Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Mapping regional forest management units: a road-based framework in Southeastern Coastal Plain and Piedmont

Mapping regional forest management units: a road-based framework in Southeastern Coastal Plain... Management practices are one of the most important factors affecting forest structure and function. Landowners in southern United States manage forests using appropriately sized areas, to meet management objectives that include economic return, sustainability, and esthetic enjoyment. Road networks spatially designate the socio- environmental elements for the forests, which represented and aggregated as forest management units. Road networks are widely used for managing forests by setting logging roads and firebreaks. We propose that common types of forest management are practiced in road-delineated units that can be determined by remote sensing satellite imagery coupled with crowd-sourced road network datasets. Satellite sensors do not always capture road- caused canopy openings, so it is difficult to delineate ecologically relevant units based only on satellite data. By integrating citizen-based road networks with the National Land Cover Database, we mapped road-delineated management units across the regional landscape and analyzed the size frequency distribution of management units. We found the road-delineated units smaller than 0.5 ha comprised 64% of the number of units, but only 0.98% of the total forest area. We also applied a statistical similarity test (Warren’s Index) to access the equivalency of road-delineated units with forest disturbances by simulating a serious of neutral landscapes. The outputs showed that the whole southeastern U.S. has the probability of road-delineated unit of 0.44 and production forests overlapped significantly with disturbance areas with an average probability of 0.50. Keywords: Forest management unit, Warren’s index, Neutral landscape, OpenStreetMap, Road ecology Introduction management patterns, reflecting long-term land-use leg- The Southeastern United States (SEUS) forest comprises acies (Josephson 1989; Haynes 2002), contribute to the 32% of the total U.S. forestland (Oswalt et al. 2014), complex land mosaic of SEUS. which combined with the productivity of the forest, It is challenging to quantify the ecological and anthropo- places this region at the forefront of American forestry genic mechanisms that control the spatial structure of the production (Fox et al. 2007). This heterogeneous land- forest landscape and its surrounding areas in these com- scape is composed of heavily managed forests, intensive plex forest mosaics. Forest management is the predomin- agriculture, and multiple metropolitan areas. SEUS, al- ant factor in forest ecology and structural patterns though one of the most densely forested regions in the (Becknell et al. 2015), but little is known about how man- United States (Hanson 2010), is also heavily dissected by agement practices are related to surrounding land-use at road networks (Coffin 2007). The diverse forest the regional scale. One thing that is known is that, in the SEUS, significant expansions of urban areas tend to con- * Correspondence: dyang1@uwyo.edu vert forested land to urban uses and that croplands tend Wyoming Geographic Information Science Center, University of Wyoming, to transition to pine plantations (Davis et al. 2006;Haynes 1000E. University Ave, Laramie, Wyoming 82071, USA 2002; Wear and Greis 2002, 2012, 2013;Stanturf et al. Department of Geography, University of Florida, 330 Newell Dr, Gainesville, Florida 32611, USA © The Author(s). 2021 Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. Yang and Fu Forest Ecosystems (2021) 8:17 Page 2 of 17 2003; Becknell et al. 2015). To understand the eco- Faccoli and Bernardinelli 2014), change water yield logical and anthropogenic influences of differently and hydrologic regulation (Douglass 1983), produce managed forests on ecosystem processes, all land- wood products, provide places for hunting and recre- scapes should be understood at multiple scales, from ation, and conserve habitat biodiversity. the local scale (forest management unit) to regional Forestland structures, functions, and ecological pro- scales, with the regional scale referring to broad forest cesses are scale-dependent (Battaglia and Sands 1998; mosaics that are formed from management patches Niemelä 1999; Drever et al. 2006). Regionally, for the (O’Neill et al. 1996). Understanding forest manage- purpose of sustainable forest management, we need to ment spatial patterns require defining a map-based develop criteria and indicators of management units. management unit, which is the subdivision regarding Forest management units in this study are zones or effects of land use on forest ecosystems. This research patches, which can be identified, mapped and managed seeks to further the understanding of how spatial pat- according to the land-use objectives. Road networks link terns of forest management affect land use, by posing human activities (e.g. management practices) and sur- the following questions: Do roads delineate forest rounding physical environments (land cover). For pro- management units? What is the spatial distribution of duction, preservation, and ecological forestry, in many road-defined forest management units in the SEUS? cases, forest management units are harvest or burn Moreover, how are the distributions of management units. Roads are built to create access for the managers units affected by different forest management ap- and harvesters. For example, the preserved forest in the proaches? And how does forest management affect Ordway Swisher Biological Station (OSBS) in northcen- nearby land use, and how does nearby land use affect tral Florida is subdivided by road networks into manage- forest management? ment (burn) units, which is the smallest unit of land that Forest management is the main driving force of for- is actively managed (Ordway-Swisher Biological Station est structure in SEUS (Becknell et al. 2015)and alters 2015). In the Joseph W. Jones Ecological Research Cen- forest properties and processes, which affects forest ter in Ichauway, Georgia, and OSBS, the internal road ecosystem services (Kurz et al. 2008;Stephensetal. network provides access to the research site and serves 2012;Oswalt etal. 2014). Forest management can be as prescribed fire breaks (Ordway-Swisher Biological classified into four categories: production forestry, Station 2015). For passive management practices, there ecological forestry, passive management and preserva- are currently no clear criteria defining the management tion management (Becknell et al. 2015). Production units. However, in national forest systems (mostly with management harvests forest products and sustains the ecologically managed forests and multi-use production bio-productivity of the system with the sole objective forests), existing roads and trails are used for controlling of producing wood, pulp, and other forest products. prescribed fire and wildfires (e.g. Apalachicola National In SEUS, production management based on silvicul- Forest, Osceola National Forest, United States Depart- ture systems, which homogenize parts of the land- ment of Agriculture Forest Service 1999). For an ex- scape, has predominated the SEUS (Siry 2002). ample of a privately-owned forest, the Red Hills region However, forest conservation systems have evolved in Georgia uses roads to delineate burn units ranging ap- considerably over recent decades (Mitchener and proximately from five to 30 ha (Robertson and Ostertag Parker 2005; Franklin et al. 2007). Ecological manage- 2007). ment uses legacies of disturbance, including inter- Road networks facilitate movements of humans and mediate stand disturbance processes such as variable connect natural resources with societies and economies. density thinning and fire, and variable and appropri- As conduits for human access to nature, the physical ate recovery times to manage forests that still pro- footprint of approximately 6.6 million km of roads in the duce economically valuable wood products while United States FHWA (Federal Highway Administration preserving many of the values of natural forests 2013) has significant primary and secondary impacts on (Franklin et al. 2007). Passive management is defined ecosystems and the distribution of species (Bennett as a practice with little or no active management. We 1991). Fifteen to 20 % of American land is subject to the arguethatall forestsare managedtosomedegree ecological effects of road networks (Forman and and that doing nothing is a form of management. Alexander 1998; Forman and Deblinger 2000). The most Preservation forestry aims to minimize the ecological noticeable effects of road networks on forest structures footprint of society with the objectives of protecting are landscape structure changes, including reduced wildlife and maintaining ecosystem services. Further- mean patch size, increased patch shape complexity, in- more, certain forest management practices can sup- creased edge densities, and reduced unit connectivity. In press wildfires (Waldrop et al. 1992), prevent insect one case, McGarigal et al. (2001) investigated the land- /pathogen outbreaks (Netherer and Nopp-Mayr 2005; scape structure changes of the San Juan Mountains from Yang and Fu Forest Ecosystems (2021) 8:17 Page 3 of 17 1950 to 1993 and found that roads had a more significant effects are not yet well developed (Ries et al. 2004; Hou ecological impact (e.g. core forest areas and patch sizes de- et al. 2013). Roads and streams may be challenging to crease) on landscape structure than logging activities. identify, or invisible because they do not open the can- In addition to management practices (e.g., harvesting, opy so that many roads are not detectable on satellite fertilizing), the construction of road networks divides imagery or even aerial photography. The reliability of the forested land into smaller patches, thereby increasing large-scale road data is also challenged due to issues of the potential intensity of the effects of management accuracy, coverage and immediacy, all of which can practices. Road networks in managed forests provide underestimate the extent and ecological impacts of roads easy access for managers and harvesters to extract and on forest structures (Riitters et al. 2004). We propose regenerate resources (Demir and Hasdemir 2005). Roads that common types of forest management are practiced may influence fire regimes by increasing fire ignition as in road-delineated units that are detectable by remote a result of human activities (Franklin and Forman 1987). sensing satellite images coupled with crowd-sourced Moreover, road networks alter the spatial configuration road network datasets. We also describe and study the of management patches by functioning as firebreaks, patterns of forest management units in response to land which form new patterns in landscapes (Franklin and ownership and different management practices. Forman 1987; Nelson and Finn 1991; Eker and Coban 2010). By quantifying the spatial patterns of manage- Study area ment units created by roads we may gain insight into We focus on the Southeastern U.S. Coastal Plain and the ecological effects of road networks on spatial forest Piedmont (SEUS) region (Fig. 1). The SEUS is located structures within differently managed areas. between Piedmont to the north and the Atlantic Ocean The impacts and ecological effects of roads on the to the east and covers a significant portion of the south- landscape might be misestimated because methods eastern United States. The SEUS is the home to the measuring the road-effect zones and landscape scale most densely production-forested region nationwide, Fig. 1 Study area of Southeastern United States Coastal Plain and Piedmont, with level III ecoregions (Bailey 2004) Yang and Fu Forest Ecosystems (2021) 8:17 Page 4 of 17 which makes up 32% of total U. S forest cover (Oswalt GeoTIFF rasters with a 250-m resolution from PANGAEA et al. 2014). Based on EPA eco-region descriptions, land (Marsik et al. 2017). cover in the SEUS is a mosaic of cropland, pasture, woodland and forests (Bailey 2004). Major silvicultural Road networks forests in SEUS are pine forests, such as slash pine We selected OpenStreetMap as the primary road data (Pinus elliottii Engelm.) and loblolly pine (Pinus taeda and the USDA National Forest service trail and road L.) forests. European settlement and the extensive har- maps (https://data.fs.usda.gov/geodata/edw/datasets. vesting in the early 1900s removed 98% of the original php?dsetCategory=transportation, accessed Dec 2016) as longleaf pine (Pinus palustris Mill.) forests, which was secondary in this study. OpenStreetMap is a collabora- one of the most dominant ecosystems in SEUS (Outcalt tive, crowdsourced project that creates free, open, and 2000) and converted them to plantations of native slash accessible maps of road networks. OpenStreetMap is pine. The SEUS forest system is a fire-dominated system one of the most popular and well-supported Volun- with native trees adapted to short-period stand-clearing teered Geographic Information (VGI) datasets (Mooney events. The primary forest management types are pro- et al. 2010). Community volunteers collect geographic duction and passive management due to the dominant information and submit it to the global OpenStreetMap ownership of private owners, logging companies, and in- database (Ciepluch et al. 2009). OpenStreetMap moni- vestment institutions (Real Estate Investment Trusts tors road networks at near real-time and includes add- (REITs) and Timber Investment Management Organiza- itional classes of roads such as private access roads and tions (TIMOs)) (Zhang et al. 2012). This diversity of driveways in rural areas, small service roads or alleys in land cover types is spatially heterogenous, and patch urban areas, and forest access roads. All those road fea- sizes of the numerous vegetation classes vary across a tures are critical for this research and no distinctions wide range of scales (Fig. 1). were drawn between the types of road, traffic volume, or other factors. OpenStreetMap shows up-to-date road Data sources networks information, which the other official road data- Forest extent bases do not offer. The accuracy of the OpenStreetMap In this study, forest extent is determined by a composite in our study region has been studied. For the state of of the 2006 and 2011 USGS National Land Cover Data- Florida, Zielstra and Hochmair (2011) compared road base (NLCD), which was constructed from Landsat im- networks dataset from different sources and concluded agery at 30-m spatial resolution (Jin et al. 2013). We that OpenStreetMap was significantly better the other aggregated 21 NLCD classes into two classes: forest (de- road databases. All OpenStreetMap data were down- ciduous, evergreen, mixed forest and woody wetland) loaded from the website of Geofabrik (http://download. and non-forest (including water). Only the pixels that Geofabrik.de, Accessed Dec 2016). USDA National For- contain 50% or more forest area in NLCD will be con- est Services Trails and Road Map provide the coverage sidered as forested pixels. We also extracted the SEUS of detailed transportation map in National Forests urban areas by using the most recent 2015 US Census (Coghlan and Sowa 1997). Bureau’s TIGER cartographic boundary urban areas Road density in SEUS was measured as the total length (TIGER 2015) dataset to remove urban areas from the of all roads (in kilometers) in a district divided by the analysis. total land coverage area of the district (km ) based on our developed road networks map (Figure S2). Forest management type An integrated random forest classifier was built from the Landscape fire and resource management planning tools analysis of long-term phenological features derived from (LANDFIRE) BFAST outputs and spectral entropy calculated from the Disturbance data from the Landscape Fire and Resource Terra-MODIS enhanced vegetation index (EVI: MOD13Q Management Planning Tools (LANDFIRE) disturbance data product), along with ancillary data such as land own- database were used to evaluate the management unit ership, and disturbance history to classify different forest map. LANDFIRE is a combination of Landsat images, management types (Breiman 2001; Verbesselt et al. 2010; fire program data, and cooperator-provided field data Zaccarelli et al. 2013). The forest management type map and other ancillary databases (e.g., PAD-US), and is a has a spatial grain of 250 m and is a composite of pheno- shared program between the wildland fire management logical patterns and changes in the patterns from February programs of the U.S. Department of Agriculture Forest 2001 through December 2016 (Figure S1). The SEUS for- Services and U.S. Department of the Interior (Rollins est management type map has an overall accuracy of 89% 2009). LANDFIRE also describes land cover/use both for a 10-fold cross-validation. The forest management spatially and temporally from 1999 to 2014 and provides raster is available for each region as georeferenced the existing vegetation composition map based on Yang and Fu Forest Ecosystems (2021) 8:17 Page 5 of 17 dominant species or group of dominant species. Road-delineated forests units were mapped on the for- Spatially, LANDFIRE is a Landsat-based (30 m) database, est extent map from the section of Forest Extent after which matches the 30-m spatial resolution of this study superimposing detailed road networks with the Region (https://www.landfire.gov/disturbance.php). We chose Group tool to identify forest clusters as units. When the LANDFIRE project data because the spatial scale is superimposing road maps, all road networks were con- small enough to detect subtle changes brought about by verted to one-pixel segments. After one-pixel wide road land management practices, and large enough to reflect segments were derived, we converted all the forest pixels the characteristic variability of essential ecological pro- to 1 and the pixels that contained at least one road seg- cesses (such as wildfire) in the appropriate spatial con- ment to non-forest pixels (30-m spatial resolution) to 0. text. The disturbance data from LANDFIRE will be used to evaluate management units delineated by the road Geospatial assessment network in the context of intensely managed forests in In SEUS, regional forest management activities were rep- SEUS (Figure S3). resented as disturbances as described by LANDFIRE data, such as clear cuts, fires, and thinning. Figure 2 Forest ownership shows the example view of LANDFIRE cumulative dis- Geospatial land-ownership data from federal and non- turbance with delineated forest extent, and it can be governmental agencies were integrated for land owner- clearly observed (with the Warren’s Index of 0.62) that ship mapping (Figure S4). Forest ownership in SEUS is the disturbed areas and delineated unit shared boundar- broadly categorized as publicly owned and privately ies and show a large degree of equivalence. The spatial owned according to the landowners. There are six sub- coincidence has been shown to facilitate the interpret- types of public ownership, which are federally protected, ation and integration of defining regional forest manage- federal, state protected, state, military, and local. Also, ment units. In this study, we propose a test based on there are four sub-types of private ownership: non- forest management unit and the geographic correspond- governmental organization, private, family, and corpor- ing forest disturbances to compare the geographical ate. The ownership classification implies different man- similarity between the management unit and corre- agement objectives, as well as landowner skills, budgets sponding forest disturbances. The assessment of geo- and interests. Datasets from other federal and state gov- graphic image overlap is analogous to quantifying the ernment agencies were regrouped and classified into ten niche similarity of two species in two dimensions. As sub-types to create a comprehensive dataset that in- two-dimensional rasters, both disturbance and forest ex- cludes public land ownership and privately protected tent data can be treated as homogenous and spatial- easements as well as specially designated areas and asso- explicit datasets. ciated protection level (see Table S1). The final product Testing the overlap between pairs of road- is a 250-m spatial resolution raster data depicting the delineated forest extent with disturbance regions was forest ownership types and resampled to 30-m in this compared using the similarity statistics of Warren’s study to match the spatial resolution of the NLCD Index (Warren et al. 2008). The values of Warren’s database. Index range from 0 to 1. The value of 0 means forest management units have no spatial overlap with forest Methods disturbance areas, and 1 means all forest management Mapping road-delineated units units are identical to disturbance areas. The statistics In each forest management type, the fundamental elem- of Warren’s Index assume probability distribution ent of management practice is the management unit. In defined over geographic space, in which the p (or this study, we define the individual forest clusters that X, i p ) denotes the spatial probability distribution of X delineated by road-networks as “management units” and Y, i (road-delineated forests compartments), or Y (prob- the clusters that directly derived from forest extent map ability distribution of forest disturbances) to cell i.In as “management patches.” We hereby developed two nature, Warren’s equivalency index carries no bio- comparative methods for landscape analysis by using logical assumptions concerning the parameters as be- two sets of input data, with and without incorporating ing from any probability distributions. Spatially, we OpenStreetMap. For the method without incorporating applied it into assessing road-delineated forest man- OpenStreetMap, management patches were mapped on agement units versus the areas of corresponding the forest extent map resulting from the map described disturbances. in the section of Forest Extent and the Region Group Firstly, the Hellinger distance was calculated to meas- tool in ESRI ArcGIS 10.X (ESRI Inc.,) to identify clusters ure and compare the probability distance (Van der Vaart of forest pixels that formed unique and unconnected 1998): forests. Yang and Fu Forest Ecosystems (2021) 8:17 Page 6 of 17 Fig. 2 Spatial visualization of road-delineated forest extent (road delineated units - yellow colored polygon) matching with LANDFIRE disturbance data. Warren’s Index for this sub-landscape is 0.62 sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 1) Spatially. In this study, we used the Worldwide Ref- pffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffi erence System 2 (WRS-2) row 17 path 39 (17_39 here- HpðÞ ; p ¼ p − p ð1Þ X Y X ;i Y ;i after), which is one scene of Landsat Thematic Mapper (TM). As shown in Fig. 3, the forests in 17_39 are heav- ily disturbed and fragmented, but it also shared the while the Warren’s statistic is: larges contiguous forests in SEUS (Okefenokee National Refuge). This heterogeneous landscape consists of a mix- IpðÞ ; p ¼ 1 − HpðÞ ; p ð2Þ X Y X Y ture of natural and plantation forests, urban centers, urban and rural residential areas, and commercial and To measure and quantify the similarity of differently small-scale agricultural operations. Instead of using managed management units with the corresponding dis- LANDFIRE disturbance data, we simulated a neutral dis- turbance areas, all the units were classified into four turbance layer by mimicking the disturbance patch groups (ecological, passive, preservation and production shape and area called “Random NLM” (Saura and management). A 10 km × 10 km grid was utilized to Martinez-Millán 2000; Sciaini et al. 2018). 17_39 was compute Warren’s Index over the SEUS. The inputs of divided into 325 individual landscapes with the spatial this iterative Warren’s Index spatial analysis are: 1) resolution of 30 m (10 km × 10 km in size), which means stacked disturbance area data from LANDFIRE and 2) each landscape in 17_39 contains 333 × 333 pixels corresponding road-delineated forest management units. (10,000 ha). We applied the neutral landscape generation These analyses were carried out using the nivheOverlap algorithm in each landscape in 17_39 for 325 times and function of dismo package in R 3.3.X (Hijmans et al. recalculated the Warren’s Index for each 10 km × 10 km 2015). landscape with the simulated disturbance layer with the We further tested the hypothesis of road delineated road networks delineated forest extent map. forest management units by generating neutral land- 2) Iteratively. We applied a range-based attribute filter scape models from two perspective as: to the SEUS road density map (Figure S2). The road Yang and Fu Forest Ecosystems (2021) 8:17 Page 7 of 17 Fig. 3 Case study region of Worldwide Reference System 2 (WRSII) row 17 path 39 (17_39). As the LANDFIRE EVT product shows, it encompasses a diversity of landcover and disturbance types. Much of the conifer and conifer-hardwood land cover found outside of the large riparian area of the Okefenokee Swamp (upper central) are heavily managed, privately owned plantations and mixed agriculture/timber land use − 2 densities range from 0 to 91.48 km∙km , so we used the By replotting the spatial probability of road networks interval of 5 to randomly select a serious spatial noncor- delineating management units in 17_39 (Fig. 6c), we related 10 km × 10 km grids with a total number of 19 recalculated the histogram of how the Warren’s Index landscapes in SEUS (Fig. 4). For each spot, we iteratively distributed over 17_39 region after replacing the LAND run the Random NLM algorithm for 500 times and cal- FIRE data with simulated disturbance layer, as shown in culated the Warren’s Index with overlaying the simu- Fig. 6d. The mean value of the probability distribution is lated disturbance resulted from NLM with the road 0.14 with the standard deviation of 0.082. The results networks delineated forest extent map. show that human-derived forest disturbances can result in a non-random association with forest extent. By com- Results paring with the LANDFIRE disturbance derived War- Spatial assessment rans’ I map (Fig. 6a), road networks make a great We calculated the Warren’s Index with overlaying the contribution in identifying and shaping forest patterns in road networks delineated forest extent with LANDFIRE the case study area 17_39, wherein SEUS reflect the disturbance map on a 10 km × 10 km grid in SEUS management practices on the ground, and so for the (Fig. 5). The total number of 10 km × 10 km grids is road-delineated forest compartments, we hereby call 11,072. Figure 5 shows the probability distribution of them forest management units. − 2 road network delineating management unit, with an Road density in SEUS ranges from 0 to 91 km∙km . average of 0.44 and the standard deviation is 0.28. We Table 1 lists the info of all 19 10 km × 10 km grids with a also computed the distribution and found that most for- variation of road densities, the mean values of 500 simu- ested areas have the Warren’s Index from 0.3 to 0.5. lated landscapes. We compared the original Warren’s When we extracted 17_39 (Fig. 6a), shown as Fig. 6b Index of all 19 landscapes with the mean values calcu- with the mean value of 0.54 and the standard deviation lated from simulated landscapes, and found all the simu- of 0.17. lated landscape show significant difference on each Yang and Fu Forest Ecosystems (2021) 8:17 Page 8 of 17 − 2 Fig. 4 Spatial distribution of 19 points with 5 km∙km intervals landscape with the original Warren’s Index. We also plot represented the diversity and heterogeneity of road- the histogram of the landscapes to see how the simu- delineated forests clusters. lated neutral landscape distributed (Fig. 7). A linear rela- As the unit size increases, the number of units de- tionship between the Warren’s Index overlaying creases in an approximately logarithmic manner (Fig. 8). simulated disturbances and roads with the Warren’s Overall, 177,400 units occupied the SEUS with the mean Index overlaying LANDFIRE disturbances with roads size of 29.5 ha. We found that 35.94% of the forest man- (Figure S5) with R of 0.4755. It shows the evidence that agement units represent more than 99% of the whole roads help to shape forest patterns by statistical test and forest and ranged in area from 0.5 to 172,886 ha (the approve the hypnosis that road delineates forest manage- Okefenokee National Wildlife Refuge). The remaining ment patterns. 64.06% of small forest compartments, which are smaller than 0.5 ha, covers only 0.98% of the forested area. The SEUS forests is dominated by management unit Forest management unit map size ranging from 100 to 10,000 ha. The forest manage- A total area of 5.24 × 10 ha managed forest was mea- ment unit-size class map was reclassified based on the sured in the SEUS from the forest extent map based on forest management unit sizes (Fig. 9). From Fig. 9, it can NLCD. When we compared the NLCD derived forest/ be clearly seen that riparian forests stand out as large, non-forest maps between 2006 and 2011, a strong ten- unbroken linear features (most of the orange colored dency of deforestation was found on our study area units), which cover average unit sizes from 10,000 to (1.7 × 10 ha). The total length of the roads in SEUS is 100,000 ha. A characteristic feature for the southeastern 2.26 million km, which was calculated based on the forest is that relatively small and large management number of rasterized road network multiply the spatial units locate close to each other, surrounded with small- resolution (30 m). The road densities in SEUS range sized units (Figs. 9 and 10). Figure 10 shows five close- − 2 from zero to 49.29 km∙km (Figure S2). We mapped ups/examples from multiple locations and landscapes. In the forest management units at regional scales, and the our study area, the largest management unit is an area spatial distribution of those management units of the Okefenokee National Wildlife Refuge (Georgia Yang and Fu Forest Ecosystems (2021) 8:17 Page 9 of 17 Fig. 5 The probability of road networks delineating management unit in SEUS. The spatial resolution of this map is 10 km, and the variable is the value of Warren’s Index calculated for the overlap of road-defined areas and disturbance areas from the LANDFIRE database and Florida) with the size of 172,886 ha (Figs. 9 and 10d), networks) with the histogram of the management followed by the Atchafalaya River basin in Louisiana at patches (without incorporating road networks. The red 102,367 ha (Fig. 9 and 10c). Riparian forests are often histogram shows the size-based frequency distribution of undisturbed in SEUS without much road access because patches without incorporating road networks and the their soils are too wet for harvesting machinery and the purple histogram illustrates the size frequency of forest trees are not as valuable. There is also a cluster of large management units. When refining the management forest management units (yellow: 1000–10,000 ha) along units with road networks, there are 17 times more the Piedmont ecoregion, and the middle part of the patches (management patches) compared with the map southeastern plains. The Atchafalaya River basin in without incorporating road networks (management southern Louisiana is full of canals and river channels units). but no roads. There are many smaller units in Alabama and Mississippi and central Louisiana due to the rela- Discussion tively high road density in this region, so the landscape Management units under different management is broken up according to the road density map. Another approaches example is Fig. 10e, Great Dismal Swamp National Road density (Figure S2) has been proposed as a broad Wildlife Refuge, is the largest intact forest across south- index of roads’ ecological effects in a landscape (Forman eastern Virginia and northeastern North Carolina and and Alexander 1998). The magnitude of average road − 2 was established for protecting and managing the densities ranges from 0.63 to 2.2 km∙km in all man- swamp’s ecosystem (USFWS 2006). aged forests in SEUS. Preservation forest holds the low- − 2 To illustrate the contributions of road networks to our est average road density of 0.63 km∙km , and the regional SEUS management unit map, a representative passively managed forest has the highest average road − 2 comparison set was done with and without incorporat- density of 2.2 km∙km . Ecological forest and production ing road networks data. In Fig. 11, we overlay the histo- forest have the average road density of 1.86 and 1.29 − 2 gram of the management unit (incorporating road km∙km , respectively. One reason is the building of Yang and Fu Forest Ecosystems (2021) 8:17 Page 10 of 17 Fig. 6 The spatial distribution map of Warren’s Index overlaying the road network delineated forested extent with neutral simulated disturbance map over 17_39 area (a). (b) is the one of 10 km × 10 km simulated landscapes, (c) is the distribution of the warrens’ I with the average value of 0.14 and the standard deviation of 0.08. The sub-figure shows the randomly simulated patches in a plot of 10 km by 10 km forest service roads and private logging roads, which ob- management units under different functional forest viously increase the road density in production forest- management types. For ecological management, the spe- land. Road density is a predictor of forest management cific practice is designed to emulate the outcome of nat- intensity (Wendland et al. 2011), and the indicator of ural disturbance, which is to create an uneven-age stand human interactions with forests (Forman et al. 2003). structure to manage competition between and within We compared the size-frequency distributions of man- multi-cohort stands. The distribution of ecological man- agement units with a map of different kinds of manage- agement units shows spatial heterogeneity with structur- ment (production, ecological, preservation, and passive ally complex stands. For passive management forest management) derived independently. Preservation and lands, as the passively managed forests mostly adopt production management had the largest patches, with many irregular shapes with blurred boundaries, and rup- means of 109.6 and 82.6 ha, respectively. Ecological and ture of connectivity. For preservation management for- passively managed units averaged about half as large as est: mostly large government-managed land for 73.8 and 73.0 ha, respectively (Table 2). multiple-purpose including watershed, wildlife, recre- We incorporated Warren’s Index to assess quantita- ation and wilderness aspects. Accordingly, various prac- tively the geographic overlap between forest tices may be applied to it such as harvest, cutting, Yang and Fu Forest Ecosystems (2021) 8:17 Page 11 of 17 Table 1 Spatial assessment covering the different road densities (km/km ) ID Road density Mean of simulated landscapes Standard error of simulated landscapes Warren's I 1 0 0.16 0.008 0.41 2 30.23 0.06 0.009 0.10 3 35.15 0.12 0.008 0.66 4 15.03 0.10 0.009 0.25 5 40.63982391 0.02 0.008 0.20 6 4.9943223 0.26 0.006 0.52 7 85.59053802 0.19 0.007 0.49 8 49.92194366 0.23 0.007 0.55 9 24.98642921 0.25 0.006 0.47 10 9.908494949 0.13 0.009 0.30 11 65.24546814 0.10 0.008 0.38 12 75.59457397 0.10 0.008 0.22 13 19.80970001 0.24 0.007 0.66 14 59.10257721 0.02 0.007 0.31 15 45.15394592 0.11 0.008 0.33 16 55.00973892 0.12 0.008 0.23 17 68.21031952 0.17 0.008 0.34 18 79.92221832 0.03 0.008 0.34 19 91.88208771 0.11 0.008 0.24 means significant different with a 95% confidence interval Fig. 7 Distributions of 19 simulated landscapes of SEUS (Numbers are corresponding with Fig. 4) Yang and Fu Forest Ecosystems (2021) 8:17 Page 12 of 17 Fig. 8 The spatial distribution of SEUS forest management units based on unit size retention cutting, thinning and prescribed fire. As shown road-delineated management units I = 0.33. As the dom- in Table 3, Warren’s Index represents the overlap be- inant forest: production forests, the average size of plan- tween forest management units under different manage- tation management units tends to be large, have a ment types with the corresponding forest disturbance uniform composition (Figure S9), are internally homoge- area. neous and involve practice such as clearcutting and thin- The 10 km × 10 km based spatial grid analysis of War- ning. For preservation/wilderness, across the whole ren’s Index is shown in Fig. 5 and Figure S6,S7,S8 and SEUS, the probability of road-delineated management S9. Among all four forest management types, production unit is 0.44, with the standard deviation of 0.29, because forestry showed the highest probability of road- the large areas of wildfires happened but with road net- delineated management units with I = 0.50, and the pas- work setting as firebreaks. Roads provide access and fire- sive managed forests showed the lowest probability of breaks, as the use of prescribed fires is widespread in Fig. 9 Logarithmic size-based frequency distribution of SEUS management unit. The blue dotted line indicates the mean value of the management units within the SEUS Yang and Fu Forest Ecosystems (2021) 8:17 Page 13 of 17 Fig. 10 Examples of forest management unit distributions Fig. 11 Comparison between with and without incorporating OpenStreetMap on unit size frequency distribution Yang and Fu Forest Ecosystems (2021) 8:17 Page 14 of 17 Table 2 Average management unit size in each forest management type Average management unit size Standard error Average road density Standard error −2 (ha) (Size) (km∙km ) (Density) Ecological management 73.80 0.124 1.86 0.0017 Passive management 73.0 0.037 2.20 0.0005 Preservation 109.6 0.079 0.63 0.0012 management Production management 82.6 0.015 1.29 0.0052 SEUS and much of the prescribed fires are on private addition, as the road networks increased, the number of lands (Haines et al. 2001). In this study, we used a small-sized parcels also shows a substantial increase. threshold of 50% of the similarity score although many The mean forest management units under different useful criteria for establishing such thresholds have been ownerships range from 73.2 (privately owned forests) to proposed (Jimenez-Valverde and Lobo 2007). In the ana- 115.9 ha (state protected forests) with the standard devi- lysis of Table 3, the criteria were just used to show bin- ation of 33.4 ha. By comparing with Table 2, we found ary predictions of four differently managed forests. that the average unit size of military land is 74.9 ha, which is the closest to the average management unit size of ecological forestry at 73.8 ha. The land with the own- Ownership representation of forest management unit ership of state protected represents the largest average Forest management activities are important links be- management unit at 115.9 ha. tween human and environmental factors, especially at regional scale. Forest ownership patterns also explain Conclusions different types of land management practices and trajec- Quantifying forest management units under different tories of land cover change (Turner et al. 1996). The aim management approaches is a key step to ensure that ap- of this part of the research is to link regional land own- propriate management practices and policies are in place ership to management. We produced the SEUS owner- to maintain the array of forest ecosystem services. Re- ship database (Figure S4; Table S1) by collecting the gionally, forest harvesting operations are conducted data from different sources, where the ownership was di- within road-defined boundaries. Forest management vided as Private and Public. Based on our understanding practices in SEUS are specifically the activities primarily of the SEUS forest ownership, we reclassified the forest dictated by forest harvest and the needs of management owner types into public and private with 10 sub-classes for recreation and sustainability. Roads networks in (Table 4): 1) Public: Federal protected, federal, state pro- SEUS are often used for setting firebreaks and timber tected, state, military, local and NGO lands; and 2) Pri- harvest boundaries, so there should be a spatial coinci- vate: Private, family and corporate forests. Our dence between road-delineated management units and ownership data indicate that 18.7% of the landowners disturbance. In this study, we assessed the forest man- are under public forest and 81.3% of private forest land- agement from stands to regional scale, by incorporating owners, which covers 41.5% and 58.5% total forestland. road networks and multi-temporal disturbance remote We can see that the type of land ownership affects sensing database. units’ distribution (Table 4). The special characteristics A number of conclusions can be drawn from the ana- of SEUS forest ownership patterns can result in strong lysis in this study. contrasts in management unit distribution. The major ownership types in the region are family, corporate and 1) Road networks play a role in delineating forests state, which have different management objectives. In from local to regional scale. By defining the individual forest clusters delineated by road Table 3 Warren’s Index between different managed units and networks as “forest management units” and the corresponding disturbances overlap with stacked LANDFIRE clusters that directly derived from forest extent map disturbance as “management patches”, we mapped the forest Management type Warrens I Standard Dev extent map of both “units” and “patches” and compared them with treating “patches” map as Ecological 0.38 0.26 background. There were 17 times more “units” than Passive 0.38 0.28 “patches” over the whole SEUS. And we also Preservation 0.44 0.29 summarized the size distribution road delineated Production 0.50 0.24 units, with units smaller than 0.5 ha comprised 64% Yang and Fu Forest Ecosystems (2021) 8:17 Page 15 of 17 Table 4 Forest mean unit sizes based on different ownership types in SEUS Owner Class Owner Type Mean Unit Area (ha) Stdev Area (ha) Public Federal Protected 95.4 54.0 1,291,925 Public Federal 89.7 32.9 2,750,912 Public State Protected 112.6 33.4 2,793,425 Public State 96.8 52.1 1,277,169 Public Military 73.4 40.7 740,831 Public Local 89.4 45.5 1,635,244 Public NGO 93.6 42.9 130,331 Private Private 73.6 42.9 11,695,831 Private Family 72.9 36.3 23,247,681 Private Corporate 86.9 35.8 13,628,388 of the counts of units, these small units altogether management perspective, more road landscape area covered only 0.98% of the total forest area. leads to less available land for trees at the 2) We quantitatively tested the probability distribution macrosystems scale. On the other hand, logging patterns by using Warren’s Index of road-delineated roads and trails are an efficient way to manage management units and the corresponding forest dis- forests, which we can see the relatively high road turbances area. The average probability of road- network densities in ecological and production delineated management units is 0.44, and we also forestry. visualized the probabilities by setting a 10 km × 10 km grid. In SEUS, the high equivalency between the This study represents benefits so society in that future road-delineated units and the corresponding areas management decisions can be evaluated cross scales, were found at most production forests, and large- taking account of both climate and disturbance regimes. size preserved areas (e.g. Okefenokee National More information on the effects of land ownership and Wildlife Preserve and St. Marks National Wildlife forest management, combined with the detailed road Refuge). network and a continental coverage land cover maps can 3) The combination of remote sensing data and aid in thwarting further forest fragmentation by promot- OpenStreetMap constitutes a useful tool to ing more reasonable road planning by land planners and monitor, characterize and quantify land cover and decision makers. management unit distributions at macrosystems scale. By using the NLCD as the forest reference Supplementary Information The online version contains supplementary material available at https://doi. data and OpenStreetMap as road networks dataset, org/10.1186/s40663-021-00289-w. we produced the OpenStreetMap refined management unit pattern map and analyzed the Additional file 1. spatial size distribution of forest patterns. In addition, by incorporating the OpenStreetMap, the Abbreviations roads are shown to play an important role in GEE: Google earth engine; LANDFIRE: Landscape fire and resource causing fragmentation of the remnant forestlands. management planning tools; MODIS: Moderate resolution imaging spectroradiometer; NLCD: National land cover database; NLM: Neutral The size frequency distribution tells us that all of landscape model; OSM: OpenStreetMap; SEUS: Southeastern U.S. Coastal the 64% of management units are small Plain and Piedmont; VGI: Volunteered geographic information management units (< 0.5 ha) making up just 0.98% Acknowledgments forestlands. Acknowledgement is made of the assistance of Dr. Michael Binford and Dr. 4) Our land ownership product indicates that 18.7% of Peter Waylen, Department of Geography, University of Florida, for English public forest and 81.3% of private and industrial writing and reviewing, for suggestions and discussion. Funding for this research was provided by the National Science Foundation Macrosystems forestland owners, cover 41.5% and 58.5% area of Biology Program Grant EF #1241860. total forestland, respectively. Management practices affected units are represented not only at the stand Authors’ contributions or local scale, but also will change the forest pattern YD and FC conceptualized the idea for the study, FC generated the SEUS forest ownership database; YD performed data analysis and led the writing dramatically at the regional scale. We provided of the manuscript; FC critically reviewed the data analysis, and contributed substantial evidence that road networks occupy a substantially to the writing. Both authors read and approved the final substantial proportion of forest. From a forest manuscript. Yang and Fu Forest Ecosystems (2021) 8:17 Page 16 of 17 Funding Newtown Square, PA: U.S. Department of Agriculture, Forest Service, We acknowledge funding from the Macrosystems Biology Program Grant EF Northern Research Station #1241860 from United States National Science Foundation (NSF). Haines TK, Busby RL, Cleaves DA (2001) Prescribed burning in the south: trends, purpose, and barriers. South J Appl For 25(4):149–153 Hanson C (2010) Southern forests for the future. https://www.wri.org/publication/ Availability of data and materials southern-forests-future. Accessed 22 May 2020 The dataset and code used during the current study are available at: https:// Haynes RW (2002) Forest management in the 21st century: changing numbers, doi.org/10.6084/m9.figshare.11406612.v1 changing context. J For 100(2):38–43 Hijmans RJ, Phillips S, Leathwick J, Elith J, Hijmans MRJ (2015) Package ‘dismo’. https:// Ethics approval and consent to participate cran.r-project.org/web/packages/dismo/dismo.pdf. Accessed 22 May 2020 Not applicable. Hou Z, Xu Q, Nuutinen T, Tokola T (2013) Extraction of remote sensing-based forest management units in tropical forests. Remote Sens Environ 130:1–10 Consent for publication Jiménez-Valverde A, Lobo J M (2007) Threshold criteria for conversion of Not applicable. probability of species presence to either–or presence–absence. Acta Ecol 31(3):361–369 Jin S, Yang L, Danielson P, Homer C, Fry J, Xian G (2013) A comprehensive Competing interests change detection method for updating the National Land Cover Database to The authors declare that they have no competing interests. circa 2011. Remote Sens Environ 132:159–175 Josephson H (1989) A history of forestry research in the southern United States. Received: 22 June 2020 Accepted: 23 January 2021 Notes 1462:78 http://www.srs.fs.usda.gov/pubs/3175. Accessed 22 May 2020 Kurz WA, Dymond CC, Stinson G, Rampley GJ, Neilson ET, Carroll AL, Ebata T, Safranyik L (2008) Mountain pine beetle and forest carbon feedback to References climate change. Nature 452(7190):987–990 Bailey RG (2004) Identifying ecoregion boundaries. Environ Manag 34(1):S14–S26 Marsik M, Staub C, Kleindl W, Fu CS, Hall J, Yang D, Stevens FR, Binford M (2017) PANG Battaglia M, Sands PJ (1998) Process-based forest productivity models and their AEA. doi: https://doi.org/10.1594/PANGAEA.880304. Accessed 22 May 2020 application in forest management. Forest Ecol Manag 102(1):13–32 McGarigal K, Romme WH, Crist M, Roworth E (2001) Cumulative effects of roads Becknell JM, Desai AR, Dietze MC, Schultz CA, Starr G, Duffy PA, Franklin JF, and logging on landscape structure in the San Juan Mountains, Colorado Pourmokhtarian A, Hall J, Stoy PC, Binford MW, Boring LR, Staudhammer CL (USA). Landsc Ecol 16:327–349 (2015) Assessing interactions among changing climate, management, and Mitchener LJ, Parker AJ (2005) Climate, lightning, and wildfire in the national forests disturbance in forests: a macrosystems approach. BioScience 65(3):263–274 of the southeastern United States: 1989-1998. Phys Geogr 26(2):147–162 Bennett AF (1991) Roads, roadsides and wildlife conservation: a review. http:// Mooney P, Corcoran P, Winstanley AC (2010) Towards quality metrics for worldcat.org/isbn/0949324353. Accessed 22 May 2020 OpenStreetMap. Proceedings of the 18th SIGSPATIAL International Breiman L (2001) Random forests. Mach Learn 45(1):5–32 Conference on Advances in Geographic Information Systems. ACM, San Jose Ciepluch B, Mooney P, Jacob R, Winstanley AC (2009) Using OpenStreetMap to 514–517 deliver location-based environmental information in Ireland. SIGSPATIAL Nelson JD, Finn ST (1991) The influence of cut-block size and adjacency rules on Special 1(3):7–22 harvest levels and road networks. Can J For Res 21(5):595–600 Coffin AW (2007) From roadkill to road ecology: a review of the ecological effects Netherer S, Nopp-Mayr U (2005) Predisposition assessment systems (PAS) as of roads. J Transp Geogr 15(5):396–406 supportive tools in forest management—rating of site and stand-related Coghlan G, Sowa R (1997) National forest road system and use. Draft rep. US hazards of bark beetle infestation in the high Tatra Mountains as an example Department of Agriculture, Forest Service, Engineering Staff, Washington, DC for system application and verification. Forest Ecol Manag 207(1):99–107 Davis DE, Colten CE, Nelson MK, Saikku M, Allen BL (2006) Southern United Niemelä J (1999) Management in relation to disturbance in the boreal forest. States: an environmental history. ABC-CLIO Forest Ecol Manag 115(2):127–134 Demir M, Hasdemir M (2005) Functional planning criterion of forest road network O’Neill RV, Hunsaker CT, Timmins SP, Jackson BL, Jones KB, Riitters KH, Wickham systems according to recent forestry development and suggestion in Turkey. JD (1996) Scale problems in reporting landscape pattern at the regional Am J Environ Sci 1(1):22–28 scale. Landsc Ecol 11(3):169–180 Douglass JE (1983) The potential for water yield augmentation from forest Ordway-Swisher Biological Station (2015) OSBS User Guide V. http://www. management in the eastern United States. J Am Water Res Assoc 19(3):351– ordway-swisher.ufl.edu/docs/OSBS_User_Guide.pdf. Accessed 22 May 2020 Oswalt SN, Smith WB, Miles PD et al (2014) Forest resources of the United States, Drever CR, Peterson G, Messier C, Bergeron Y, Flannigan M (2006) Can forest 2012: a technical document supporting the Forest Service 2010 update of management based on natural disturbances maintain ecological resilience? the RPA assessment. Gen tech rep WO-91. Washington, DC: U.S. Department Can J For Res 36(9):2285–2299 of Agriculture, Forest Service, Washington Office, p 218 Eker M, Coban HO (2010) Impact of road network on the structure of a Outcalt KW (2000) The longleaf pine ecosystem of the south. Native Plants J 1(1): multifunctional forest landscape unit in southern Turkey. J Environ Biol 31(1– 42–53 2):157–168 Ries L, Fletcher RJ Jr, Battin J, Sisk TD (2004) Ecological responses to habitat Faccoli M, Bernardinelli I (2014) Composition and elevation of spruce forests edges: mechanisms, models, and variability explained. Ann Rev Ecol Evol Syst affect susceptibility to bark beetle attacks: implications for forest 35:491–522 management. Forests 5(1):88–102 Riitters KH, Wickham JD, Coulston JW (2004) Use of road maps in national FHWA (Federal Highway Administration) (2013). Highway statistics 2013. assessments of forest fragmentation in the United States. Ecol Soc 9(2):13 Forman RT, Sperling D, Bissonette JA, Clevenger AP, Cutshall CD, Dale VH, Fahrig Robertson KM, Ostertag TE(2007) Effects of land use on fuel characteristics and L, France RL, Goldman CR, Heanue K, Jones J, Swanson F, Turrentine T, fire behavior in pinelands of Southwest Georgia. Proceedings of the 23rd tall Winter TC (2003) Road ecology: science and solutions. Island Press, timbers fire ecology conference: fire in Grassland & Shrubland Ecosystems, Washington San Diego, California Forman RTT, Alexander LE (1998) Roads and their major ecological effects. Ann Rollins MG (2009) LANDFIRE: a nationally consistent vegetation, wildland fire, and Rev Ecol Syst 29:207–231 fuel assessment. Int J Wildland Fire 18(3):235–249 Forman RTT, Deblinger RD (2000) The ecological road-effect zone of a Saura S, Martinez-Millán J (2000) Landscape patterns simulation with a modified Massachusetts (U.S.a.) suburban highway. Conserv Biol 14(1):36–46 Fox TR, Jokela EJ, Allen HL (2007) The development of pine plantation silviculture random clusters method. Landsc Ecol 15(7):661–678 in the southern United States. J For 105(7):337–347 Sciaini M, Fritsch M, Scherer C, Simpkins CE (2018) NLMR and landscapetools: An Franklin JF, Forman RTT (1987) Creating landscape patterns by forest cutting: integrated environment for simulating and modifying neutral landscape ecological consequences and principles. Landsc Ecol 1(1):5–18 models in R bioRxiv 307306 Franklin JF, Mitchell RJ, Palik BJ (2007) Natural disturbance and stand Siry JP (2002) Intensive timber management practices. Southern Forest Res development principles for ecological forestry. Gen tech rep NRS-19. Assess 14:327–340 Yang and Fu Forest Ecosystems (2021) 8:17 Page 17 of 17 Stanturf JA, Kellison RC, Broerman FS, Jones SB (2003). Productivity of southern pine plantations: where are we and how did we get here?. J Fores 101(3):26–31 Stephens SL, McIver JD, Boerner REJ, Fettig CJ, Fontaine JB, Hartsough BR, Kennedy PL, Schwilk DW (2012) The effects of forest fuel-reduction treatments in the United States. BioScience 62(6):549–560 Turner MG, Wear DN, Flamm RO (1996) Land ownership and land-cover change in the southern Appalachian highlands and the Olympic peninsula. Ecol Appl 6(4):1150–1172 TIGER Cartographic Boundary – Urban Areas, prepared by the U.S. Census Bureau (2015) Available from https://www.census.gov/geo/mapsdatadata/kml/kml_ ua.html. United States Department of Agriculture Forest Service (1999) Final environmental impact statement for the land and resource management plans for National Forests in Florida. https://www.fs.usda.gov/Internet/FSE_ DOCUMENTS/fseprd500375.pdf. Accessed 22 May 2020 US Fish, Wildlife Service (2006) Great Dismal Swamp National Wildlife Refuge and Nansemond National Wildlife Refuge Final Comprehensive Conservation Plan. http://www.fws.gov/uploadedFiles/Region_5/NWRS/South_Zone/Great_ Dismal_Swamp_Complex/Great_Dismal_Swamp/FinalCCP_GDS.pdf. Accessed 22 May 2020 Van der Vaart AW (1998) Asymptotic statistics (Vol. 3). Cambridge University Press, UK Verbesselt J, Hyndman R, Zeileis A, Culvenor D (2010) Phenological change detection while accounting for abrupt and gradual trends in satellite image time series. Remote Sensing Environ 114(12):2970–2980 Waldrop TA, White DL, Jones SM (1992) Fire regimes for pine-grassland communities in the southeastern United States. Forest Ecol Manag 47(1–4): 195–210 Warren DL, Glor RE, Turelli M (2008) Environmental niche equivalency versus conservatism: quantitative approaches to niche evolution. Evolution 62(11): 2868–2883 Wear DN, Greis J (2002) Southern forest resource assessment: summary of findings. J For 100(7):6–14 Wear DN, Greis JG (2012) The southern forest futures project: summary report. Gen tech rep SRS-GTR-168. USDA-Forest Service, Southern Research Station, Asheville, p 54 Wear DN, Greis JG (2013) The southern forest futures project: technical report. Gen tech rep SRS-GTR-178. USDA-Forest Service, Southern Research Station, Asheville, p 542 Wendland KJ, Lewis DJ, Alix-Garcia J, Ozdogan M, Baumann M, Radeloff VC (2011) Regional-and district-level drivers of timber harvesting in European Russia after the collapse of the Soviet Union. Glob Environ Chang 21(4):1290–1300 Zaccarelli N, Li BL, Petrosillo I, Zurlini G (2013) Order and disorder in ecological time-series: introducing normalized spectral entropy. Ecol Indic 28:22–30 Zhang D, Butler BJ, Nagubadi RV (2012) Institutional timberland ownership in the US south: magnitude, location, dynamics, and management. J For 110(7): 355–361 Zielstra D, Hochmair HH (2011) Digital street data: Free versus proprietary. GIM Int 25(7):29–33 http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png "Forest Ecosystems" Springer Journals

Mapping regional forest management units: a road-based framework in Southeastern Coastal Plain and Piedmont

"Forest Ecosystems" , Volume 8 (1) – Feb 25, 2021

Loading next page...
 
/lp/springer-journals/mapping-regional-forest-management-units-a-road-based-framework-in-y4HPmdgm0e
Publisher
Springer Journals
Copyright
Copyright © The Author(s) 2021
eISSN
2197-5620
DOI
10.1186/s40663-021-00289-w
Publisher site
See Article on Publisher Site

Abstract

Management practices are one of the most important factors affecting forest structure and function. Landowners in southern United States manage forests using appropriately sized areas, to meet management objectives that include economic return, sustainability, and esthetic enjoyment. Road networks spatially designate the socio- environmental elements for the forests, which represented and aggregated as forest management units. Road networks are widely used for managing forests by setting logging roads and firebreaks. We propose that common types of forest management are practiced in road-delineated units that can be determined by remote sensing satellite imagery coupled with crowd-sourced road network datasets. Satellite sensors do not always capture road- caused canopy openings, so it is difficult to delineate ecologically relevant units based only on satellite data. By integrating citizen-based road networks with the National Land Cover Database, we mapped road-delineated management units across the regional landscape and analyzed the size frequency distribution of management units. We found the road-delineated units smaller than 0.5 ha comprised 64% of the number of units, but only 0.98% of the total forest area. We also applied a statistical similarity test (Warren’s Index) to access the equivalency of road-delineated units with forest disturbances by simulating a serious of neutral landscapes. The outputs showed that the whole southeastern U.S. has the probability of road-delineated unit of 0.44 and production forests overlapped significantly with disturbance areas with an average probability of 0.50. Keywords: Forest management unit, Warren’s index, Neutral landscape, OpenStreetMap, Road ecology Introduction management patterns, reflecting long-term land-use leg- The Southeastern United States (SEUS) forest comprises acies (Josephson 1989; Haynes 2002), contribute to the 32% of the total U.S. forestland (Oswalt et al. 2014), complex land mosaic of SEUS. which combined with the productivity of the forest, It is challenging to quantify the ecological and anthropo- places this region at the forefront of American forestry genic mechanisms that control the spatial structure of the production (Fox et al. 2007). This heterogeneous land- forest landscape and its surrounding areas in these com- scape is composed of heavily managed forests, intensive plex forest mosaics. Forest management is the predomin- agriculture, and multiple metropolitan areas. SEUS, al- ant factor in forest ecology and structural patterns though one of the most densely forested regions in the (Becknell et al. 2015), but little is known about how man- United States (Hanson 2010), is also heavily dissected by agement practices are related to surrounding land-use at road networks (Coffin 2007). The diverse forest the regional scale. One thing that is known is that, in the SEUS, significant expansions of urban areas tend to con- * Correspondence: dyang1@uwyo.edu vert forested land to urban uses and that croplands tend Wyoming Geographic Information Science Center, University of Wyoming, to transition to pine plantations (Davis et al. 2006;Haynes 1000E. University Ave, Laramie, Wyoming 82071, USA 2002; Wear and Greis 2002, 2012, 2013;Stanturf et al. Department of Geography, University of Florida, 330 Newell Dr, Gainesville, Florida 32611, USA © The Author(s). 2021 Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. Yang and Fu Forest Ecosystems (2021) 8:17 Page 2 of 17 2003; Becknell et al. 2015). To understand the eco- Faccoli and Bernardinelli 2014), change water yield logical and anthropogenic influences of differently and hydrologic regulation (Douglass 1983), produce managed forests on ecosystem processes, all land- wood products, provide places for hunting and recre- scapes should be understood at multiple scales, from ation, and conserve habitat biodiversity. the local scale (forest management unit) to regional Forestland structures, functions, and ecological pro- scales, with the regional scale referring to broad forest cesses are scale-dependent (Battaglia and Sands 1998; mosaics that are formed from management patches Niemelä 1999; Drever et al. 2006). Regionally, for the (O’Neill et al. 1996). Understanding forest manage- purpose of sustainable forest management, we need to ment spatial patterns require defining a map-based develop criteria and indicators of management units. management unit, which is the subdivision regarding Forest management units in this study are zones or effects of land use on forest ecosystems. This research patches, which can be identified, mapped and managed seeks to further the understanding of how spatial pat- according to the land-use objectives. Road networks link terns of forest management affect land use, by posing human activities (e.g. management practices) and sur- the following questions: Do roads delineate forest rounding physical environments (land cover). For pro- management units? What is the spatial distribution of duction, preservation, and ecological forestry, in many road-defined forest management units in the SEUS? cases, forest management units are harvest or burn Moreover, how are the distributions of management units. Roads are built to create access for the managers units affected by different forest management ap- and harvesters. For example, the preserved forest in the proaches? And how does forest management affect Ordway Swisher Biological Station (OSBS) in northcen- nearby land use, and how does nearby land use affect tral Florida is subdivided by road networks into manage- forest management? ment (burn) units, which is the smallest unit of land that Forest management is the main driving force of for- is actively managed (Ordway-Swisher Biological Station est structure in SEUS (Becknell et al. 2015)and alters 2015). In the Joseph W. Jones Ecological Research Cen- forest properties and processes, which affects forest ter in Ichauway, Georgia, and OSBS, the internal road ecosystem services (Kurz et al. 2008;Stephensetal. network provides access to the research site and serves 2012;Oswalt etal. 2014). Forest management can be as prescribed fire breaks (Ordway-Swisher Biological classified into four categories: production forestry, Station 2015). For passive management practices, there ecological forestry, passive management and preserva- are currently no clear criteria defining the management tion management (Becknell et al. 2015). Production units. However, in national forest systems (mostly with management harvests forest products and sustains the ecologically managed forests and multi-use production bio-productivity of the system with the sole objective forests), existing roads and trails are used for controlling of producing wood, pulp, and other forest products. prescribed fire and wildfires (e.g. Apalachicola National In SEUS, production management based on silvicul- Forest, Osceola National Forest, United States Depart- ture systems, which homogenize parts of the land- ment of Agriculture Forest Service 1999). For an ex- scape, has predominated the SEUS (Siry 2002). ample of a privately-owned forest, the Red Hills region However, forest conservation systems have evolved in Georgia uses roads to delineate burn units ranging ap- considerably over recent decades (Mitchener and proximately from five to 30 ha (Robertson and Ostertag Parker 2005; Franklin et al. 2007). Ecological manage- 2007). ment uses legacies of disturbance, including inter- Road networks facilitate movements of humans and mediate stand disturbance processes such as variable connect natural resources with societies and economies. density thinning and fire, and variable and appropri- As conduits for human access to nature, the physical ate recovery times to manage forests that still pro- footprint of approximately 6.6 million km of roads in the duce economically valuable wood products while United States FHWA (Federal Highway Administration preserving many of the values of natural forests 2013) has significant primary and secondary impacts on (Franklin et al. 2007). Passive management is defined ecosystems and the distribution of species (Bennett as a practice with little or no active management. We 1991). Fifteen to 20 % of American land is subject to the arguethatall forestsare managedtosomedegree ecological effects of road networks (Forman and and that doing nothing is a form of management. Alexander 1998; Forman and Deblinger 2000). The most Preservation forestry aims to minimize the ecological noticeable effects of road networks on forest structures footprint of society with the objectives of protecting are landscape structure changes, including reduced wildlife and maintaining ecosystem services. Further- mean patch size, increased patch shape complexity, in- more, certain forest management practices can sup- creased edge densities, and reduced unit connectivity. In press wildfires (Waldrop et al. 1992), prevent insect one case, McGarigal et al. (2001) investigated the land- /pathogen outbreaks (Netherer and Nopp-Mayr 2005; scape structure changes of the San Juan Mountains from Yang and Fu Forest Ecosystems (2021) 8:17 Page 3 of 17 1950 to 1993 and found that roads had a more significant effects are not yet well developed (Ries et al. 2004; Hou ecological impact (e.g. core forest areas and patch sizes de- et al. 2013). Roads and streams may be challenging to crease) on landscape structure than logging activities. identify, or invisible because they do not open the can- In addition to management practices (e.g., harvesting, opy so that many roads are not detectable on satellite fertilizing), the construction of road networks divides imagery or even aerial photography. The reliability of the forested land into smaller patches, thereby increasing large-scale road data is also challenged due to issues of the potential intensity of the effects of management accuracy, coverage and immediacy, all of which can practices. Road networks in managed forests provide underestimate the extent and ecological impacts of roads easy access for managers and harvesters to extract and on forest structures (Riitters et al. 2004). We propose regenerate resources (Demir and Hasdemir 2005). Roads that common types of forest management are practiced may influence fire regimes by increasing fire ignition as in road-delineated units that are detectable by remote a result of human activities (Franklin and Forman 1987). sensing satellite images coupled with crowd-sourced Moreover, road networks alter the spatial configuration road network datasets. We also describe and study the of management patches by functioning as firebreaks, patterns of forest management units in response to land which form new patterns in landscapes (Franklin and ownership and different management practices. Forman 1987; Nelson and Finn 1991; Eker and Coban 2010). By quantifying the spatial patterns of manage- Study area ment units created by roads we may gain insight into We focus on the Southeastern U.S. Coastal Plain and the ecological effects of road networks on spatial forest Piedmont (SEUS) region (Fig. 1). The SEUS is located structures within differently managed areas. between Piedmont to the north and the Atlantic Ocean The impacts and ecological effects of roads on the to the east and covers a significant portion of the south- landscape might be misestimated because methods eastern United States. The SEUS is the home to the measuring the road-effect zones and landscape scale most densely production-forested region nationwide, Fig. 1 Study area of Southeastern United States Coastal Plain and Piedmont, with level III ecoregions (Bailey 2004) Yang and Fu Forest Ecosystems (2021) 8:17 Page 4 of 17 which makes up 32% of total U. S forest cover (Oswalt GeoTIFF rasters with a 250-m resolution from PANGAEA et al. 2014). Based on EPA eco-region descriptions, land (Marsik et al. 2017). cover in the SEUS is a mosaic of cropland, pasture, woodland and forests (Bailey 2004). Major silvicultural Road networks forests in SEUS are pine forests, such as slash pine We selected OpenStreetMap as the primary road data (Pinus elliottii Engelm.) and loblolly pine (Pinus taeda and the USDA National Forest service trail and road L.) forests. European settlement and the extensive har- maps (https://data.fs.usda.gov/geodata/edw/datasets. vesting in the early 1900s removed 98% of the original php?dsetCategory=transportation, accessed Dec 2016) as longleaf pine (Pinus palustris Mill.) forests, which was secondary in this study. OpenStreetMap is a collabora- one of the most dominant ecosystems in SEUS (Outcalt tive, crowdsourced project that creates free, open, and 2000) and converted them to plantations of native slash accessible maps of road networks. OpenStreetMap is pine. The SEUS forest system is a fire-dominated system one of the most popular and well-supported Volun- with native trees adapted to short-period stand-clearing teered Geographic Information (VGI) datasets (Mooney events. The primary forest management types are pro- et al. 2010). Community volunteers collect geographic duction and passive management due to the dominant information and submit it to the global OpenStreetMap ownership of private owners, logging companies, and in- database (Ciepluch et al. 2009). OpenStreetMap moni- vestment institutions (Real Estate Investment Trusts tors road networks at near real-time and includes add- (REITs) and Timber Investment Management Organiza- itional classes of roads such as private access roads and tions (TIMOs)) (Zhang et al. 2012). This diversity of driveways in rural areas, small service roads or alleys in land cover types is spatially heterogenous, and patch urban areas, and forest access roads. All those road fea- sizes of the numerous vegetation classes vary across a tures are critical for this research and no distinctions wide range of scales (Fig. 1). were drawn between the types of road, traffic volume, or other factors. OpenStreetMap shows up-to-date road Data sources networks information, which the other official road data- Forest extent bases do not offer. The accuracy of the OpenStreetMap In this study, forest extent is determined by a composite in our study region has been studied. For the state of of the 2006 and 2011 USGS National Land Cover Data- Florida, Zielstra and Hochmair (2011) compared road base (NLCD), which was constructed from Landsat im- networks dataset from different sources and concluded agery at 30-m spatial resolution (Jin et al. 2013). We that OpenStreetMap was significantly better the other aggregated 21 NLCD classes into two classes: forest (de- road databases. All OpenStreetMap data were down- ciduous, evergreen, mixed forest and woody wetland) loaded from the website of Geofabrik (http://download. and non-forest (including water). Only the pixels that Geofabrik.de, Accessed Dec 2016). USDA National For- contain 50% or more forest area in NLCD will be con- est Services Trails and Road Map provide the coverage sidered as forested pixels. We also extracted the SEUS of detailed transportation map in National Forests urban areas by using the most recent 2015 US Census (Coghlan and Sowa 1997). Bureau’s TIGER cartographic boundary urban areas Road density in SEUS was measured as the total length (TIGER 2015) dataset to remove urban areas from the of all roads (in kilometers) in a district divided by the analysis. total land coverage area of the district (km ) based on our developed road networks map (Figure S2). Forest management type An integrated random forest classifier was built from the Landscape fire and resource management planning tools analysis of long-term phenological features derived from (LANDFIRE) BFAST outputs and spectral entropy calculated from the Disturbance data from the Landscape Fire and Resource Terra-MODIS enhanced vegetation index (EVI: MOD13Q Management Planning Tools (LANDFIRE) disturbance data product), along with ancillary data such as land own- database were used to evaluate the management unit ership, and disturbance history to classify different forest map. LANDFIRE is a combination of Landsat images, management types (Breiman 2001; Verbesselt et al. 2010; fire program data, and cooperator-provided field data Zaccarelli et al. 2013). The forest management type map and other ancillary databases (e.g., PAD-US), and is a has a spatial grain of 250 m and is a composite of pheno- shared program between the wildland fire management logical patterns and changes in the patterns from February programs of the U.S. Department of Agriculture Forest 2001 through December 2016 (Figure S1). The SEUS for- Services and U.S. Department of the Interior (Rollins est management type map has an overall accuracy of 89% 2009). LANDFIRE also describes land cover/use both for a 10-fold cross-validation. The forest management spatially and temporally from 1999 to 2014 and provides raster is available for each region as georeferenced the existing vegetation composition map based on Yang and Fu Forest Ecosystems (2021) 8:17 Page 5 of 17 dominant species or group of dominant species. Road-delineated forests units were mapped on the for- Spatially, LANDFIRE is a Landsat-based (30 m) database, est extent map from the section of Forest Extent after which matches the 30-m spatial resolution of this study superimposing detailed road networks with the Region (https://www.landfire.gov/disturbance.php). We chose Group tool to identify forest clusters as units. When the LANDFIRE project data because the spatial scale is superimposing road maps, all road networks were con- small enough to detect subtle changes brought about by verted to one-pixel segments. After one-pixel wide road land management practices, and large enough to reflect segments were derived, we converted all the forest pixels the characteristic variability of essential ecological pro- to 1 and the pixels that contained at least one road seg- cesses (such as wildfire) in the appropriate spatial con- ment to non-forest pixels (30-m spatial resolution) to 0. text. The disturbance data from LANDFIRE will be used to evaluate management units delineated by the road Geospatial assessment network in the context of intensely managed forests in In SEUS, regional forest management activities were rep- SEUS (Figure S3). resented as disturbances as described by LANDFIRE data, such as clear cuts, fires, and thinning. Figure 2 Forest ownership shows the example view of LANDFIRE cumulative dis- Geospatial land-ownership data from federal and non- turbance with delineated forest extent, and it can be governmental agencies were integrated for land owner- clearly observed (with the Warren’s Index of 0.62) that ship mapping (Figure S4). Forest ownership in SEUS is the disturbed areas and delineated unit shared boundar- broadly categorized as publicly owned and privately ies and show a large degree of equivalence. The spatial owned according to the landowners. There are six sub- coincidence has been shown to facilitate the interpret- types of public ownership, which are federally protected, ation and integration of defining regional forest manage- federal, state protected, state, military, and local. Also, ment units. In this study, we propose a test based on there are four sub-types of private ownership: non- forest management unit and the geographic correspond- governmental organization, private, family, and corpor- ing forest disturbances to compare the geographical ate. The ownership classification implies different man- similarity between the management unit and corre- agement objectives, as well as landowner skills, budgets sponding forest disturbances. The assessment of geo- and interests. Datasets from other federal and state gov- graphic image overlap is analogous to quantifying the ernment agencies were regrouped and classified into ten niche similarity of two species in two dimensions. As sub-types to create a comprehensive dataset that in- two-dimensional rasters, both disturbance and forest ex- cludes public land ownership and privately protected tent data can be treated as homogenous and spatial- easements as well as specially designated areas and asso- explicit datasets. ciated protection level (see Table S1). The final product Testing the overlap between pairs of road- is a 250-m spatial resolution raster data depicting the delineated forest extent with disturbance regions was forest ownership types and resampled to 30-m in this compared using the similarity statistics of Warren’s study to match the spatial resolution of the NLCD Index (Warren et al. 2008). The values of Warren’s database. Index range from 0 to 1. The value of 0 means forest management units have no spatial overlap with forest Methods disturbance areas, and 1 means all forest management Mapping road-delineated units units are identical to disturbance areas. The statistics In each forest management type, the fundamental elem- of Warren’s Index assume probability distribution ent of management practice is the management unit. In defined over geographic space, in which the p (or this study, we define the individual forest clusters that X, i p ) denotes the spatial probability distribution of X delineated by road-networks as “management units” and Y, i (road-delineated forests compartments), or Y (prob- the clusters that directly derived from forest extent map ability distribution of forest disturbances) to cell i.In as “management patches.” We hereby developed two nature, Warren’s equivalency index carries no bio- comparative methods for landscape analysis by using logical assumptions concerning the parameters as be- two sets of input data, with and without incorporating ing from any probability distributions. Spatially, we OpenStreetMap. For the method without incorporating applied it into assessing road-delineated forest man- OpenStreetMap, management patches were mapped on agement units versus the areas of corresponding the forest extent map resulting from the map described disturbances. in the section of Forest Extent and the Region Group Firstly, the Hellinger distance was calculated to meas- tool in ESRI ArcGIS 10.X (ESRI Inc.,) to identify clusters ure and compare the probability distance (Van der Vaart of forest pixels that formed unique and unconnected 1998): forests. Yang and Fu Forest Ecosystems (2021) 8:17 Page 6 of 17 Fig. 2 Spatial visualization of road-delineated forest extent (road delineated units - yellow colored polygon) matching with LANDFIRE disturbance data. Warren’s Index for this sub-landscape is 0.62 sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 1) Spatially. In this study, we used the Worldwide Ref- pffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffi erence System 2 (WRS-2) row 17 path 39 (17_39 here- HpðÞ ; p ¼ p − p ð1Þ X Y X ;i Y ;i after), which is one scene of Landsat Thematic Mapper (TM). As shown in Fig. 3, the forests in 17_39 are heav- ily disturbed and fragmented, but it also shared the while the Warren’s statistic is: larges contiguous forests in SEUS (Okefenokee National Refuge). This heterogeneous landscape consists of a mix- IpðÞ ; p ¼ 1 − HpðÞ ; p ð2Þ X Y X Y ture of natural and plantation forests, urban centers, urban and rural residential areas, and commercial and To measure and quantify the similarity of differently small-scale agricultural operations. Instead of using managed management units with the corresponding dis- LANDFIRE disturbance data, we simulated a neutral dis- turbance areas, all the units were classified into four turbance layer by mimicking the disturbance patch groups (ecological, passive, preservation and production shape and area called “Random NLM” (Saura and management). A 10 km × 10 km grid was utilized to Martinez-Millán 2000; Sciaini et al. 2018). 17_39 was compute Warren’s Index over the SEUS. The inputs of divided into 325 individual landscapes with the spatial this iterative Warren’s Index spatial analysis are: 1) resolution of 30 m (10 km × 10 km in size), which means stacked disturbance area data from LANDFIRE and 2) each landscape in 17_39 contains 333 × 333 pixels corresponding road-delineated forest management units. (10,000 ha). We applied the neutral landscape generation These analyses were carried out using the nivheOverlap algorithm in each landscape in 17_39 for 325 times and function of dismo package in R 3.3.X (Hijmans et al. recalculated the Warren’s Index for each 10 km × 10 km 2015). landscape with the simulated disturbance layer with the We further tested the hypothesis of road delineated road networks delineated forest extent map. forest management units by generating neutral land- 2) Iteratively. We applied a range-based attribute filter scape models from two perspective as: to the SEUS road density map (Figure S2). The road Yang and Fu Forest Ecosystems (2021) 8:17 Page 7 of 17 Fig. 3 Case study region of Worldwide Reference System 2 (WRSII) row 17 path 39 (17_39). As the LANDFIRE EVT product shows, it encompasses a diversity of landcover and disturbance types. Much of the conifer and conifer-hardwood land cover found outside of the large riparian area of the Okefenokee Swamp (upper central) are heavily managed, privately owned plantations and mixed agriculture/timber land use − 2 densities range from 0 to 91.48 km∙km , so we used the By replotting the spatial probability of road networks interval of 5 to randomly select a serious spatial noncor- delineating management units in 17_39 (Fig. 6c), we related 10 km × 10 km grids with a total number of 19 recalculated the histogram of how the Warren’s Index landscapes in SEUS (Fig. 4). For each spot, we iteratively distributed over 17_39 region after replacing the LAND run the Random NLM algorithm for 500 times and cal- FIRE data with simulated disturbance layer, as shown in culated the Warren’s Index with overlaying the simu- Fig. 6d. The mean value of the probability distribution is lated disturbance resulted from NLM with the road 0.14 with the standard deviation of 0.082. The results networks delineated forest extent map. show that human-derived forest disturbances can result in a non-random association with forest extent. By com- Results paring with the LANDFIRE disturbance derived War- Spatial assessment rans’ I map (Fig. 6a), road networks make a great We calculated the Warren’s Index with overlaying the contribution in identifying and shaping forest patterns in road networks delineated forest extent with LANDFIRE the case study area 17_39, wherein SEUS reflect the disturbance map on a 10 km × 10 km grid in SEUS management practices on the ground, and so for the (Fig. 5). The total number of 10 km × 10 km grids is road-delineated forest compartments, we hereby call 11,072. Figure 5 shows the probability distribution of them forest management units. − 2 road network delineating management unit, with an Road density in SEUS ranges from 0 to 91 km∙km . average of 0.44 and the standard deviation is 0.28. We Table 1 lists the info of all 19 10 km × 10 km grids with a also computed the distribution and found that most for- variation of road densities, the mean values of 500 simu- ested areas have the Warren’s Index from 0.3 to 0.5. lated landscapes. We compared the original Warren’s When we extracted 17_39 (Fig. 6a), shown as Fig. 6b Index of all 19 landscapes with the mean values calcu- with the mean value of 0.54 and the standard deviation lated from simulated landscapes, and found all the simu- of 0.17. lated landscape show significant difference on each Yang and Fu Forest Ecosystems (2021) 8:17 Page 8 of 17 − 2 Fig. 4 Spatial distribution of 19 points with 5 km∙km intervals landscape with the original Warren’s Index. We also plot represented the diversity and heterogeneity of road- the histogram of the landscapes to see how the simu- delineated forests clusters. lated neutral landscape distributed (Fig. 7). A linear rela- As the unit size increases, the number of units de- tionship between the Warren’s Index overlaying creases in an approximately logarithmic manner (Fig. 8). simulated disturbances and roads with the Warren’s Overall, 177,400 units occupied the SEUS with the mean Index overlaying LANDFIRE disturbances with roads size of 29.5 ha. We found that 35.94% of the forest man- (Figure S5) with R of 0.4755. It shows the evidence that agement units represent more than 99% of the whole roads help to shape forest patterns by statistical test and forest and ranged in area from 0.5 to 172,886 ha (the approve the hypnosis that road delineates forest manage- Okefenokee National Wildlife Refuge). The remaining ment patterns. 64.06% of small forest compartments, which are smaller than 0.5 ha, covers only 0.98% of the forested area. The SEUS forests is dominated by management unit Forest management unit map size ranging from 100 to 10,000 ha. The forest manage- A total area of 5.24 × 10 ha managed forest was mea- ment unit-size class map was reclassified based on the sured in the SEUS from the forest extent map based on forest management unit sizes (Fig. 9). From Fig. 9, it can NLCD. When we compared the NLCD derived forest/ be clearly seen that riparian forests stand out as large, non-forest maps between 2006 and 2011, a strong ten- unbroken linear features (most of the orange colored dency of deforestation was found on our study area units), which cover average unit sizes from 10,000 to (1.7 × 10 ha). The total length of the roads in SEUS is 100,000 ha. A characteristic feature for the southeastern 2.26 million km, which was calculated based on the forest is that relatively small and large management number of rasterized road network multiply the spatial units locate close to each other, surrounded with small- resolution (30 m). The road densities in SEUS range sized units (Figs. 9 and 10). Figure 10 shows five close- − 2 from zero to 49.29 km∙km (Figure S2). We mapped ups/examples from multiple locations and landscapes. In the forest management units at regional scales, and the our study area, the largest management unit is an area spatial distribution of those management units of the Okefenokee National Wildlife Refuge (Georgia Yang and Fu Forest Ecosystems (2021) 8:17 Page 9 of 17 Fig. 5 The probability of road networks delineating management unit in SEUS. The spatial resolution of this map is 10 km, and the variable is the value of Warren’s Index calculated for the overlap of road-defined areas and disturbance areas from the LANDFIRE database and Florida) with the size of 172,886 ha (Figs. 9 and 10d), networks) with the histogram of the management followed by the Atchafalaya River basin in Louisiana at patches (without incorporating road networks. The red 102,367 ha (Fig. 9 and 10c). Riparian forests are often histogram shows the size-based frequency distribution of undisturbed in SEUS without much road access because patches without incorporating road networks and the their soils are too wet for harvesting machinery and the purple histogram illustrates the size frequency of forest trees are not as valuable. There is also a cluster of large management units. When refining the management forest management units (yellow: 1000–10,000 ha) along units with road networks, there are 17 times more the Piedmont ecoregion, and the middle part of the patches (management patches) compared with the map southeastern plains. The Atchafalaya River basin in without incorporating road networks (management southern Louisiana is full of canals and river channels units). but no roads. There are many smaller units in Alabama and Mississippi and central Louisiana due to the rela- Discussion tively high road density in this region, so the landscape Management units under different management is broken up according to the road density map. Another approaches example is Fig. 10e, Great Dismal Swamp National Road density (Figure S2) has been proposed as a broad Wildlife Refuge, is the largest intact forest across south- index of roads’ ecological effects in a landscape (Forman eastern Virginia and northeastern North Carolina and and Alexander 1998). The magnitude of average road − 2 was established for protecting and managing the densities ranges from 0.63 to 2.2 km∙km in all man- swamp’s ecosystem (USFWS 2006). aged forests in SEUS. Preservation forest holds the low- − 2 To illustrate the contributions of road networks to our est average road density of 0.63 km∙km , and the regional SEUS management unit map, a representative passively managed forest has the highest average road − 2 comparison set was done with and without incorporat- density of 2.2 km∙km . Ecological forest and production ing road networks data. In Fig. 11, we overlay the histo- forest have the average road density of 1.86 and 1.29 − 2 gram of the management unit (incorporating road km∙km , respectively. One reason is the building of Yang and Fu Forest Ecosystems (2021) 8:17 Page 10 of 17 Fig. 6 The spatial distribution map of Warren’s Index overlaying the road network delineated forested extent with neutral simulated disturbance map over 17_39 area (a). (b) is the one of 10 km × 10 km simulated landscapes, (c) is the distribution of the warrens’ I with the average value of 0.14 and the standard deviation of 0.08. The sub-figure shows the randomly simulated patches in a plot of 10 km by 10 km forest service roads and private logging roads, which ob- management units under different functional forest viously increase the road density in production forest- management types. For ecological management, the spe- land. Road density is a predictor of forest management cific practice is designed to emulate the outcome of nat- intensity (Wendland et al. 2011), and the indicator of ural disturbance, which is to create an uneven-age stand human interactions with forests (Forman et al. 2003). structure to manage competition between and within We compared the size-frequency distributions of man- multi-cohort stands. The distribution of ecological man- agement units with a map of different kinds of manage- agement units shows spatial heterogeneity with structur- ment (production, ecological, preservation, and passive ally complex stands. For passive management forest management) derived independently. Preservation and lands, as the passively managed forests mostly adopt production management had the largest patches, with many irregular shapes with blurred boundaries, and rup- means of 109.6 and 82.6 ha, respectively. Ecological and ture of connectivity. For preservation management for- passively managed units averaged about half as large as est: mostly large government-managed land for 73.8 and 73.0 ha, respectively (Table 2). multiple-purpose including watershed, wildlife, recre- We incorporated Warren’s Index to assess quantita- ation and wilderness aspects. Accordingly, various prac- tively the geographic overlap between forest tices may be applied to it such as harvest, cutting, Yang and Fu Forest Ecosystems (2021) 8:17 Page 11 of 17 Table 1 Spatial assessment covering the different road densities (km/km ) ID Road density Mean of simulated landscapes Standard error of simulated landscapes Warren's I 1 0 0.16 0.008 0.41 2 30.23 0.06 0.009 0.10 3 35.15 0.12 0.008 0.66 4 15.03 0.10 0.009 0.25 5 40.63982391 0.02 0.008 0.20 6 4.9943223 0.26 0.006 0.52 7 85.59053802 0.19 0.007 0.49 8 49.92194366 0.23 0.007 0.55 9 24.98642921 0.25 0.006 0.47 10 9.908494949 0.13 0.009 0.30 11 65.24546814 0.10 0.008 0.38 12 75.59457397 0.10 0.008 0.22 13 19.80970001 0.24 0.007 0.66 14 59.10257721 0.02 0.007 0.31 15 45.15394592 0.11 0.008 0.33 16 55.00973892 0.12 0.008 0.23 17 68.21031952 0.17 0.008 0.34 18 79.92221832 0.03 0.008 0.34 19 91.88208771 0.11 0.008 0.24 means significant different with a 95% confidence interval Fig. 7 Distributions of 19 simulated landscapes of SEUS (Numbers are corresponding with Fig. 4) Yang and Fu Forest Ecosystems (2021) 8:17 Page 12 of 17 Fig. 8 The spatial distribution of SEUS forest management units based on unit size retention cutting, thinning and prescribed fire. As shown road-delineated management units I = 0.33. As the dom- in Table 3, Warren’s Index represents the overlap be- inant forest: production forests, the average size of plan- tween forest management units under different manage- tation management units tends to be large, have a ment types with the corresponding forest disturbance uniform composition (Figure S9), are internally homoge- area. neous and involve practice such as clearcutting and thin- The 10 km × 10 km based spatial grid analysis of War- ning. For preservation/wilderness, across the whole ren’s Index is shown in Fig. 5 and Figure S6,S7,S8 and SEUS, the probability of road-delineated management S9. Among all four forest management types, production unit is 0.44, with the standard deviation of 0.29, because forestry showed the highest probability of road- the large areas of wildfires happened but with road net- delineated management units with I = 0.50, and the pas- work setting as firebreaks. Roads provide access and fire- sive managed forests showed the lowest probability of breaks, as the use of prescribed fires is widespread in Fig. 9 Logarithmic size-based frequency distribution of SEUS management unit. The blue dotted line indicates the mean value of the management units within the SEUS Yang and Fu Forest Ecosystems (2021) 8:17 Page 13 of 17 Fig. 10 Examples of forest management unit distributions Fig. 11 Comparison between with and without incorporating OpenStreetMap on unit size frequency distribution Yang and Fu Forest Ecosystems (2021) 8:17 Page 14 of 17 Table 2 Average management unit size in each forest management type Average management unit size Standard error Average road density Standard error −2 (ha) (Size) (km∙km ) (Density) Ecological management 73.80 0.124 1.86 0.0017 Passive management 73.0 0.037 2.20 0.0005 Preservation 109.6 0.079 0.63 0.0012 management Production management 82.6 0.015 1.29 0.0052 SEUS and much of the prescribed fires are on private addition, as the road networks increased, the number of lands (Haines et al. 2001). In this study, we used a small-sized parcels also shows a substantial increase. threshold of 50% of the similarity score although many The mean forest management units under different useful criteria for establishing such thresholds have been ownerships range from 73.2 (privately owned forests) to proposed (Jimenez-Valverde and Lobo 2007). In the ana- 115.9 ha (state protected forests) with the standard devi- lysis of Table 3, the criteria were just used to show bin- ation of 33.4 ha. By comparing with Table 2, we found ary predictions of four differently managed forests. that the average unit size of military land is 74.9 ha, which is the closest to the average management unit size of ecological forestry at 73.8 ha. The land with the own- Ownership representation of forest management unit ership of state protected represents the largest average Forest management activities are important links be- management unit at 115.9 ha. tween human and environmental factors, especially at regional scale. Forest ownership patterns also explain Conclusions different types of land management practices and trajec- Quantifying forest management units under different tories of land cover change (Turner et al. 1996). The aim management approaches is a key step to ensure that ap- of this part of the research is to link regional land own- propriate management practices and policies are in place ership to management. We produced the SEUS owner- to maintain the array of forest ecosystem services. Re- ship database (Figure S4; Table S1) by collecting the gionally, forest harvesting operations are conducted data from different sources, where the ownership was di- within road-defined boundaries. Forest management vided as Private and Public. Based on our understanding practices in SEUS are specifically the activities primarily of the SEUS forest ownership, we reclassified the forest dictated by forest harvest and the needs of management owner types into public and private with 10 sub-classes for recreation and sustainability. Roads networks in (Table 4): 1) Public: Federal protected, federal, state pro- SEUS are often used for setting firebreaks and timber tected, state, military, local and NGO lands; and 2) Pri- harvest boundaries, so there should be a spatial coinci- vate: Private, family and corporate forests. Our dence between road-delineated management units and ownership data indicate that 18.7% of the landowners disturbance. In this study, we assessed the forest man- are under public forest and 81.3% of private forest land- agement from stands to regional scale, by incorporating owners, which covers 41.5% and 58.5% total forestland. road networks and multi-temporal disturbance remote We can see that the type of land ownership affects sensing database. units’ distribution (Table 4). The special characteristics A number of conclusions can be drawn from the ana- of SEUS forest ownership patterns can result in strong lysis in this study. contrasts in management unit distribution. The major ownership types in the region are family, corporate and 1) Road networks play a role in delineating forests state, which have different management objectives. In from local to regional scale. By defining the individual forest clusters delineated by road Table 3 Warren’s Index between different managed units and networks as “forest management units” and the corresponding disturbances overlap with stacked LANDFIRE clusters that directly derived from forest extent map disturbance as “management patches”, we mapped the forest Management type Warrens I Standard Dev extent map of both “units” and “patches” and compared them with treating “patches” map as Ecological 0.38 0.26 background. There were 17 times more “units” than Passive 0.38 0.28 “patches” over the whole SEUS. And we also Preservation 0.44 0.29 summarized the size distribution road delineated Production 0.50 0.24 units, with units smaller than 0.5 ha comprised 64% Yang and Fu Forest Ecosystems (2021) 8:17 Page 15 of 17 Table 4 Forest mean unit sizes based on different ownership types in SEUS Owner Class Owner Type Mean Unit Area (ha) Stdev Area (ha) Public Federal Protected 95.4 54.0 1,291,925 Public Federal 89.7 32.9 2,750,912 Public State Protected 112.6 33.4 2,793,425 Public State 96.8 52.1 1,277,169 Public Military 73.4 40.7 740,831 Public Local 89.4 45.5 1,635,244 Public NGO 93.6 42.9 130,331 Private Private 73.6 42.9 11,695,831 Private Family 72.9 36.3 23,247,681 Private Corporate 86.9 35.8 13,628,388 of the counts of units, these small units altogether management perspective, more road landscape area covered only 0.98% of the total forest area. leads to less available land for trees at the 2) We quantitatively tested the probability distribution macrosystems scale. On the other hand, logging patterns by using Warren’s Index of road-delineated roads and trails are an efficient way to manage management units and the corresponding forest dis- forests, which we can see the relatively high road turbances area. The average probability of road- network densities in ecological and production delineated management units is 0.44, and we also forestry. visualized the probabilities by setting a 10 km × 10 km grid. In SEUS, the high equivalency between the This study represents benefits so society in that future road-delineated units and the corresponding areas management decisions can be evaluated cross scales, were found at most production forests, and large- taking account of both climate and disturbance regimes. size preserved areas (e.g. Okefenokee National More information on the effects of land ownership and Wildlife Preserve and St. Marks National Wildlife forest management, combined with the detailed road Refuge). network and a continental coverage land cover maps can 3) The combination of remote sensing data and aid in thwarting further forest fragmentation by promot- OpenStreetMap constitutes a useful tool to ing more reasonable road planning by land planners and monitor, characterize and quantify land cover and decision makers. management unit distributions at macrosystems scale. By using the NLCD as the forest reference Supplementary Information The online version contains supplementary material available at https://doi. data and OpenStreetMap as road networks dataset, org/10.1186/s40663-021-00289-w. we produced the OpenStreetMap refined management unit pattern map and analyzed the Additional file 1. spatial size distribution of forest patterns. In addition, by incorporating the OpenStreetMap, the Abbreviations roads are shown to play an important role in GEE: Google earth engine; LANDFIRE: Landscape fire and resource causing fragmentation of the remnant forestlands. management planning tools; MODIS: Moderate resolution imaging spectroradiometer; NLCD: National land cover database; NLM: Neutral The size frequency distribution tells us that all of landscape model; OSM: OpenStreetMap; SEUS: Southeastern U.S. Coastal the 64% of management units are small Plain and Piedmont; VGI: Volunteered geographic information management units (< 0.5 ha) making up just 0.98% Acknowledgments forestlands. Acknowledgement is made of the assistance of Dr. Michael Binford and Dr. 4) Our land ownership product indicates that 18.7% of Peter Waylen, Department of Geography, University of Florida, for English public forest and 81.3% of private and industrial writing and reviewing, for suggestions and discussion. Funding for this research was provided by the National Science Foundation Macrosystems forestland owners, cover 41.5% and 58.5% area of Biology Program Grant EF #1241860. total forestland, respectively. Management practices affected units are represented not only at the stand Authors’ contributions or local scale, but also will change the forest pattern YD and FC conceptualized the idea for the study, FC generated the SEUS forest ownership database; YD performed data analysis and led the writing dramatically at the regional scale. We provided of the manuscript; FC critically reviewed the data analysis, and contributed substantial evidence that road networks occupy a substantially to the writing. Both authors read and approved the final substantial proportion of forest. From a forest manuscript. Yang and Fu Forest Ecosystems (2021) 8:17 Page 16 of 17 Funding Newtown Square, PA: U.S. Department of Agriculture, Forest Service, We acknowledge funding from the Macrosystems Biology Program Grant EF Northern Research Station #1241860 from United States National Science Foundation (NSF). Haines TK, Busby RL, Cleaves DA (2001) Prescribed burning in the south: trends, purpose, and barriers. South J Appl For 25(4):149–153 Hanson C (2010) Southern forests for the future. https://www.wri.org/publication/ Availability of data and materials southern-forests-future. Accessed 22 May 2020 The dataset and code used during the current study are available at: https:// Haynes RW (2002) Forest management in the 21st century: changing numbers, doi.org/10.6084/m9.figshare.11406612.v1 changing context. J For 100(2):38–43 Hijmans RJ, Phillips S, Leathwick J, Elith J, Hijmans MRJ (2015) Package ‘dismo’. https:// Ethics approval and consent to participate cran.r-project.org/web/packages/dismo/dismo.pdf. Accessed 22 May 2020 Not applicable. Hou Z, Xu Q, Nuutinen T, Tokola T (2013) Extraction of remote sensing-based forest management units in tropical forests. Remote Sens Environ 130:1–10 Consent for publication Jiménez-Valverde A, Lobo J M (2007) Threshold criteria for conversion of Not applicable. probability of species presence to either–or presence–absence. Acta Ecol 31(3):361–369 Jin S, Yang L, Danielson P, Homer C, Fry J, Xian G (2013) A comprehensive Competing interests change detection method for updating the National Land Cover Database to The authors declare that they have no competing interests. circa 2011. Remote Sens Environ 132:159–175 Josephson H (1989) A history of forestry research in the southern United States. Received: 22 June 2020 Accepted: 23 January 2021 Notes 1462:78 http://www.srs.fs.usda.gov/pubs/3175. Accessed 22 May 2020 Kurz WA, Dymond CC, Stinson G, Rampley GJ, Neilson ET, Carroll AL, Ebata T, Safranyik L (2008) Mountain pine beetle and forest carbon feedback to References climate change. Nature 452(7190):987–990 Bailey RG (2004) Identifying ecoregion boundaries. Environ Manag 34(1):S14–S26 Marsik M, Staub C, Kleindl W, Fu CS, Hall J, Yang D, Stevens FR, Binford M (2017) PANG Battaglia M, Sands PJ (1998) Process-based forest productivity models and their AEA. doi: https://doi.org/10.1594/PANGAEA.880304. Accessed 22 May 2020 application in forest management. Forest Ecol Manag 102(1):13–32 McGarigal K, Romme WH, Crist M, Roworth E (2001) Cumulative effects of roads Becknell JM, Desai AR, Dietze MC, Schultz CA, Starr G, Duffy PA, Franklin JF, and logging on landscape structure in the San Juan Mountains, Colorado Pourmokhtarian A, Hall J, Stoy PC, Binford MW, Boring LR, Staudhammer CL (USA). Landsc Ecol 16:327–349 (2015) Assessing interactions among changing climate, management, and Mitchener LJ, Parker AJ (2005) Climate, lightning, and wildfire in the national forests disturbance in forests: a macrosystems approach. BioScience 65(3):263–274 of the southeastern United States: 1989-1998. Phys Geogr 26(2):147–162 Bennett AF (1991) Roads, roadsides and wildlife conservation: a review. http:// Mooney P, Corcoran P, Winstanley AC (2010) Towards quality metrics for worldcat.org/isbn/0949324353. Accessed 22 May 2020 OpenStreetMap. Proceedings of the 18th SIGSPATIAL International Breiman L (2001) Random forests. Mach Learn 45(1):5–32 Conference on Advances in Geographic Information Systems. ACM, San Jose Ciepluch B, Mooney P, Jacob R, Winstanley AC (2009) Using OpenStreetMap to 514–517 deliver location-based environmental information in Ireland. SIGSPATIAL Nelson JD, Finn ST (1991) The influence of cut-block size and adjacency rules on Special 1(3):7–22 harvest levels and road networks. Can J For Res 21(5):595–600 Coffin AW (2007) From roadkill to road ecology: a review of the ecological effects Netherer S, Nopp-Mayr U (2005) Predisposition assessment systems (PAS) as of roads. J Transp Geogr 15(5):396–406 supportive tools in forest management—rating of site and stand-related Coghlan G, Sowa R (1997) National forest road system and use. Draft rep. US hazards of bark beetle infestation in the high Tatra Mountains as an example Department of Agriculture, Forest Service, Engineering Staff, Washington, DC for system application and verification. Forest Ecol Manag 207(1):99–107 Davis DE, Colten CE, Nelson MK, Saikku M, Allen BL (2006) Southern United Niemelä J (1999) Management in relation to disturbance in the boreal forest. States: an environmental history. ABC-CLIO Forest Ecol Manag 115(2):127–134 Demir M, Hasdemir M (2005) Functional planning criterion of forest road network O’Neill RV, Hunsaker CT, Timmins SP, Jackson BL, Jones KB, Riitters KH, Wickham systems according to recent forestry development and suggestion in Turkey. JD (1996) Scale problems in reporting landscape pattern at the regional Am J Environ Sci 1(1):22–28 scale. Landsc Ecol 11(3):169–180 Douglass JE (1983) The potential for water yield augmentation from forest Ordway-Swisher Biological Station (2015) OSBS User Guide V. http://www. management in the eastern United States. J Am Water Res Assoc 19(3):351– ordway-swisher.ufl.edu/docs/OSBS_User_Guide.pdf. Accessed 22 May 2020 Oswalt SN, Smith WB, Miles PD et al (2014) Forest resources of the United States, Drever CR, Peterson G, Messier C, Bergeron Y, Flannigan M (2006) Can forest 2012: a technical document supporting the Forest Service 2010 update of management based on natural disturbances maintain ecological resilience? the RPA assessment. Gen tech rep WO-91. Washington, DC: U.S. Department Can J For Res 36(9):2285–2299 of Agriculture, Forest Service, Washington Office, p 218 Eker M, Coban HO (2010) Impact of road network on the structure of a Outcalt KW (2000) The longleaf pine ecosystem of the south. Native Plants J 1(1): multifunctional forest landscape unit in southern Turkey. J Environ Biol 31(1– 42–53 2):157–168 Ries L, Fletcher RJ Jr, Battin J, Sisk TD (2004) Ecological responses to habitat Faccoli M, Bernardinelli I (2014) Composition and elevation of spruce forests edges: mechanisms, models, and variability explained. Ann Rev Ecol Evol Syst affect susceptibility to bark beetle attacks: implications for forest 35:491–522 management. Forests 5(1):88–102 Riitters KH, Wickham JD, Coulston JW (2004) Use of road maps in national FHWA (Federal Highway Administration) (2013). Highway statistics 2013. assessments of forest fragmentation in the United States. Ecol Soc 9(2):13 Forman RT, Sperling D, Bissonette JA, Clevenger AP, Cutshall CD, Dale VH, Fahrig Robertson KM, Ostertag TE(2007) Effects of land use on fuel characteristics and L, France RL, Goldman CR, Heanue K, Jones J, Swanson F, Turrentine T, fire behavior in pinelands of Southwest Georgia. Proceedings of the 23rd tall Winter TC (2003) Road ecology: science and solutions. Island Press, timbers fire ecology conference: fire in Grassland & Shrubland Ecosystems, Washington San Diego, California Forman RTT, Alexander LE (1998) Roads and their major ecological effects. Ann Rollins MG (2009) LANDFIRE: a nationally consistent vegetation, wildland fire, and Rev Ecol Syst 29:207–231 fuel assessment. Int J Wildland Fire 18(3):235–249 Forman RTT, Deblinger RD (2000) The ecological road-effect zone of a Saura S, Martinez-Millán J (2000) Landscape patterns simulation with a modified Massachusetts (U.S.a.) suburban highway. Conserv Biol 14(1):36–46 Fox TR, Jokela EJ, Allen HL (2007) The development of pine plantation silviculture random clusters method. Landsc Ecol 15(7):661–678 in the southern United States. J For 105(7):337–347 Sciaini M, Fritsch M, Scherer C, Simpkins CE (2018) NLMR and landscapetools: An Franklin JF, Forman RTT (1987) Creating landscape patterns by forest cutting: integrated environment for simulating and modifying neutral landscape ecological consequences and principles. Landsc Ecol 1(1):5–18 models in R bioRxiv 307306 Franklin JF, Mitchell RJ, Palik BJ (2007) Natural disturbance and stand Siry JP (2002) Intensive timber management practices. Southern Forest Res development principles for ecological forestry. Gen tech rep NRS-19. Assess 14:327–340 Yang and Fu Forest Ecosystems (2021) 8:17 Page 17 of 17 Stanturf JA, Kellison RC, Broerman FS, Jones SB (2003). Productivity of southern pine plantations: where are we and how did we get here?. J Fores 101(3):26–31 Stephens SL, McIver JD, Boerner REJ, Fettig CJ, Fontaine JB, Hartsough BR, Kennedy PL, Schwilk DW (2012) The effects of forest fuel-reduction treatments in the United States. BioScience 62(6):549–560 Turner MG, Wear DN, Flamm RO (1996) Land ownership and land-cover change in the southern Appalachian highlands and the Olympic peninsula. Ecol Appl 6(4):1150–1172 TIGER Cartographic Boundary – Urban Areas, prepared by the U.S. Census Bureau (2015) Available from https://www.census.gov/geo/mapsdatadata/kml/kml_ ua.html. United States Department of Agriculture Forest Service (1999) Final environmental impact statement for the land and resource management plans for National Forests in Florida. https://www.fs.usda.gov/Internet/FSE_ DOCUMENTS/fseprd500375.pdf. Accessed 22 May 2020 US Fish, Wildlife Service (2006) Great Dismal Swamp National Wildlife Refuge and Nansemond National Wildlife Refuge Final Comprehensive Conservation Plan. http://www.fws.gov/uploadedFiles/Region_5/NWRS/South_Zone/Great_ Dismal_Swamp_Complex/Great_Dismal_Swamp/FinalCCP_GDS.pdf. Accessed 22 May 2020 Van der Vaart AW (1998) Asymptotic statistics (Vol. 3). Cambridge University Press, UK Verbesselt J, Hyndman R, Zeileis A, Culvenor D (2010) Phenological change detection while accounting for abrupt and gradual trends in satellite image time series. Remote Sensing Environ 114(12):2970–2980 Waldrop TA, White DL, Jones SM (1992) Fire regimes for pine-grassland communities in the southeastern United States. Forest Ecol Manag 47(1–4): 195–210 Warren DL, Glor RE, Turelli M (2008) Environmental niche equivalency versus conservatism: quantitative approaches to niche evolution. Evolution 62(11): 2868–2883 Wear DN, Greis J (2002) Southern forest resource assessment: summary of findings. J For 100(7):6–14 Wear DN, Greis JG (2012) The southern forest futures project: summary report. Gen tech rep SRS-GTR-168. USDA-Forest Service, Southern Research Station, Asheville, p 54 Wear DN, Greis JG (2013) The southern forest futures project: technical report. Gen tech rep SRS-GTR-178. USDA-Forest Service, Southern Research Station, Asheville, p 542 Wendland KJ, Lewis DJ, Alix-Garcia J, Ozdogan M, Baumann M, Radeloff VC (2011) Regional-and district-level drivers of timber harvesting in European Russia after the collapse of the Soviet Union. Glob Environ Chang 21(4):1290–1300 Zaccarelli N, Li BL, Petrosillo I, Zurlini G (2013) Order and disorder in ecological time-series: introducing normalized spectral entropy. Ecol Indic 28:22–30 Zhang D, Butler BJ, Nagubadi RV (2012) Institutional timberland ownership in the US south: magnitude, location, dynamics, and management. J For 110(7): 355–361 Zielstra D, Hochmair HH (2011) Digital street data: Free versus proprietary. GIM Int 25(7):29–33

Journal

"Forest Ecosystems"Springer Journals

Published: Feb 25, 2021

There are no references for this article.