Assessment of public and private land cover change in the United States from 1985–2018

Analysis of land cover composition and land cover change over time in the conterminous United States (CONUS) reveals important details of the Earth's surface as technological capability of Earth observing satellites improves (Brown et al 2020, Wulder et al 2012). Analyzing annual land cover change with satellites provides frequent observations of unique changes across large geographic areas that have the potential to go undetected if limited to infrequent airborne or in situ observations. Gaining a better understanding of the frequency and magnitude of land cover change is vital to inform effective land use management, conservation efforts, and initiatives aimed at protection of natural resources across the CONUS. Incorporating analysis of composition and changes to existing natural resource studies can improve our understanding of effectiveness of previous management initiatives and aid refinement of future strategies of publicly and privately managed land.

Delineation of public and private land can be accomplished through rigorous compilation of county- and state-level information but can be cumbersome and, sometimes, proprietary. To maximize the potential for widespread application of land cover change analysis, freely available data are advantageous. One such resource is the Protected Areas Database of the United States (PAD-US). This publicly available inventory of protected areas across the United States was originally developed to support the U.S. Geological Survey (USGS) Gap Analysis Project (GAP) (U.S. Geological Survey Gap Analysis Project USGS-GAP 2021) goal to identify species and plant communities not adequately represented in existing conservation lands. The PAD-US database (version 2.1) provides the most comprehensive available data of both terrestrial and marine protected areas (U.S. Geological Survey Gap Analysis Project USGS-GAP 2020). PAD-US geospatial data provides assignment about local, state, and federal management responsibility, as well as details about lands that are managed (i.e., 'protected') to preserve biological diversity, and support conservation, recreation, and public health. Combining the PAD-US database with land cover data, and land cover change data establishes a new capability of examining dynamics of natural resources in land uses such as forestry, agriculture, and development on public and private land across CONUS.

This study is not an attempt to compare LCMAP data with land change found in other multi-temporal land cover and land use assessments of the CONUS, such as the USDA's National Resource Inventory, Major Land Uses, Forest Inventory Analysis or older USGS efforts (e.g., Land Cover Trends). An investigation with that focus may be a worthy effort to answer other specific land change questions, although it would wrestle in dealing with combinations of differing land use or land cover definitions, different methodologies, and different geographic frameworks or constraints. Our study is a novel investigation of annual land cover and its change for the CONUS against a dataset that contains what we consider a 'first-order' land attribute, i.e., management by the main types of land ownership within the United States. As we show later, many of the drivers of contemporary U.S. land use change are associated with or impacted by this first-order land attribute.

Publicly and privately managed forests provide environmental services and resources including food, fuel, fiber, air and water supply filtration, flood and erosion control, sustainability of biodiversity and genetic resources, and carbon sequestration, in addition to providing opportunities for recreation, education, and cultural enrichment (Balloffet et al 2012). However, forests are subject to change as a result of a variety of different local, regional, and national forest management strategies and disturbances related to climate change, extreme weather events (e.g., drought and flood), invasive species, insect infestations (e.g., bark beetles and defoliators), wildfire, and other natural hazards (e.g., tornados, hurricanes, and derechos). Currently, privately managed forests experience heightened levels of disturbance to above-ground biomass compared to publicly managed forests (Zheng et al 2010). Current wildfire regimes are shifting toward greater intensity and frequency (Halofsky et al 2020). The value of understanding the dynamics of Tree Cover change is important to many of today's most pressing environmental issues such as climate change (Jain et al 2021, McGuire et al 2016), carbon sequestration (Pugh et al 2019, Sedjo and Sohngen 2012, Zheng et al 2013), forest age and health (Mcdowell et al 2020, Oswalt et al 2019), wildfires (Jain et al 2021), land use change (Butler and Wear 2013), and biodiversity (Hansen et al 2001).

Established in 1985, the federal Conservation Reserve Program (CRP) provided private landowners the voluntary option of removing land from agricultural production with the goal of protecting wildlife habitat, improving water quality, and reducing soil erosion through 10 to 15-year contracts. However, these allotments do not always represent a permanent conversion, as seen by the 25% decline in the amount of land enrolled in CRP between 2007 and 2015 (U.S. Department of Agriculture – Farm Service Agency USDA-FSA 2021). Much of this land returned to high-intensity agriculture with monocultural production of crops such as wheat, corn, and soy (Morefield et al 2016). Thus, cropland dynamics are an important aspect of the land change scenario across CONUS.

Synthesizing land cover data and land management delineations can aid our ability to address how land management decisions will affect land change dynamics in the context of whether they are managed by public or private entities. Because the majority of CONUS is privately managed, future goals of protection, conservation, biodiversity, carbon cycling, etc could require more collaborations between private land managers and public policymakers that influence how private lands are managed. For example, private land encompasses most lower elevation lands with productive soils (Scott et al 2001), so aspects of landscape position and natural resource derivatives would require careful consideration in the context of effective future management decision-making. There needs to be a clear understanding of public and private land cover distribution, spatial and temporal features of land cover change, and prioritization of how to address future effects of natural (e.g. climate change) and anthropogenic (e.g. development) disturbances on different land cover conversions. This research explores the continuously changing status of land cover across the CONUS through the lens of land management by addressing the following questions: (1) How does the composition of land cover differ between public and private management?; (2) How has land cover change on public land differed from land cover change on private land from 1985 to 2018?

Delineation of public and private lands was accomplished using the USGS Protected Areas Database for the United States (PAD-US) v2.1 data (U.S. Geological Survey Gap Analysis Project USGS-GAP 2020) that provides a comprehensive dataset for land management attributes. To analyze land cover and land cover change from 1985–2018 we used the PAD-US 2.1 dataset to differentiate publicly and privately managed areas. Because the PAD-US data are intended to be used for purposes related to conservation, land management, planning, and recreation, we focus on labeling land as either public or private based on assumed management responsibility, not ownership. A goal of this study is to analyze publicly available data even though it inherently assumes public management responsibility for some parcels that are technically privately owned (ex. private inholdings within National Park boundaries). Public lands in this study are defined as lands managed by federal agencies, local, state, and regional offices, and non-government organizations (NGOs) that manage public lands in the PAD-US data. All other land is considered managed by private entities in this study.

A few slight modifications to the PAD-US data were made with the intention of improving the accuracy of some land management boundaries. For example, National Park boundaries found in the PAD-US data were refined using geospatial data from the National Park Service's Land Resources Division (U.S. Department of Interior – National Park Service USNPS 2021). Although the Bureau of Indian Affairs (BIA) oversees Native American lands and reservations, these lands are deemed private in this study because of tribal and individual Native American sovereignty in land management decisions. Defining geospatial boundaries for tribal land was not solely based on the PAD-US data, but a blend of geospatial data from the PAD-US, BIA (USBIA 2021), and the U.S. Census Bureau (U.S. Census Bureau 2021a).

The USGS Land Change Monitoring, Assessment, and Projection (LCMAP) initiative provides a suite of freely available map and reference data products of annual land cover composition and land cover change (Brown et al 2020, Pengra et al 2020a, 2021a). The LCMAP's eight land cover classes are similar to Anderson et al (1976) Level I but may have their own definition that are different than that seminal work as well as other subsequent USGS and other U.S. government land-cover class criteria (Brown et al 2020). LCMAP's reference data were developed to quantify accuracy and validation of thematic land cover and its change (Stehman et al 2021) and to statistically estimate annual land cover composition and change across the CONUS. The LCMAP project's reference data started as version 1.0, which was a large random sample of 24,971 (30- × 30-m) plots located across the CONUS that were evaluated by trained interpreters for the years 1985–2018 (Pengra et al 2020a, 2020b). Pengra et al (2020b) report that extensive interpreter training, feedback, and other quality assurance/quality control (QA/QC) efforts were implemented to ensure consistent quality of the reference dataset. Version 1.1 of the LCMAP reference data includes updated interpretations for the years 1984–2018. This study examines version 1.2 of the LCMAP reference data published by Pengra et al (2020c) which includes an additional 2,000 plots which were selected with a stratified sampling method (Stehman 2013, Olofsson et al 2014) that were mostly based on land cover change found in LCMAP version 1.0 Annual Land Cover Change (LCACHG) map products 1986 through 2017 (Pengra et al 2021a). For this analysis, 813 coastal/offshore LCMAP reference plots were excluded from the v1.2 data resulting in a total of 24,158 plots. Removal of these plots reduced the extent of the study area and affected land cover composition and area estimates, most notably in land cover area classified as water, compared to previously published LCMAP CONUS-wide research (e.g., Pengra et al 2020a, Pengra et al 2021b, Auch et al 2022). The remaining v1.2 reference plots were then grouped into public and private subsets as they aligned with our delineation of the PAD-US dataset. Regional analyses of four megaregions were based on Omernik ecoregions (Omernik and Griffith 2014) using boundaries generally defined in previous USGS regional studies (Sleeter et al 2012, Taylor et al 2015, Auch and Karstensen 2015, Sayler et al 2016, Pengra et al 2020b) (figure 1).

Zoom In Zoom Out Reset image size

Published freely available LCMAP Reference data v1.2 (Pengra et al 2020c) is selected for this study because it allows for statistically rigorous area estimates of land cover composition and change on public and private land across CONUS when combined with the PAD-US data. Pengra et al (2020b) report that all LCMAP reference data are generated by trained interpreters who assign (1) land use, (2) land cover, and (3) change processes for every year between 1985 and 2018 to each reference sample plot using the TimeSync (Cohen et al 2010) Landsat time series visualization and data collection tool (Pengra et al 2020c, Xian et al 2022). After TimeSync interpretation, the information was translated to the appropriate LCMAP land cover class, providing a single land cover reference label for each sample plot (Pengra et al 2020b, Pengra et al 2020c). Criteria that define each of the eight LCMAP land cover classes can be found in appendix A. Statistical procedures for estimations of land cover composition and land cover change are outlined in Stehman (2014) and included in Appendix B. Accuracy of annual land cover maps and land cover change maps are available in Stehman et al (2021), and additional detailed tables of results are available in Pengra et al (2020c). The reference-based land cover area estimates or change areas do not have an accuracy measure except what the standard errors of the sampling describes. Stehman et al (2021) note that full transparency is provided in Pengra et al (2020b) for anyone interested in independent evaluation of the LCMAP reference data by providing information about reference class assignments and reference plot locations. Finally, all versions of the LCMAP reference data are available from Pengra et al (2020a, 2020b, 2020c, 2021a, 2021b).

In many cases, the land cover classes are the result of biophysical conditions, land use, or both. But in other cases, some reference plots may get labeled a specific class because it fits the definition or criteria of the label, but these years could be in more of a transitory condition. Also, from the LCMAP classes, the results can be a mixture of both. For example, 'Barren' in the biophysical aspect could be a sandy beach or sandbar, upper elevations of a mountain, a rocky outcrop, or playa, but 'Barren' in a forested area could be bare ground from a very fresh clearcut timber harvest. Young forest regrowth after harvest or wildfire is also typically classed as 'Grass/Shrub' for a period of varying years. Thus, the labeling criteria for reference data interpreters was defined yet some uncertainty in annual classification of some plots may exist due to differences in interpretation of analysis. Consistent and accurate interpretation was more difficult for commonly challenging classes, which decreased interpreter agreement for Disturbed (46% - A term used during the reference dataset testing that primarily affected change in Tree Cover land cover class), Barren (56%), and Wetland (74%) (Pengra et al et al 2020b). However, CONUS-level agreement for the four most prevalent classes (Tree Cover, Grass/Shrub, Cropland, and Water) ranged from 89% to 94% (Pengra et al 2020b). Overall, interpreter agreement for all reference samples was 88%, whereas from 1985 to 2016 agreement ranged from 87.4 to 88.9% (Pengra et al 2020b).

3.1. Land cover composition

Far more land is privately managed (73% or 5,694,000 km²) than publically mananged (27% or 2,094,000 km²) across CONUS and estimated land cover composition differs greatly. We present composition as estimated area of each LCMAP class as percentages for CONUS, public, and private land in figure 2 and table 1 (also see appendix C). The most common among the eight LCMAP land cover classes is Grass/Shrub averaging 40 ± 0.5% (standard error) of CONUS from 1985–2018. Grass/Shrub made up 35.0 ± 0.5% of private land and 52.2 ± 0.6% of public land. Tree Cover, the second most common land cover, averaged 29 ± 0.4% of CONUS. Tree Cover on private land nearly doubles the area of publically managed Tree Cover although private Tree Cover represents only 26.6 ± 0.3% of private land while Tree Cover represents 36.2 ± 0.5% of public land. Cropland, the third most common land cover, averaged 18 ± 0.3% of CONUS (figure 2(a)). Cropland reprsents 24.8 ± 0.4% of private land but just 0.93 ± 0.2 % of public land. Developed land cover averaged 5 ± 0.2% of CONUS. Developed land represents 6.5 ± 0.3% of private land (figure 2(b)), but only 0.90 ± 0.2% of public land. Water, Wetland, and Barren made up 1.8 ± 0.1%, 5 ± 0.3%, and 0.9 ± 0.1% of CONUS, respectively. Water averaged 1.9 ± 0.1% on private land, and 1.7 ± 0.1% on public land. Wetlands made up 5.0 ± 0.2% of private land, and 5.2 ± 0.3% of public land. Barren land cover represented 0.2 ± 0.04% of private land, and 2.9 ± 0.2% of public land. Only three reference plots represented the Snow/Ice LCMAP class; therefore, subsequent analysis of this class was excluded from this study.

Zoom In Zoom Out Reset image size

Table 1. Estimated area and normalized percentages and standard errors (SE) of land cover composition on CONUS, public, and private land for each LCMAP class.

CONUS		Area (km2)	SE (± km2)	% CONUS	SE (± %)
Developed	392,030	17,375	5	0.2
	Cropland	1,430,144	1,522	18	0.3
	Grass/Shrub	3,084,779	38,193	40	0.5
	Tree cover	2,274,977	2,298	29	0.4
	Water	143,152	8,785	1.8	0.1
	Wetland	392,030	19,805	5	0.3
	Snow/Ice	919	433	0.01	0.006
	Barren	70,786	6,170	0.9	0.1
		Area (km2)	SE (± km2)	Normalized %	SE (± %)
Private	Developed	372,313	14,314	6.5	0.3
	Cropland	1,410,725	22,769	24.8	0.4
	Grass/Shrub	1,992,568	26,766	35.0	0.5
	Tree cover	1,516,943	21,717	26.6	0.3
	Water	106,631	5,863	1.9	0.1
	Wetland	283,988	13,004	5.0	0.2
	Snow/Ice	—	—	—	—
	Barren	11,136	2,057	0.2	0.04
Public	Developed	18,875	3,061	0.90	0.2
	Cropland	19,419	3,544	0.93	0.2
	Grass/Shrub	1,092,211	11,428	52.2	0.6
	Tree cover	758,034	10,906	36.2	0.5
	Water	36,521	2,922	1.7	0.1
	Wetland	108,041	6,801	5.2	0.3
	Snow/Ice	919	433	0.04	0.006
	Barren	11,135	2,057	2.9	0.04

3.2. Net composition changes from 1985 to 2018

Examining net changes to land cover composition provides an indication of how land cover composition has shifted over time. Figure 3 and table 2 describe net changes from CONUS, private, and public land for each of the land cover classes in this study from 1985 to 2018. One similarity of both public and private land is that overall land cover class change averaged less than 1% per year. The largest proportion of net land cover change for all of CONUS and on private land was an increase in Developed. The second largest proportional change for all of CONUS and on private land was a net loss of Cropland. A net increase of Grass/Shrub and a net decrease of Tree Cover were the two largest proportional changes on public land. Net Tree Cover loss on private land was just slightly more than what was lost on public land. The remaining land cover classes experienced small net changes. Barren land had the smallest net change overall.

Zoom In Zoom Out Reset image size

Table 2. Estimated area and normalized percentages and standard errors (SE) of net land cover changes on CONUS, public, and private land for each LCMAP class from 1985–2018. Note: CONUS-wide estimates were calculated without the regional stratification; therefore, while they are consistent considering the estimation uncertainty, the mean estimate is not the exact sum of the regional estimates.

CONUS		Area (km2)	SE (± km2)	% CONUS	SE (± %)
	Developed	134,938	5,881	1.73	0.08
	Cropland	−113,229	9,044	−1.45	0.12
	Grass/Shrub	25,941	11,973	0.33	0.15
	Tree cover	−55,224	9,626	−0.71	0.12
	Water	10,356	2,657	0.13	0.03
	Wetland	−4,215	2,636	−0.05	0.03
	Snow/Ice	—	—	—	—
	Barren	1,432	1,182	0.02	0.02
		Area (km2)	SE (± km2)	Normalized %	SE (± %)
Private	Developed	131,021	5,760	2.30	0.10
	Cropland	−107,003	8,889	−1.88	0.16
	Grass/Shrub	4,483	11,064	0.08	0.19
	Tree cover	−33,053	8,598	−0.58	0.15
	Water	9,692	2,265	0.17	0.04
	Wetland	−4,377	2,363	−0.08	0.04
	Snow/Ice	—	—	—	—
	Barren	−762	806	−0.01	0.01
Public	Developed	3,988	853	0.19	0.01
	Cropland	−6,238	1,509	−0.30	0.03
	Grass/Shrub	21,291	4,447	1.02	0.08
	Tree cover	−22,001	4,224	−1.05	0.07
	Water	528	1,280	0.03	0.02
	Wetland	423	1,151	0.02	0.02
	Snow/Ice	—	—	—	—
	Barren	2,008	828	0.10	0.01

3.3. Annual land cover change and regional characteristics of change

Annual land cover change for each individual LCMAP land cover class is reported in the following sections separately. We present estimates of annual gross change, which combines estimates of annual gains and losses from each class based on the reference sample (table 3, appendix C). Each section presents net annual change across CONUS, annual private and public change, where net gains and losses occurred regionally, and contributions of net change by each of the other LCMAP classes.

Table 3. Estimated area and normalized percentages and standard errors (SE) of annual average gains and losses (gross change) on CONUS, public, and private land for each LCMAP class from 1985–2018. Asterisks symbolize that the value represents a normalized percentage based on the total area of either public or private land.

CONUS		gain	gain SE	gain	gain SE	loss	loss SE	loss	loss SE
		km2	± km2	%	± %	km2	± km2	%	± %
	Developed	4,784	1,061	0.06	0.01	−695	389	−0.01	0.005
	Cropland	5,235	1,203	0.07	0.02	−8,666	1,511	−0.11	0.02
	Grass/Shrub	19,800	2,367	0.25	0.03	−19,014	2,295	−0.24	0.03
	Tree cover	12,128	1,855	0.16	0.02	−13,802	1,990	−0.18	0.03
	Water	2,073	569	0.03	0.01	−1,759	502	−0.02	0.01
	Wetland	1,338	496	0.02	0.01	−1,466	519	−0.02	0.01
	Snow/Ice	—	—	—	—	—	—	—	—
	Barren	904	382	0.01	0	−860	371	−0.01	0
		gain	gain SE	gain*	gain SE*	loss	loss SE	loss*	loss SE*
		km2	± km2	%	± %	km2	± km2	%	± %
Private	Developed	4,631	2,940	0.08	0.05	−670	605	−0.01	0.01
	Cropland	5,153	2,627	0.09	0.05	−8,379	3,751	−0.15	0.07
	Grass/Shrub	17,082	6,222	0.30	0.11	−16,937	6,567	−0.30	0.12
	Tree cover	10,444	3,326	0.18	0.06	−11,480	4,302	−0.20	0.08
	Water	1,379	955	0.02	0.02	−1,068	749	−0.02	0.01
	Wetland	1,038	739	0.02	0.01	−1,171	827	−0.02	0.01
	Snow/Ice	—	—	—	—	—	—	—	—
	Barren	625	518	0.01	0.01	−649	527	−0.01	0.01
Public	Developed	134	128	0.006	0.006	−18	18	−0.001	0.001
	Cropland	61	61	0.003	0.003	−275	266	−0.01	0.01
	Grass/Shrub	2,641	1,504	0.13	0.07	−1,999	1,330	−0.10	0.06
	Tree cover	1,633	1,020	0.08	0.05	−2,282	1,196	−0.11	0.06
	Water	625	467	0.03	0.02	−617	453	−0.03	0.02
	Wetland	298	238	0.01	0.01	−268	193	−0.01	0.01
	Snow/Ice	—	—	—	—	—	—	—	—
	Barren	245	175	0.01	0.008	−179	137	−0.009	0.007

3.3.1. Grass/Shrub

Grass/Shrub is the most common land cover on both private and public land. Net annual gains occurred across CONUS in consecutive years from 1986 to 1992 but then net losses occurred in all but 6 of the 22 years from 1993 to 2014 (figure 4(a)). The largest annual net losses on private land occurred in 1997, 2008, and from 2011 to 2014, and the greatest net gains occurred from 1986 to 1992 (figure 4(b)). On public land, the largest net loss occurred in 1993 and the largest net increase occurred in 1988 (figure 4(c)). On private land, Grass/Shrub gains and losses mirrored one another except at the beginning and the end of the study period where gains were much larger than losses (figure 4(b)), with most of the net losses occurring in the East-Central and West regions and all of the net gains occurred in the West-Central region (figure 4(d)). The average private Grass/Shrub gain per year was 0.30 ± 0.11% and loss per year was −0.30 ± 0.12% (figure 4(b)). Grass/Shrub that converted to Developed occurred primarily on private land (figure 4(e)). Grass/Shrub gains and losses on public land varied greatly over time, yet average yearly gain was greater than the average yearly loss (figure 4(c)). Average public Grass/Shrub gain per year was 0.13% ± 0.07, while average loss per year was −0.10% ± 0.06 (figure 4(c)). All net losses of public Grass/Shrub were in the East region and the majority net gains were in the West region (figure 4(f)). Overall, the greatest changes to private Grass/Shrub were associated with conversions with Tree Cover and Cropland (figure 4(e)) and the greatest changes to public Grass/Shrub were associated with conversions with Tree Cover on public land (figure 4(g)).

Zoom In Zoom Out Reset image size

3.3.2. Tree cover

Tree Cover is the second most common land cover on both private and public land. Across CONUS, net tree cover losses occurred in consecutive years from 1986 to 1990 and from 2012 to 2018, with short periods (1–3 years) of both gains and losses from 1991 to 2011 (figure 5(a)). From an annual perspective, the largest net losses on private land occurred in 1986, 1987, 2016, and 2018 and the greatest net gain occurred in 2008 (figure 5(a)). On public land, the largest net loss occurred in 1988. Public land did not experience net gain in tree cover until 1993 and, apart from 1994, those gains continued through 1999 (figure 5(a)). On private land, Tree Cover gains and losses mirrored one another except at the beginning and the end of the study period where losses were much larger than gains (figure 5(b)), which is nearly a reciprocal of Grass/Shrub fluctuations (see Section 3.3.1). The average private Tree Cover gain per year was 0.18 ± 0.06% and loss per year was −0.20 ± 0.08% (figure 5(c)). Tree Cover gains and losses on public land varied greatly over time, yet average yearly gain was one-half of the average yearly loss. Average public Tree Cover gain per year was 0.08 ± 0.05%, while average loss per year was −0.11% ± 0.06 (figure 5(c)). Most of the net Tree Cover losses on private land occurred in the East region and most of the net gains were in the East-Central region (figure 5(d)). Tree Cover that converted to Developed occurred primarily on private land (figure 5(e)). The majority net losses of public Tree Cover were in the West region and the majority net gains were in the East region (figure 5(f)). Overall, the greatest changes to Tree Cover were associated with conversions to and from Grass/Shrub on both private (figure 5(e)) and public (figure 5(g)) land.

Zoom In Zoom Out Reset image size

3.3.3. Cropland

Net Cropland losses persisted from 1985 to 2007 across CONUS, then net gains occurred every year from 2008 to 2018 except for 2017. The largest annual net loss on private land occurred in 1989 and the greatest net gains were in 2012 and 2014 (figure 6(a)). Private Cropland losses were much greater than gains from 1985 to 2007 (figure 6(b)). The average private Cropland gain per year was 0.09 ± 0.05% and loss per year was −0.15 ± 0.07% (figure 6(b)). Cropland gains (0.003 ± 0.003%) and losses (−0.01 ± 0.01%) on public land were very minimal (figure 6(c)). Cropland that converted to Developed occurred primarily on private land, but the greatest net changes in private Cropland were in the West-Central and the East regions (figure 6(d)) and were associated with conversions with Grass/Shrub (figure 6(e)). Net losses to public Cropland primarily occurred in the West-Central and East-Central regions (figure 6(f)) resulting from conversions to Grass/Shrub and Wetland classes (figure 6(g)).

Zoom In Zoom Out Reset image size

3.3.4. Developed

Across CONUS, net gains in Developed occurred in all years, reaching a maximum in 2002 and a minimum in 2014 (figure 7(a)). On private land, net gains of Developed steadily increased between 1986 and 2006, and then a large decline in net gains persisted from 2007 to 2018 (figure 7(a)). Private Developed land cover losses were minimal (figure 7(b)). The average annual private Developed gain rate was 0.08 ± 0.05% per year and loss was −0.01 ± 0.01% per year (figure 7(b)) across the CONUS during the study period. Public gains (0.006 ± 0.006%) and losses (−0.001 ± 0.001%) were very small (figure 7(c)). Most of the net gains of private Developed occurred in the East region and remaining gains evenly distributed between the other three regions (figure 7(d)). Overall, the greatest changes to private Developed were associated with conversions from Grass/Shrub, Tree Cover, and Cropland (figure 7(e)). It is worth noting that in the East and East-Central regions, LCMAP Grass/Shrub also includes 'pasture' land. Developed gains and losses on public land were minimal (figure 7(c)), although most estimated net gains occurred in the West region with near equal distribution of additional Developed in the other three regions (figure 7(f)) which was, like on private land, associated with conversions with Grass/Shrub, Tree Cover, and Cropland (figure 7(g)).

Zoom In Zoom Out Reset image size

3.3.5. Water

Net annual changes in Water on both private and public land had similar patterns (figure 8(a)). Across CONUS, net Water gains occurred in from 1991 to 1998 and from 2005 to 2012 with the exceptions of 1994 and 2007, whereas net losses primarily occurred from 1987 to 1990 and from 1999 to 2004 (figure 8(a)). Private Water gain per year was 0.02 ± 0.02% and loss per year was −0.02 ± 0.01% (figure 8(b)). Public land experienced Water gains of 0.03 ± 0.02% and losses of −0.03 ± 0.02% (figure 8(c)). Most of the private net gains occurred in the East and West-Central regions while most net losses of Water were in the West region (figure 8(d)). Private water conversions were primarily associated with Wetlands, Grass/Shrub, and Barren classes (figure 8(e)). Approximately one-half of the public net gains were in the West-Central region and the remaining fraction of net gain was split nearly equally between the East and East-Central regions, while all estimated net loss of public Water was in the West (figure 8(f)). As with private land, the greatest changes to Water on public land were associated with conversions to and from Wetland, Grass/Shrub, and Barren (figure 8(g)).

Zoom In Zoom Out Reset image size

3.3.6. Wetland

Fluctuations of Wetlands across CONUS varied annually with net gains quickly followed by net losses. The longest period of sustained net gains in Wetland across CONUS occurred from 1999 to 2007 with a maximum in 2000, whereas the periods of net losses primarily occurred from 1995 to 1997, and all years from 2008 to 2018 with the exceptions of 2012 and 2016 (figure 9(a)). Patterns of net changes to Wetlands on both private and public land were not always the same (figure 9(a)). On private land, the average private Wetland gain per year was 0.02 ± 0.01% and loss per year was −0.02 ± 0.01% (figure 9(b)). Public Wetlands experienced annual average gains of 0.01 ± 0.01% and losses of −0.01 ± 0.01% (figure 9(c)). Most of the net gains occurring on private land were in the West-Central and West regions and losses in the East and West-Central (figure 9(d)). Overall, the greatest changes to Wetlands on private land were associated with conversions to and from Water and Cropland (figure 9(e)), and conversions with Water were most common on public land (figure 9(g)).

Zoom In Zoom Out Reset image size

3.3.7. Barren

Net change of Barren across CONUS was minimal when compared to other land cover classes but the greatest net gains were in 1988 and 1990, whereas the greatest losses were in 1991, 2004, and 2017 (figure 10(a)). On both private and public land, annual net changes varied greatly from year to year but remained very minimal (figure 10(a)). The average annual private Barren gain per year was 0.01 ± 0.01% (figure 10(b)) and loss per year was −0.01 ± 0.01% (figure 10(c)). All net gains on private land occurred in the West-Central region and net losses occurred in the West and East regions (figure 10(d)), where primary land cover conversions were associated with Grass/Shrub, Tree Cover, and Water (figure 10(e)). On public land, Barren average annual gains of 0.01 ± 0.008% per year and losses of −0.01 ± 0.007% per year occurred (figure 10(c)). Estimated regional net gains in Barren on public land were all in the West region and net losses were split nearly evenly between the East and West-Central regions (figure 10(f)). Overall, the greatest changes to Barren on public land were associated with conversions involving Water (figure 10(g)).

Zoom In Zoom Out Reset image size

3.4. Tree cover-grass/shrub-cropland-developed connections

Net conversions of Grass/Shrub to Tree Cover each year were mirrored by net conversions of Tree Cover to Grass/Shrub (figures 11(a)–(b)). Reciprocal annual conversions were similar between Cropland and Grass/Shrub (figures 11(c)–(d)). The three largest land cover conversions to Developed are from Grass/Shrub, Tree Cover, and Cropland. Private land was converted to Developed in greater area and frequency than public land (figures 11(e)–(f)). On private land, the total net conversion to Developed land is represented by 42% from Grass/Shrub, 33% from Tree Cover, and 25% from Cropland. On public land, 30% of conversions to Developed land are from Grass/Shrub, 16% are from Tree Cover, and 5% are from Cropland.

Zoom In Zoom Out Reset image size

This study provides a comprehensive assessment of how the composition of land cover change across CONUS has fluctuated from 1985–2018 at the land cover class scale. Furthermore, details of net changes and annual gross changes offer new insights to how new land cover products (i.e. LCMAP) can be used to understand land cover dynamics. With annual estimations, rather than estimations over longer intervals, this study demonstrates how regular monitoring of land cover change can enhance understanding of public and private land management. We expand upon these findings in the following section with explanations of how we address our research questions.

We find the composition of land cover differs substantially between public and private management. Private land management is dominant on a national scale, so private land management is highly representative of the whole. However, the proportions of land cover change between these two different management sectors are different. Public land shows more fluctuation in Grass/Shrub and Tree Cover, while private land shows more fluctuation in Developed and Cropland. Interpretation of any land cover change with a relatively large net change standard error indicates that there is a lot of back-and-forth gross annual change through time. To constrain the net change better and reduce the standard error, the data must include enough gains and losses to determine more precisely the balance across gain and loss. For example, if less gross change occurs, the net change standard error is lower, and thus, a small net change is quite a robust result even when changes are normalized to the proportion of the respective land management category.

To address examination of public and private lands separately, identifying a publicly available resource is a necessary first step, and we chose to use the PAD-US data because these data not only serve our purposes, but allow for expansion of this research without proprietary restrictions. The PAD-US data is an extremely valuable resource for delineation of public and privately managed lands. Without it, determination of boundaries between these categories would entail a complicated blend of federal, state, and local jurisdictional datasets sometimes requiring digitization. Private land holdings can change ownership somewhat frequently and records of private land parcels can be prohibitive because they are proprietary. A common approach when land cover data are limited and/or do not provide temporally continuous information is to examine discrete time periods of available data.

A major advantage of LCMAP products and reference data is that they provide annual coverage of land cover from 1985 to 2018. Like Auch et al (2022), this study emphasizes the strengths of the LCMAP data that show varying rates of change over time and cumulative interannual gross changes representing a more complete story of change. In this study, however, we examine publicly available land management and land cover data at a high frequency to improve our understanding of land cover composition and change under different management strategies.

Grass/Shrub is the most common land cover across both public and private land and is prolific in much of the southwestern United States and the Great Plains regions. Grass/Shrub is a highly altered terrestrial ecosystem (Henwood 2010) that is closely tied to natural resource land use cycles involving forestry and agriculture. Overall, our study indicates a net increase of 0.33% in Grass/Shrub across CONUS. Our results indicate that the greatest net changes on both private and public land occurred in the late 1980s when clearcut forest harvests were common across the western United States (Cohen et al 2002). If reforestation is not implemented after a harvest, wildfire, insect infestation, or drought-killed trees in forested areas, Grass/Shrub often becomes the dominant land cover prior to forest regrowth (Harvey et al 2014, Oswalt et al 2019). Our data analysis shows the net increase in Grass/Shrub was much greater on public land than private (1.02% versus 0.08%, respectively), attributed to conversions of Tree Cover in the West region likely from wildfire and forest harvest (Easterday et al 2018). Our analysis shows net losses of Grass/Shrub on public land were in the East region, and primarily attributed to forest regrowth. The West-Central region's widespread agriculture experienced all the net gains in Grass/Shrub on private land, predominantly from conversions of Cropland and Tree Cover. Net losses of Grass/Shrub on private land in the East-Central and the East regions were attributed to increased development, conversion to Cropland, and forest regrowth. Our data analysis shows that Grass/Shrub makes up 42% of all land that converted to Developed.

Annually, we find that gross change was greater on private land than on public land such that the average annual gains and losses of Grass/Shrub on private land were comparable to the net increase in Grass/Shrub on public land over the entire study period. Our analysis also shows that these large annual changes occurring on private land were dominated by conversions both to and from Tree cover from Forest harvest and replanting of new trees to replace the harvested volume (ex. plantation forestry in the Southeast) (Oswalt et al 2019). Our results show that Grass/Shrub on public land had little gain through the 1990s as forest harvests were scaled back with the 1994 Northwest Forest Plan (U.S. Department of Agriculture, Forest Service; U.S. Department of the Interior, Bureau of Land Management USDA and USDI 1994), but then net increases occurred in all but four years from 2000 to 2018. Finally, from 2015 to 2018 Grass/Shrub had net gains on both public and private land, which may be due to increased wildfire frequency and extent, logging activities, and reductions of cropland (Easterday et al 2018, Homer et al 2020). More detailed connections of Grass/Shrub with other land cover classes are expanded upon later in the discussion (ex. Tree Cover and Developed).

The second most common land cover across CONUS is Tree Cover, comprising 27% of private and 36% of public land. During the study period, Tree Cover had a net decrease on both private and public land, which is supported by a low standard error in both cases. Integral to the net changes were annual changes to Tree Cover (gains) and changes from Tree Cover (losses). The average gains and losses of Tree Cover on private land were much greater than on public land over the study period. Our study shows details of how the annual loss of Tree Cover to Grass/Shrub and coincident increase in Tree Cover from Grass/Shrub are related to forest harvest, wildfires, and infestations. Fluctuations in Tree Cover over time generally coincided with major economic activity across the nation, the state of the national economy, and the housing sector. Public policy changes also played a role. For example, from 1986 to 1990, LCMAP data show substantial net annual declines in Tree Cover on both private and public land. Our data analysis reflects forest harvest practices on both public and private land, especially in the western United States, which relied on clear-cutting practices throughout the 1980s (Elliott et al 2019), and disturbances from insect infestations (Harvey et al 2014) and extensive stand-clearing wildfires in publicly managed forests like what occurred in the forests of Yellowstone National Park in 1988 (Turner et al 2016, Starrs et al 2018, Vogeler et al 2020). Current wildfire and infestation regimes resulting in further decreases in Tree Cover continue to challenge forest managers (Mcdowell et al 2020). Land ownership, firefighting strategies, and reserve status may be key features in future management of Tree Cover because wildfires are predicted to occur more frequently on federal lands (Starrs et al 2018). Findings from our study highlighting how innovative new tools and data analysis examining interannual dynamics of Tree Cover change on public and private lands can be beneficial for future forest managers in this context.

Our analysis also reflects connections between development and Tree Cover. In the early 1990s, during the Savings and Loan crisis and the Gulf War, new home construction was hindered (Macrotrends 2020) from added volatility to timber markets. Our analysis shows that between 1991 and 1992, Tree Cover was decreasing on public land but increasing on private land. The 1994 Northwest Forest Plan (Spies et al 2019), combined with decreasing demand from Asian markets and increased softwood imports from Canada (Sleeter et al 2013), resulted in reduced timber harvested from public and private land in the Pacific Northwest Region (Wear and Murray 2004). Areas that had been clearcut in the northwest were still regrowing, as private tree plantations in the southeast were becoming more common (Fox et al 2004, Wear and Murray 2004), where the cycle of regrowth on pine plantations is stocked with specific endemic species is much faster than in many other forested regions of CONUS.

Land management responsibility dictates planning of silviculture intensities, rotation lengths, fire suppression, and salvage harvesting. Changing ownership of private forest land changed the way forests were harvested and managed, as many large timber companies have divested from manufacturing and forests have become a commodity in the money markets (Zhang et al 2012, Butler and Wear 2013). The downturn of the American technological industry (i.e the Dotcom Bubble) in the early 2000s came at a time of low interest rates, which kept the housing economy afloat (Sealey et al 2018) and forest harvests remained steady on private land. However, by 2008 increasingly risky mortgage lending practices eventually led to a housing market decline and a strain on the global banking system leading to the Great Recession. Our data analysis reflects a rapid net decrease in forest harvests on both public and private land as the housing market crashed. The economy rebounded and our results show a steady decrease in annual Tree Cover beginning in 2012 on both public and private land although increasing area affected by wildfire in the West is also part of this trend (Abatzoglou et al 2021).

From a regional perspective on private land, Tree Cover gains outpaced losses only in the eastern region resulting from tree planting and regrowth in the forest industry. Most Tree Cover losses on private land were distributed between the East primarily going to Developed, and the West primarily to Grass/Shrub from timber harvest, wildfire, and/or insect infestations. On public land, the West had nearly all (96%) net Tree Cover gains, all of which came from Grass/Shrub, a result of tree regrowth. The net Tree Cover losses converting primarily to Grass/Shrub on public land were distributed between the East (61%) and East-Central (39%).

Developed land is a small portion of the overall land cover across CONUS although our study indicates that the change in Developed land represents the largest net percentage gain in any land cover type. Our analysis of privately Developed land indicates two distinct rates of annual gains during different socioeconomic periods, divided by the Great Recession from 2007 to 2009. Our results show a net increase in Developed land between 1986 and 2006 nearly every year because of easy access to home mortgage funds (Sealey et al 2018) leading to a robust housing market. Although Developed land continues to see net annual increases, this rate of change falls dramatically in 2008, which is a direct reflection of the crash in the U.S. housing market and the global financial crisis. After 2009, the average rate of increase in Developed land was less than one-half the earlier rate.

The U.S. population increased more than 30% between 1990 and 2019 (U.S. Census Bureau 2021b). The rate of land development outpaced population growth betwen 1980 and 2000 (Theobald 2005) because of government-backed home mortgages and tax credits for commercial development (Sealey et al 2018). Since 2000, U.S. population growth has slowed (U.S. Census Bureau 2022), but Developed land continues to expand (Brown et al 2005, Hammer et al 2009, Theobald 2014, Bosch et al 2019) and, in most cases, the conversion to Developed land is rarely reversed. We found conversion to private Developed land came from net losses of Grass/Shrub, Tree Cover, and Cropland and net change standard error was low. From a regional perspective, net increases in private Developed land occurred in the East (42%), in the East-Central (21%), in the West-Central (19%), and in the West (18%).

Cropland experienced a net loss of −1.45% with a low standard error. Large losses of Cropland in our analysis between 1986 and 1993 and between 2000 and 2007 reflect increased enrollment in the U.S. Department of Agriculture's (USDA) Conservation Reserve Program (CRP), where marginal and other lands were removed from agricultural production and replanted with native grasses, shrub, and tree species. The CRP program experienced a rapid and substantial enrollment increase from its inception, starting with just 1,929,064 acres (7807 km²) in 1986 and reaching 35,015,042 acres (141,701 km²) in 1993 (U.S. Department of Agriculture – Farm Service Agency USDA-FSA 2021). Enrollment then declined by 15% from 1996 to 1999 but rebounded to reach peak enrollment in 2007 (36,770,984 acres or 148,807 km²). Starting in 2008, enrollment in CRP has steadily declined each year to an area 39% lower in 2018 (22,609,442 acres or 91,497 km²) than in 2008 (U.S. Department of Agriculture – Farm Service Agency USDA-FSA 2021) and much of this land returned to intensive agriculture (Morefield et al 2016). The patterns of CRP enrollment described above are evident in our data analysis and provide a plausible explanation for the interannual fluctuations of Cropland we report in this study. Thus, fluctuations of CRP enrollment describe gains and losses noted in our analysis of Cropland on private land.

Public Cropland is rare, but the 1986 North American Waterfowl Management Plan (NAWMP) opened public lands to benefit agricultural producers while serving as habitat protection, restoration, and enhancement on some National Wildlife Refuges (NWRs). The U.S. Fish and Wildlife Service's Cooperative Agriculture program allows farmers and ranchers permits to grow grain, hay or other crops, and support livestock grazing on publicly managed land (U.S. Fish and Wildlife Service USFWS 2022) (ex. Sequoyah NWR - Oklahoma, USA; Cypress Creek NWR - Illinois, USA). These mutually beneficial cooperative agreements that are uniquely designed for specific species management objectives and strategies pertaining to the NWR provide (1) profits for farmers from harvesting and selling a portion of the crop, and (2) improvement of natural habitat for species of interest like migratory waterfowl. Our analysis indicates that most of the −0.3% net loss of public Cropland during this study is attributed to conversions to Grass/Shrub or Water in the East-Central and West-Central regions.

Trade-offs between Water and Wetland are expected transitions as surface water area increases and decreases. Most of these transitions occur naturally due to precipitation patterns. Naturally, Water often transitions to Wetland before a transition to another land cover type. Wetlands can become either inundated and convert to Water or are drained and convert to a different category of land cover. Our data analysis shows that Wetlands most often convert to Water, but less frequently they dry enough to convert to Grass/Shrub, Cropland, or Tree Cover, or are Developed.

Our study finds that of the net changes to Water, 97% occurred on private land. Of this total, 86% of the Water lost on private land and 100% lost on public land was converted primarily to Barren land in the West region, which reflects prolonged drought conditions that have become more prevalent in the 2010s and beyond (Cook et al 2021). Over 50% of the roughly 894,000 km² original Wetlands in the 1780s has been lost by being drained for agricultural production, filled for development, or converted to other land cover by the 1980s (Dahl 1990). Gains in Wetlands since 1998 are attributed to abandoned or flooded Cropland (Dahl 2006). We estimate net Wetland losses were primarily on private lands and represent conversions to Developed land and inundation that converted Wetlands to Water. However, a small gain of Wetlands occurred, most as a conversion from Cropland in the West-Central region on both private and public land during wet periods, similar to Dahl's (2006) findings. When Cropland becomes inundated, the land cover may change; however, during drier periods it is possible for those lands to convert to agricultural production when private land managers are able to maximize all available land (Shrestha et al 2017). Our analysis shows logical progressions from Water loss and coinciding Wetlands gain during dry periods, to Water gain and Wetlands loss during wet periods.

Aside from substantial Water loss in the West, we find that the some of the largest gains in Water and Wetlands occurs in the West-Central region. There was no net loss of Wetlands on public land in this study. However, it is worth noting that although the standard error for Water on private land was reasonably low, the standard error on public land was very near the net change value. Standard errors for Wetland on both public and private land are very near the net change value for the study period.

Our findings indicate a net gain in Barren land cover on public land with a low standard error. All net gains of public Barren land occurred in the West region, converting primarily from Water and Tree Cover. This is likely a reflection of Water drying up and leaving Barren land behind and potentially the result of wildland fire and/or logging in areas where the understory is sparse. Net gains in Barren lands can affect wildlife, landscape ecology, and climate because Barren land tends to have higher albedo. Net losses on public Barren land occurred nearly half in the West-Central (56%) and half in the East (44%) regions mainly from Water conversion.

Our findings showed a net loss of Barren on private land. However, standard errors for private land are very near the net change value for the study period. Net gains on private land were associated with conversion to Water, Tree Cover, and Grass/Shrub in the West-Central region. The semi-arid portions of the West-Central region may be playing a role in Tree Cover to Barren transitions. The net loss on private land was either revegetated as Grass/Shrub or converted to Water. The West had the greatest proportion (60%) of net loss of private Barren land returning to Grass/Shrub, while the East region nearly consumed the rest of this transition (36%) and only a small fraction (4%) occurred in the East-Central region. This is an indication of Grass/Shrub expansion in the West, albeit for a relatively small area.

Most land cover changes occur at the local scale, but this study addresses their effects, through accumulation and how they manifest themselves at regional and national scales. Our analysis examines national- and regional-scale details of interannual land cover composition and change from eight broad land cover types. Through analysis that combined the PAD-US data with LCMAP land cover change, we explored interannual land cover dynamics on lands under public and private management at a high temporal frequency. Our results compare how public and private land cover change evolved from 1985 to 2018. Statistically attributing estimates of reference data for analysis of dynamics associated with compositional and land cover class changes enabled us to analyze differences in lands managed by public and private entities.

This study highlights patterns in land cover change across different general land management strategies and can inform policy and decision-makers about land change strategies at scales appropriate for national and regional-level mitigation, adaptation, and conservation of future natural resource management in the United States. A better understanding of how major land cover types have fluctuated on an annual frequency from 1985 to 2018 allows public and private land managers a comprehensive analysis to evaluate the effectiveness of previous policies, effects of economic volatility, and evaluation of conservation measures through time. Land use management is challenging due to the complexity of regulations and goals that can vary widely when public and private management strategies are the focus. Development of new approaches to evaluate the most current land cover data is important for refinement of future management strategies.

Our findings highlight links between the state of the economy and land cover change, which is a key component to understanding reasons for the amount of Developed land, fluctuations in Tree Cover, and Cropland expansion or contraction. We found that when the economy was strong, private development increased somewhat rapidly and resulted in the largest net percentage increase in Developed land, greater than any land cover type in this study, but not the most gross change by area. We find that Cropland decreased from 1985 to 2018 and because it is almost solely privately managed, it is more likely to be placed into conservation programs like CRP when the economy has slowed and/or incentives for such programs are enhanced. Grass/shrub dynamics will inevitably remain an important component of future natural resource management on both public and private lands because our findings show close connections with this cover type and a variety of others including Tree Cover, Cropland, and Wetlands. This study shows how examination of interannual expansion or contraction of Grass/Shrub is connected to Cropland and Tree Cover dynamics, which inevitably influences conservation and restoration management planning focused on landscape ecology. For example, expansion of Grass/Shrub on public land in the West region is an important finding from this study. Our study highlights interactions between environmental conditions and Tree Cover demography (recruitment, growth, and mortality) across CONUS since 1985 that can be expanded upon in future research. Our findings show how annual regional differences in silvicultural practices, forest succession, disturbance regimes (e.g., wildfire, insect infestations, drought), and connections to economic activity (e.g., housing market) affects net gains and losses in Tree Cover through time.

Connections between Wetlands and Cropland on both public and private land is an important finding from this study. We find that the primary drivers of these land cover types are rooted in precipitation patterns and management objectives. We found that across CONUS the largest losses in Water were on private land in the West region, which is likely due to prolonged and widespread droughts during the study period (i.e., 2010 and beyond), while the largest gains were on private lands in the West-Central region. Net Barren land gains on public lands occurred solely in the West-Central region using our sampling strategy as conversions from Tree Cover and Water, while net Barren losses on public land were split nearly equally between the West-Central and East regions where revegetation to Grass/Shrub cover was most common. Future research could use LCMAP products and reference data to investigate spatial details of aridity, drought, and climate variables to determine drivers of Barren land expansion in the West-Central region.

In summary, our analysis of new, annual land cover data shows important national and regional differences of how climate, human activity, and socio-political conditions have affected temporal changes to natural resources in the United States. Future research of public and private lands that expands our findings could include performing a finer-scale regional analysis of these variations that would provide greater detail of spatial aspects of land cover change at a variety of scales. Shifting from the LCMAP reference data to the LCMAP map data products could assist such endeavors with awareness that differences between statistical-based estimates using the LCMAP reference data and estimates derived from the LCMAP map data products are well known and documented in Stehman et al (2021).

We thank the entire US Geological Survey (USGS) USGS Land Change Monitoring, Assessment, and Projection (LCMAP) Initiative team. We appreciate Dr Danika Wellington and Mr Bruce Pengra for feedback on manuscript preparation and analysis strategies, our internal USGS peer-reviewer, Mr Daniel Sorenson, for his invaluable feedback and input, and our three anonymous peer reviewers from the journal.

The data that support the findings of this study are openly available at the following URL: https://www.sciencebase.gov/catalog/item/5e42e54be4b0edb47be84535.

Conceptualization, N C H, J L T, R F A; methodology, N C H, J L T; investigation, N C H, J L T; writing-original draft preparation, N C H, J L T; writing-review and editing, N C H, J L T, R F A. All authors have read and agreed to the final version of the manuscript.

Funding for this work was provided by the U.S. Geological Survey's National Land Imaging program. Work performed by KBR, Inc. employees was conducted under contract 140G0121D0001.

Not applicable.

Any use of trade, firm, or product names is for descriptive purposes only and does not imply endorsement by the U.S. Government.

The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

A1. Developed

Areas of intensive use with much of the land covered with structures (e.g., high density residential, commercial, industrial, or transportation), or less intensive uses where the land cover matrix includes vegetation, bare ground, and structures (e.g., low density residential, recreational facilities, cemeteries, transportation/utility corridors, etc), including any land functionally related to the developed or built-up activity.

A2. Cropland

Land in either a vegetated or unvegetated state used in production of food, fiber, and fuels. This includes cultivated and uncultivated croplands, hay lands, orchards, vineyards, and confined livestock operations. Forest plantations are considered as forests or woodlands (Tree Cover class) regardless of the use of the wood products.

A3. Grass/Shrub

Land predominantly covered with shrubs and perennial or annual natural and domesticated grasses (e.g., pasture), forbs, or other forms of herbaceous vegetation. The grass and shrub cover must comprise at least 10% of the area and tree cover is less than 10% of the area.

A4. Tree Cover

Tree-covered land where the tree cover density is greater than 10%. Cleared or harvested trees (i.e., clearcuts) will be mapped according to current cover (e.g., Barren, Grass/Shrub).

A5. Water

Areas covered with water, such as streams, canals, lakes, reservoirs, bays, or oceans.

A6. Wetland

Lands where water saturation is the determining factor in soil characteristics, vegetation types, and animal communities. Wetlands are composed of mosaics of water, bare soil, and herbaceous or wooded vegetated cover.

A7. Barren

Land comprised of natural occurrences of soils, sand, or rocks where less than 10% of the area is vegetated.

A8. Ice/Snow

Land where accumulated snow and ice does not completely melt during the summer period (i.e., perennial ice/snow)

For each year, estimated proportion of area and standard errors of the estimate of each land cover are calculated using the following equations:

$$\begin{eqnarray}&&\hat{Y}=\displaystyle {\sum }_{h=1}^{H}{N}_{h}^{* }{\bar{y}}_{h}/N\end{eqnarray} \tag{ B.1 }$$

and

$$\begin{eqnarray}&&\hat{S}(\hat{Y})=(1/{N}^{2})\sqrt{\displaystyle {\sum }_{h=1}^{H}{N}_{h}^{* 2}(1-{n}_{h}^{* }/{N}_{h}^{* }){{s}_{yh}}^{2}/{n}_{h}^{* }}\end{eqnarray} \tag{ B.2 }$$

where ${N}_{h}^{* }$ is the total number of pixels in stratum h, N is the total size of the population, ${\mathop{y}\limits^{\unicode{x00305}}}_{h}={\sum }_{u\epsilon h}{y}_{u}/{n}_{h}^{* }$ is the sample mean of the indicator ${y}_{u}$ values defined for each sample pixel $u$ contained in stratum $h,$ ${n}_{h}^{* }$ is the number of sample pixels in stratum $h,$ and $H$ is the total number of strata. The sample variance of the ${y}_{u}$ values is ${s}_{{yh}}^{2}={\sum }_{u\epsilon h}{\left({y}_{u}-{\mathop{y}\limits^{\unicode{x00305}}}_{h}\right)}^{2}/\left({n}_{h}^{* }-1\right).$ The definition of ${y}_{u}$ depends on the quantity being estimated.

Estimates for the proportion of area (${\hat{p}}_{{ij}}$) in each error matrix cell ($i,j$), the overall accuracy ($\hat{O}$), and the proportion of area (${\hat{p}}_{.k}$) of reference class k, can be computed by defining ${y}_{u}$ as follows:

$$\begin{eqnarray}&&{y}_{u}\left[{\hat{p}}_{ij}\right]\,=1\,{\rm{if}}\,{\rm{pixel}}\,{u}\,{\rm{is}}\,{\rm{map}}\,{\rm{class}}\,{\rm{i}}\,{\rm{and}}\,{\rm{reference}}\,{\rm{class}},\,0\,{\rm{otherwise}}.\end{eqnarray} \tag{ B.3 }$$

$$\begin{eqnarray}&&{y}_{u}\left[\hat{O}\right]=1\,{\rm{if}}\,{\rm{pixel}}\,{u}\,{\rm{is}}\,{\rm{classified}}\,{\rm{correctly}},\,0\,{\rm{otherwise}}.\end{eqnarray} \tag{ B.4 }$$

$$\begin{eqnarray}&&{y}_{u}\left[{\hat{p}}_{.k}\right]=1\,{\rm{if}}\,{\rm{pixel}}\,{u}\,{\rm{is}}\,{\rm{reference}}\,{\rm{classk}},\,0\,{\rm{otherwise}}.\end{eqnarray} \tag{ B.5 }$$

To estimate the net area of 1985–2018 land-cover change for each sample plot we defined ${y}_{u}$ in square kilometers (1 plot = 0.0009 km²) based on the land cover class in 1985 and the land cover class in 2018 as follows:

$$\begin{eqnarray}&&{y}_{u}=\,+0.0009\,{\rm{if}}\,{\rm{sample}}\,{\rm{plot}}\,{u}\,{\rm{changed}}\,{\rm{to}}\,{\rm{the}}\,{\rm{target}}\,{\rm{class}}\,\left({\rm{gross}}\,{\rm{gain}}\right)\end{eqnarray} \tag{ B.6 }$$

$$\begin{eqnarray}&&{y}_{u}=0\,{\rm{if}}\,{\rm{sample}}\,{\rm{plot}}\,{u}\,{\rm{did}}\,{\rm{not}}\,{\rm{change}}\,{\rm{or}}\,{\rm{was}}\,{\rm{not}}\,{\rm{the}}\,{\rm{target}}\,{\rm{class}}\end{eqnarray} \tag{ B.7 }$$

$$\begin{eqnarray}&&{y}_{u}=-0.0009\,{\rm{if}}\,{\rm{sample}}\,{\rm{plot}}\,{u}\,{\rm{changed}}\,{\rm{from}}\,{\rm{the}}\,{\rm{target}}\,{\rm{class}}\,\left({\rm{gross}}\,{\rm{loss}}\right)\end{eqnarray} \tag{ B.8 }$$

Annual land cover composition and gross change data associated with conterminous United States (CONUS), public, and private lands. The following tables are derived from data previously released by the U.S. Geological Survey (Pengra et al 2021a).