis of landscape structure. ISPRS J. Photogramm. Remote Sens. 57, 327–345. http://dx.doi.org/10.1016/S09242716(02)00162-4. Homer, C.G., Aldridge, C.L., Meyer, D.K., Coan, M.J., Bowen, Z.H., 2009. Multiscale sagebrush rangelandhabitat modelling in southwest Wyoming. USGS Open-File Report 2008-1027. pp. 1–14. Hulet, A., Roundy, B.A., Petersen, S.L., Jensen, R.R., Bunting, S.C., 2014. An object-based image analysis of pinyon and juniper woodlands treated to reduce fuels. Environ. Manage. 53, 660–671. http://dx.doi.org/10.1007/s00267-013-0227-1. Jakubowski, M.K., Li, W., Guo, Q., Kelly, M., 2013. Delineating individual trees from lidar data: a comparison of vector- and raster-based segmentation approaches. Remote Sens. 5, 4163–4186. http://dx.doi.org/10.3390/rs5094163. Knick, S.T., Hanser, S.E., Preston, K.L., 2013. Modeling ecological minimum requirements for distribution of greater sage-grouse leks: implications for population connectivity across their western range, U.S.A. Ecol. Evol. 3, 1539–1551. http://dx.doi.org/10. 1002/ece3.557. Kormos, P.R., Marks, D., Pierson, F.B., Williams, C.J., Hardegree, S.P., Havens, S., Hedrick, A., Dates, J.D., Svejcar, T.J., 2017. Ecosystem water availability in juniper versus sagebrush snow-dominated rangelands. Rangel. Ecol. Manage. 70, 116–128. http://dx.doi.org/10.1016/j.rama.2016.05.003. Landis, J.R., Koch, G.G., 1977. The measurement of observer agreement for categorical data. Biometrics 33, 159–174. http://dx.doi.org/10.2307/2529310. Maier, B., Tiede, D., Dorren, L., 2008. Characterizing Mountain Forest Structure Using Landscape Metrics on LiDAR-based Canopy Surface Models. Object-based Image Analysis. Springer, Berlin, Germany, pp. 625–643 doi:10.1007%2F978-3-540-770589_34. Miller, R.F., Tausch, R.J., 2001. The role of fire in juniper and pinyon woodlands: a descriptive analysis. In: Fire Conference. Proceedings of the Invasive Species Workshop: the Role of Fire in the Control and Spread of Invasive Species 27. pp. 15–30. Miller, R.F., Svejcar, T.J., Rose, J.A., 2000. Impacts of western juniper on plant community composition and structure. J. Range Manage. 53, 574–585. http://dx.doi.org/ 10.2307/4003150. Miller, R.F., Bates, J.D., Svejcar, T.J., Pierson, F.B., Eddleman, L.E., 2005. Biology, Ecology, and Management of Western Juniper 152. Oregon State University Technical Bulletin, pp. 1–77. Miller, R.F., Tausch, R.J., McArthur, E.D., Johnson, D.D., Sanderson, S.C., 2008. Age Structure and Expansion of Piñon-juniper Woodlands: a Regional Perspective in the Intermountain West. Rocky Mountain Research Station, pp. 1–15. http://dx.doi.org/ 10.2737/RMRS-RP-69. Report RMRS-RP-69. Miller, R.F., Knick, S.T., Pyke, D.A., Meinke, C.W., Hanser, S.E., Wisdom, M.J., Hild, A.L., 2011. Characteristics and dynamics of greater sage-grouse populations. In: Connelly, John W., Knick, Steven T. (Eds.), Greater Sage-grouse—Ecology and Conservation of a Landscape Species and its Habitats. University of California Press, Berkeley, CA, pp. 145–184. Miller, R.F., Chambers, J.C., Pellant, M., 2014. A Field Guide for Selecting the Most Appropriate Treatment in Sagebrush and Piñon-juniper Ecosystems in the Great Basin—evaluating Resilience to Disturbance and Resistance to Invasive Annual Grasses, and Predicting Vegetation Response. General Technical Report RMRS-GTR322. U.S. Department of Agriculture, Forest Service, Rocky Mountain Research Station, pp. 1–66. http://dx.doi.org/10.2737/RMRS-GTR-322. Miller, R.F., Naugle, D.E., Maestas, J.D., Hagen, C.A., Hall, G., 2017. Special issue: targeted woodland removal to recover at-risk grouse and their sagebrush-steppe and prairie ecosystems. Rangel. Ecol. Manage. 70, 1–8. http://dx.doi.org/10.1016/j. rama.2016.10.004. Mishra, N.B., Crews, K.A., 2014. Mapping vegetation morphology types in a dry savanna ecosystem: integrating hierarchical object-based image analysis with random forest. Int. J. Remote Sens. 35, 1175–1198. http://dx.doi.org/10.1080/01431161.2013.
None. Acknowledgements This research was funded by the Nevada Sagebrush Ecosystem Program with support from the Nevada Department of Wildlife and U.S. Bureau of Land Management. Additional logistical coordination was provided by the U.S. Forest Service and California Department of Fish and Wildlife. We extend gratitude to personnel with U.S. Geological Survey, particularly W. Perry for guidance and expertise, K. Mauch and T. Kroger for assistance in developing feature extraction workflows, and R. Kelble and B. Prochazka for coordinating conifer ground referencing conducted by numerous field technicians. We also thank the Great Basin Bird Observatory for field data collection. Use of trade or product names does not imply endorsement by the U.S. Government. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at https://doi.org/10.1016/j.jag.2018.06.002. References Baruch-Mordo, S., Evans, J.S., Severson, J.P., Naugle, D.E., Maestas, J.D., Kiesecker, J.M., Falkowski, M.J., Hagen, C.A., Reese, K.P., 2013. Saving sage-grouse from the trees: a proactive solution to reducing a key threat to a candidate species. Biol. Conserv. 167, 233–241. http://dx.doi.org/10.1016/j.biocon.2013.08.017. Blaschke, T., 2010. Object based image analysis for remote sensing. ISPRS J. Photogramm. Remote Sens. 65, 2–16. http://dx.doi.org/10.1016/j.isprsjprs.2009.06. 004. BLM, USFS, 2015. Fact Sheet: BLM, USFS Greater Sage-grouse Conservation Effort. Accessed April 2017. Available online. https://www.fs.fed.us/sites/default/files/ fact-sheet-greater-sage-grouse.pdf. Boswell, A., Petersen, S., Roundy, B., Jensen, R., Summers, D., Hulet, A., 2017. Rangeland monitoring using remote sensing: comparison of cover estimates from field measurements and image analysis. AIMS Environmental Science 4, 1–16. http://dx.doi. org/10.3934/environsci.2017.1.1. Bradley, B.A., Fleishman, E., 2008. Relationships between expanding pinyon-juniper cover and topography in the central Great Basin, Nevada. J. Biogeogr. 35, 951–964. http://dx.doi.org/10.1111/j.1365-2699.2007.01847.x. Bruce, D.A., 2008. Object oriented classification: case studies using different image types with different spatial resolutions. Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci. 37, 515–520. Burnett, C., Blaschke, T., 2003. A multi-scale segmentation/object relationship modelling methodology for landscape analysis. Ecol. Model. 168, 233–249. http://dx.doi.org/ 10.1016/S0304-3800(03)00139-X. Chambers, J.C., Bradley, B.A., Brown, C.S., D’Antonio, C., Germino, M.J., Grace, J.B., Hardegree, S.P., Miller, R.F., Pyke, D.A., 2014. Resilience to stress and disturbance, and resistance to Bromus tectorum L. invasion in cold desert shrublands of western North America. Ecosystems 17 (2), 360–375. http://dx.doi.org/10.1016/j.rama. 2016.08.005. Coates, P.S., Casazza, M.L., Ricca, M.A., Brussee, B.E., Blomberg, E.J., Gustafson, K.B., Overton, C.T., Davis, D., Neill, L., Espinosa, S.P., Gardner, S.C., Delehanty, D.J., 2014. Spatially explicit modeling of greater sage-grouse (Centrocercus urophasianus) habitat in Nevada and Northeastern California—a decision-support tool for management. U.S. Geological Survey Open-File Report 2014-1163. http://dx.doi.org/10. 3133/ofr20141163. 84 p. Coates, P.S., Ricca, M.A., Prochazka, B.G., Brooks, M.L., Doherty, K.E., Kroger, T., Blomberg, E.J., Hagen, C.A., Casazza, M.L., 2016a. Wildfire, climate, and invasive grass interactions negatively impact an indicator species by reshaping sagebrush ecosystems. Proc. Natl. Acad. Sci. 113, 12745–12750. http://dx.doi.org/10.1073/ pnas.1606898113. Coates, P.S., Casazza, M.L., Ricca, M.A., Brussee, B.E., Blomberg, E.J., Gustafson, K.B., Overton, C.T., Davis, D.M., Niell, L.E., Espinosa, S.P., Gardner, S.C., Delehanty, D.J., 2016b. Integrating spatially explicit indices of abundance and habitat quality: an applied example for greater sage-grouse management. J. Appl. Ecol. 53, 83–95. http://dx.doi.org/10.1111/1365-2664.12558. Coates, P.S., Casazza, M.L., Brussee, B.E., Ricca, M.A., Gustafson, K.B., Sanchez-Chopitea, E., Mauch, K., Niell, L., Gardner, S., Espinosa, S., Delehanty, D.J., 2016c. Spatially explicit modeling of annual and seasonal habitat for greater sage-grouse (Centrocercus urophasianus) in Nevada and Northeastern California – an updated decision-support tool for management. U.S. Geological Survey Open-File Report 2016-1080. pp. 1–160. Coates, P.S., Prochazka, B.G., Ricca, M.A., Gustafson, K.B., Ziegler, P., Casazza, M.L., 2017. Pinyon and juniper encroachment into sagebrush ecosystems impacts distribution and survival of greater sage-grouse. Rangel. Ecol. Manage. 70, 25–38.
154
Int J Appl Earth Obs Geoinformation 73 (2018) 148–155
K.B. Gustafson et al.
Basin. Rangel. Ecol. Manage. 70, 50–58. http://dx.doi.org/10.1016/j.rama.2016.07. 011. Soulé, P.T., Knapp, P.A., 1999. Western juniper expansion on adjacent disturbed and near-relict sites. J. Range Manage. 52, 525–533. Soulé, P.T., Knapp, P.A., Grissino-Mayer, H.D., 2004. Human agency, environmental drivers, and western juniper establishment during the late holocene. Ecol. Appl. 14, 96–112. http://dx.doi.org/10.1890/02-5300. Strand, E.K., Bunting, S.C., Keefe, R., 2013. Influence of wildland fire along a successional gradient in sagebrush steppe and western juniper woodlands. Rangel. Ecol. Manage. 66 (6), 667–679. Tausch, R.J., Miller, R.F., Roundy, B.A., Chamber, J.C., 2009. Piñon and juniper field guide: asking the right questions to select appropriate management actions. USGS Circ. 1335, 1–95. Tsai, Y.H., Stow, D., Weeks, J., 2011. Comparison of object-based image analysis approaches to mapping new buildings in Accra, Ghana using multi-temporal QuickBird satellite imagery. Int. J. Remote Sens. 3, 2707–2726. Twidwell, D., Allred, B.W., Fuhlendorf, S.D., 2013. National-scale assessment of ecological content in the world’s largest land management framework. Ecosphere 4, 1–27. http://dx.doi.org/10.1890/ES13-00124.1. U.S. Department of the Interior, 2015. Secretarial Order No. 3336 - Rangeland Fire Prevention, Management, and Restoration. Accessed August 2015. Available online. http://www.fws.gov/greatersagegrouse/documents/Threats/20150106DOI-FireOrder.pdf. USDA, 2013. National Agriculture Imagery Program (NAIP) Digital Orthophoto Quarter Quads. Accessed November 2013. Available online. https://gdg.sc.egov.usda.gov/. USDA, 2014. “Targeted conifer Removal: a Proactive Solution to Conserving Sagegrouse.” Conservation Insight Conservation Effects Assessment Project, 1-4. Accessed April 2017. Available online. https://www.nrcs.usda.gov/Internet/FSE_ DOCUMENTS/stelprdb1261131.pdf. USDA NASS, 2014. 2012 Census of Agriculture United States Summary and State Data. 1, 1-695. Accessed April 2017. Available online. . https://www.agcensus.usda.gov/ Publications/2012/Full_Report/Volume_1,_Chapter_1_US/usv1.pdf. USFWS, 2010. Endangered and Threatened Wildlife and Plants; 12-Month Finding on a Petition to List Greater Sage-grouse (Centrocercus Urophasianus) as Threatened or Endangered. Proposed Rule: Federal Register, 75, 13910-14014. Accessed April 2017. Available online. https://www.federalregister.gov/documents/2010/03/23/ 2010-5132/endangered-and-threatened-wildlife-and-plants-12-month-findings-forpetitions-to-list-the-greater. USFWS, 2013. Greater Sage-grouse (Centrocercus Urophasianus) Conservation Objectives: Final Report, 1-92. Accessed April 2017. Available online. . https:// www.fws.gov/greatersagegrouse/documents/COT-Report-with-Dear-InterestedReader-Letter.pdf. USFWS, 2015. Endangered and Threatened Wildlife and Plants; 12-Month Finding on a Petition to List Greater Sage-grouse (Centrocercus Urophasianus) as an Endangered Species. Accessed April 2017. Available online. USFWS, Department of Interior, 80 FR 59857. https://www.federalregister.gov/documents/2015/10/02/2015-24292/ endangered-and-threatened-wildlife-and-plants-12-month-finding-on-a-petition-tolist-greater. Weiner, N.I., Strand, E.K., Bunting, S.C., Smith, A.M.S., 2016. Duff distribution influences fire severity in post-fire vegetation recovery in sagebrush steppe. Ecosystems 19, 1196–1209. http://dx.doi.org/10.1007/s10021-016-9994-x.
876120. Noss, R.F., LaRoe III, E.T., Scott, J.M., 1995. Endangered Ecosystems of the United States: a Preliminary Assessment of Loss and Degradation 28. U.S. Fish and Wildlife Service, pp. 1–95 Biological Report. O’Brien, M.A., 2003. Feature extraction with the VLS feature analyst system. Proc. the Am. Soc. Photogramm. Remote Sens. 1–9. Opitz, D., Blundell, S., 2008. Object Recognition and Image Segmentation: the Feature Analyst Approach. Object-based Image Analysis, Ch. 2.3. Springer, Berlin, Germany, pp. 153–167. http://dx.doi.org/10.1007/978-3-540-77058-9_8. Overwatch, 2013. Feature Analyst Version 5.1.2.0 for ArcGIS. Overwatch Systems Ltd. Pierson, F.B., Williams, C.J., Kormos, P.R., Hardegree, S.P., Clark, P.E., Rau, B.M., 2010. Hydrologic vulnerability of sagebrush steppe following pinyon and juniper encroachment. Rangel. Ecol. Manage. 63, 614–629. http://dx.doi.org/10.2111/REM-D09-00148.1. Poznanovic, A.J., Falkowski, M.J., Maclean, A.L., Smith, A.M.S., Evans, J.S., 2014. An accuracy assessment of tree detection algorithms in juniper woodlands. Photogramm. Eng. Remote Sens. 80, 627–637. http://dx.doi.org/10.14358/PERS.80.7.627. Prochazka, B.G., Coates, P.S., Ricca, M.A., Casazza, M.L., Gustafson, K.B., Hull, J.M., 2017. Encounters with pinyon-juniper influence riskier movements in greater sagegrouse across the great basin. Rangel. Ecol. Manage. 70, 39–49. http://dx.doi.org/10. 1016/j.rama.2016.07.004. Ricca, M.A., Coates, P.S., Gustafson, K.B., Brussee, B.E., Chambers, J.C., Espinosa, S.E., Gardner, S.C., Lisius, S., Ziegler, P., Delehanty, D.J., Casazza, M.L., 2018. A conservation planning tool for greater sage-grouse using indices of species distribution, resilience, and resistance. Ecol. Appl. 28, 878–896. Romme, W.H., Allen, C.D., Bailey, J.D., Baker, W.L., Bestelmeyer, B.T., Brown, P.M., Eisenhart, K.S., Floyd, M.L., Huffman, D.W., Jacobs, B.F., Miller, R.F., Muldavin, E.H., Swetnam, T.W., Tausch, R.J., Weisberg, P.J., 2009. Historical and modern disturbance regimes, stand structures, and landscape dynamics of piñon-juniper vegetation of the western United States. Rangel. Ecol. Manage. 62, 203–222. http://dx. doi.org/10.2111/08-188R1.1. Roth, C.L., Coates, P.S., Gustafson, K.B., Chenaille, M.P., Ricca, M.A., Sanchez-Chopitea, E., and Casazza, M.L. (In Review). A customized framework for regional classification of conifers using automated feature extraction. Manuscript submitted for publication. Roundy, D.B., 2015. Estimating pinyon and juniper cover across Utah using NAIP imagery. All Theses and Dissertations, Paper 5575. pp. 24. http://dx.doi.org/10.3934/ environsci.2016.4.765. Roundy, B.A., Young, K., Cline, N., Hulet, A., Miller, R.F., Tausch, R.J., Chambers, J.C., Rau, B., 2014. Piñon-juniper reduction increases soil water availability of the resource growth pool. Rangel. Ecol. Manage. 67, 495–505. http://dx.doi.org/10.2111/ REM-D-13-00022.1. Sankey, T., Glenn, N., 2011. Landsat-5 TM and lidar fusion for sub-pixel juniper tree cover estimates in a western rangeland. Photogramm. Eng. Remote Sens. 77, 1–8. http:// dx.doi.org/10.14358/PERS.77.12.1241. Schroeder, M.A., Aldridge, C.L., Apa, A.D., Bohne, J.R., Braun, C.E., Bunnell, S.D., Connelly, J.W., Deibert, P.A., Gardner, S.C., Hilliard, M.A., Kobriger, G.D., McAdam, S.M., McCarthy, C.W., McCarthy, J.J., Mitchell, D.L., Rickerson, E.V., Stiver, S.J., 2004. Distribution of sage-grouse in North America. Condor 106, 363–376. http://dx. doi.org/10.1650/7425. Severson, J.P., Hagen, C.A., Maestas, J.D., Naugle, D.E., Forbes, J.T., Reese, K.P., 2017. Short-term response of sage-grouse nesting to conifer removal in the northern Great
155