Ownership Patterns and Landscape Diversity: Conservation Implications in Maryland

Abstract

Land management decisions and conservation value are heavily influenced by land ownership, land cover, and land use. Our research aimed to examine ownership and land cover distribution, classify landowners based on land cover composition, and evaluate the ability of land cover clustering to be predictive of landowner motivations and behaviors in Maryland, USA. We tabulated a high-resolution land cover map against ownership boundaries, applied hierarchical clustering, and identified five landowner types characterized by a dominant land cover: (1) forest, (2) turf grass, (3) developed, (4) hay/pasture, and (5) crops. We analyzed a landowner survey of 3344 respondents to reveal how clusters predicted recreation, conservation, income, and other motivations. We found a skewed ownership distribution: 95.3% of smaller ownerships (<5 acres) cover 27.3% of the land, while 4.7% of larger owners hold 72.7%. Ownership patterns vary by cover, with forests and wetlands showing bimodal distributions, unimodal for cropland and hay/pasture, and turf grass concentrated in smaller properties. Survey analysis showed that crop, hay/pasture, and forest clusters had income percentages increasing with property size, with crop and hay/pasture accelerating more; conservation interest rose with size for forest and crop, but not hay/pasture; hunting motivation was highest in forest but increased with size similarly across clusters; non-hunting recreation motivation was highest in smaller hay/pasture properties, but decreased with size for all. Although each landowner has unique motivations and goals, our results reveal trends mediated by size of property and land cover that can be used to target outreach and improve conservation outcomes across Maryland’s diverse landscape.

1. Introduction

The loss of wildlife habitat continues to drive declines in various wildlife species [1,2]. Landscape-scale planning, management, and conservation are crucial for reversing these declines and maintaining a diverse biological community with high recreational value [3,4]. Private land holdings are a key component in maintaining biodiversity [5,6], and landowners have significant influence over land use and management [7,8]. Landowner actions affect key aspects such as habitat patch size and habitat quality, which drive species occurrence, biodiversity, and conservation value of the landscape [9,10].

Many regions worldwide feature a diverse, ever-changing mosaic of land cover types—from forests and wetlands to agricultural fields and urban areas—shaped by varying land uses, ownership patterns, and conservation priorities [11,12,13]. Forests, wetlands, native grasslands, and shrublands, unlike developed or turf grasslands, deliver critical ecological benefits, including higher biodiversity, enhanced habitat connectivity for species movement, and essential ecosystem services like carbon storage, erosion control, and water regulation [14,15]. Globally, land use changes primarily driven by urbanization and agricultural expansion lead to fragmented habitats and reduced biodiversity [16,17]. Further challenges include sub-optimal management of existing natural areas where invasive species and reduced structural and plant species diversity reduces their ability to support wildlife populations [18,19].

In the United States, these effects are particularly pronounced in regions like Maryland, where private land ownership dominates, with the vast majority of the state privately held [20,21]. For example, Maryland’s developed area has expanded by over 1 million acres between 1973 and 2010 [22]; forests increasingly lack age diversity needed by species such as the golden-winged warbler (Vermivora chrysoptera) and American woodcock (Scolopax minor) [18,19]. Similarly, grassland and shrubland habitats—often overlooked in development mitigation plans—represent a small fraction of the landscape and are often invaded by exotic grasses and shrubs, leading to dramatic declines in species like the northern bobwhite (Colinus virginianus), pollinators, and other grassland- and shrubland-dependent birds [23]. These trends highlight the urgent need for landscape-scale conservation practices that promote conservation-friendly development and enhanced management of forests, grasslands, and shrublands to support a broad range of species across the state.

Landowners vary considerably in their approaches to land management, ranging from whether they are individual, institutional, or government-owned. While public lands in Maryland are typically designated for long-term conservation, many face challenges such as even-aged forest stands, invasive species spread, and legacies of historical agricultural activities. Active management to improve biodiversity and conservation value can be stymied by public opposition to tools such as herbicides or timber harvests or a lack of dedicated funding. Institutional or family ownerships may similarly face challenges in reaching an agreement on actions to enhance conservation value. Individual private ownership further varies from small-scale recreational landowners to large-scale landholdings with strong income motivations, to absentee owners with limited engagement, and conservationists driven by ecological objectives [7,24]. These differences, influenced by factors such as ownership size, resource availability, and personal values, shape decisions that impact the landscape and influence habitat connectivity [25,26,27], which is critical for metapopulation persistence of wildlife [28]. For instance, Farmer et al. (2017) [26] shows that while economic incentives often attract large landowners to conservation programs, smaller owners may engage in conservation management with technical or educational support on a more local level, such as extension agencies.

In this analysis, we use a high-resolution land use/land cover dataset layered over all property ownership across the state to gather insights into how land use and ownership patterns might inform conservation strategies on the landscape. We then pair this analysis with a landowner survey to determine how land cover metrics might predict landowner motivations and behavior. This approach seeks to provide a framework to identify landowner priorities and target proactive education and conservation efforts in the study area and beyond.

Although research has found that parcel characteristics alone may provide an incomplete picture of landowner decision-making [7], many studies show that ownership characteristics such as size are significant explanatory variables [10]. Ferranto et al. (2014) [29] found that land ownership characteristics are key explanatory factors in land use and management. In addition, ownership type and size have been shown to influence land use change, especially in the context of changing commodity prices and policy [30]. Sorice et al. (2014) [27] found that large landowners often have different priorities, motivations, and resources than smallholders, affecting their willingness and capacity to implement conservation practices. Despite this work, past studies have not explicitly categorized landowners by land cover characteristics, and this analysis seeks to expand our knowledge with greater depth on how land cover might predict landowner motivations from which to build targeted outreach for landscape-scale conservation [31].

We identify patterns of land cover through hierarchical clustering that reveals distinct patterns of ownership, which can inform targeted policies and outreach efforts tailored to unique landowner profiles. This approach can be used to promote sustainable land management and educational initiatives that align with the needs and behaviors of each group [7]. For example, technical guidance will differ greatly for landowners with forest cover compared to those with cropland; additionally, their individual motivations may also be distinct [32]. By identifying these ownership classifications, we seek to enable targeted educational and land management efforts by individuals and organizations such as Cooperative Extension services, Conservation Districts, and the U.S. Department of Agriculture Natural Resource Conservation Service. Furthermore, the analysis will also enable policymakers to have a broad landscape-scale understanding of land cover and ownership to inform policy that impacts land use across the region.

By testing our analysis against a landowner survey, we provide insights into the benefits and limitations of GIS-based analysis to inform broader conservation initiatives. GIS-based analyses are attractive because they can be implemented with readily available public datasets across an entire landscape without the expense of landowner surveys, which can be costly and time-consuming to implement. We also illustrate broad land cover patterns, enabling strategic targeting of limited resources to achieve the greatest impact for landscape-scale conservation.

This analysis seeks to (1) examine ownership distribution, (2) evaluate land use/land cover distribution, (3) cluster landowners based on land cover composition, and (4) evaluate the ability of clustering paired with property size to be predictive of landowner motivations and behaviors. By improving our understanding of land cover and ownership metrics to predict landowner motivations and behavior, we aim to better understand how to support effective private land stewardship through targeted conservation strategies, landowner education, and policy-making decisions [27,33,34,35].

2. Materials and Methods

2.1. Description of the Study Area

Maryland is located in the Mid-Atlantic region of the United States (approximately 39.0458 N, 76.6413 W), with a land area of approximately 6.2 million acres. Maryland has a diverse landscape, with areas of high urbanization in the Central and Southern regions, which are surrounded by more rural forested land in the West and South as well as more agricultural land in the Central and Eastern regions (Figure 1). The majority of land in Maryland is privately owned and managed, and the state has seen increasing urbanization over recent decades [22].

2.2. Data Sources

2.2.1. Ownership Data

We obtained the ownership dataset for the state of Maryland, USA, from the Maryland Department of Planning, which includes unique owner name and address information for each property in the state. It consists of polygons representing a total of 6.199 million acres of land across 2.233 million unique parcels. This dataset closely matches Maryland State Archives estimates that Maryland comprised 6.212 million acres of land. The land area estimates can vary due to differences in how various types of land use are classified (e.g., tidal areas, public roads, rivers and water bodies, and marshy wetlands). We used this dataset in reporting the total area in private, federal, state, and county/city ownership in Figure 2 and Figure 4, Table 3.

We aggregated parcels with identical owner name and address information across the state into single properties for analysis. To avoid aggregating properties missing ownership information into a large mega-parcel, we removed landowners lacking owner name and address information, reducing the total land acres by 7.20% to 5.751 million acres, and reducing the number of unique parcels by 3.67% to 2.151 million unique parcels. After aggregating parcels by identical name and address fields, the dataset consisted of 1.861 million unique owner name and address combinations, with the total acreage remaining the same (5.751 million acres). We used this dataset in creating Figure 3.

For the cluster analysis and figures with land use calculations (Figure 5, Figure 6, Figure 7 and Figure 9, Tables 4–8) we processed the data to exclude landowners with parcels smaller than 0.5 acres due to the high computational costs associated with processing a large number of small parcels (often those found in cities) that contributed a small proportion to the landscape. This exclusion resulted in a total land area for these parts of our analysis at 5.489 million acres, and a total of 522,412 unique owners.

We summarized land ownership data by both county and region (Supplementary Materials), using the Maryland Department of Natural Resources regional map to categorize counties into Eastern, Central, Southern, and Western Maryland regions (Figure 1).

2.2.2. Land Use/Land Cover Data

The Chesapeake Conservancy, in collaboration with the U.S. Geological Survey (USGS) and the Chesapeake Bay Program (CBP), developed a high-resolution land use/land cover dataset for the Chesapeake Bay watershed region, covering 206 counties and spanning over 250,000 km². This dataset, produced in 2017–2018, features a 1 m spatial resolution and includes 18 land use/land cover categories, providing detailed insights into the landscape (

Table 1. Land uses of Chesapeake Conservancy land cover data [36].

	Land Use/Land Cover Categories
Water	Natural Succession	Impervious Roads
Impervious Structures	Impervious, Other	Tree Canopy over Impervious
Tree Canopy over Turf Grass	Turf Grass	Pervious Developed, Other
Harvested Forest	Extractive	Forest
Tree Canopy, Other	Wetlands, Riverine Non-forested	Wetlands, Terrene Non-forested
Cropland	Pasture/Hay	Wetlands, Tidal Non-forested

) [36].

2.2.3. Landowner Survey Data

To explore the potential of land use/land cover analyses to predict landowner behavior and motivations related to habitat management, we used a survey of private landowners in Maryland with properties exceeding 5 acres, capturing insights into a population responsible for owning a significant amount of land area in the state (survey available in Supplementary Materials). The survey collected data on income derived from property, interest, knowledge, and implementation of a variety of conservation practices, ownership motivations (e.g., hunting, income, privacy), wildlife management motivations (e.g., hunting, biodiversity), and barriers to conservation (e.g., labor, knowledge, etc.). A multi-modal survey (email, postcard, paper survey, follow-up) was sent to 37,818 landowners in 2024 and yielded a response rate of 8.82% with 3337 responses. All survey procedures adhered to protocols approved by the University of Maryland Institutional Review Board, ensuring voluntary participation and confidentiality.

2.3. Analytical Methods

2.3.1. Tabulation of Land Use by Landowner

We tabulated land cover area for each aggregated landowner parcel in Maryland with identical owner name and address combination. For each parcel, we extracted the number of pixels for each land use type and summed the pixel counts (1 m × 1 m) to quantify the land area of each land use within each parcel. The tabulated results provided acreage estimates of 18 land cover types for each parcel.

In our analysis, we present area estimates of land cover by ownership (Table 4, Figure 5) that are calculated from acreage found on individual properties. For example, if a 100-acre property is composed of 50 acres of forest, the forest land cover calculations presented in Table 4 and Figure 5 use only the 50 acres of forest present on that property in summary statistics and in the placement of that land cover into a size class.

2.3.2. Clustering of Landowners by Land Cover

We conducted clustering using multiple methods to determine the most effective approach for summarizing land cover across Maryland. After evaluating several techniques, we selected hierarchical clustering due to its reproducibility and ability to provide insights into the relationships between clusters. To enhance the clustering process, we aggregated functionally similar land cover classes into a simplified land use category, combining land covers such as ‘tree canopy/other’ and ‘forest’, various wetland types, and various developed land covers, to reduce redundancy and improve the algorithm’s ability to group landowners effectively (

Table 2. Land cover simplification used in clustering.

Original Land Use	Simplified Land Use
Pasture/Hay	Pasture/Hay
Cropland	Cropland
Impervious, Other	Developed
Impervious Structures	Developed
Pervious Developed, Other	Developed
Tree Canopy over Impervious Roads	Developed
Impervious Roads	Developed
Extractive	Developed
Forest	Forest
Tree Canopy, Other	Forest
Natural Succession	Successional
Harvested Forest	Successional
Turf Grass	Turf Grass
Tree Canopy over Turf Grass	Turf Grass
Wetlands, Tidal Non-forested	Wetlands
Water	Wetlands
Wetlands, Riverine Non-forested	Wetlands
Wetlands, Terrene Non-forested	Wetlands

).

We performed hierarchical clustering with the fastcluster package (version 1.3.0) in R (version 4.4.1). We used land cover proportions to group properties in a size agnostic manner with similar land cover characteristics and used Ward’s method for cluster linkage and the Euclidean distance metric to measure dissimilarity between data points. We used the silhouette method to validate the appropriate number of clusters, which determined 5 clusters as the optimal number.

To analyze the spatial structure and connectivity of the five identified clusters, we first converted polygon data into a raster format with a resolution of 30 m × 30 m and calculated landscape metrics using the landscapemetrics package (version 1.5.7) in R (version 4.4.1). Each raster cell was assigned a value corresponding to its respective cluster. A 30 m resolution was selected to balance computational feasibility and ecological relevance, as finer resolutions resulted in excessive computational demands and failed to run using available hardware and cloud computing capabilities. The rasterization process facilitated the calculation of landscape metrics, enabling quantitative assessment of the spatial structure, concentration, and dispersion of clusters. We calculated the following landscape metrics:

• Patch Density: Measures the number of patches per unit area (patches per 100 acres), indicating landscape fragmentation.
• Edge Density: Quantifies total edge length per unit area (m/acre), reflecting patch boundary complexity.
• Largest Patch Index: Represents the percentage of the landscape occupied by the largest patch, identifying dominant cluster extents.
• Clumpiness Index: Assesses spatial aggregation of patches, with values approaching 1 indicating high connectivity and -1 indicating dispersion.
• Mean Euclidean Nearest Neighbor (ENN) distance: Calculates the average distance (m) between patches of the same cluster, evaluating patch isolation.
• Cohesion Index: Measures physical connectedness of patches, with higher values (0–100) indicating greater landscape connectivity.
• Proportional Landscape Adjacency: Estimates the proportion of shared edges between patches (%), capturing spatial arrangement.
• Effective Mesh Size: Represents the average patch size (acres) if all patches were equal, assessing habitat connectivity.
• Interspersion and Juxtaposition Index: Evaluates the intermixing of cluster types (0–100), with higher values indicating greater interspersion.
• Standard Deviation and Coefficient of Variation of ENN: Measures variability in nearest neighbor distances, revealing clustering or dispersion patterns.

2.3.3. Clustering as a Predictor of Landowner Behavior

We utilized the landowner survey to assess statistically predictive characteristics of land cover clusters. The survey collected responses on landowner motivations (e.g., income, hunting, recreation), conservation practices, and perceived barriers, with variables measured on quantitative scales (e.g., percentages, Likert scales). To address missing data, we removed responses with more than half of the questions unanswered and used median imputation for remaining numeric variables. Of the five identified clusters (crop, forest, hay/pasture, turf grass, developed), we excluded turf grass and developed clusters from regression analysis due to low survey sample sizes (n = 79 and n = 9, respectively) and their small landscape proportions (6.9% and 4.1%, respectively), resulting in a dataset of crop (n = 236), forest (n = 1242), and hay/pasture (n = 546) clusters linked to survey responses.

We employed univariate linear regression models to assess the predictive ability of clusters, with an interaction term for property size, on dependent variables including landowner motivations (e.g., income, hunting, recreation), income percentage from the property, conservation interest (the mean value of responses to 18 conservation practices), and perceived barriers (e.g., wildlife damage, labor). Models were evaluated for fit using R² and significance of predictors (p < 0.05).

3. Results

3.1. Land Ownership Distribution

Maryland’s land ownership map shows how smaller parcels are concentrated in urbanized and suburban areas, while larger parcels dominate rural regions, particularly in the Western, Eastern, and far Southern parts of the state (Figure 2).

We found a highly skewed distribution of land ownership with 95.3% of all ownership being <5 acres, accounting for 27.3% of the total land area; the remaining 4.7% of landowners own ~72.7% of the land. Just over 82% of owners (1.53 million) possess properties smaller than 1 acre (Figure 3), and they collectively account for 14% of the state’s total land area (805,684 acres). As parcel size increases, the number of landowners decreases significantly, while the share of the average land area owned increases. Properties between 5 and 20 acres are owned by 3.2% of landowners but account for 14.7% of the land, while properties between 20 and 100 acres comprise 1.3% of owners and 28.1% of the total land in Maryland. At the extreme end of the scale, properties larger than 250 acres are owned by just 1084 organizations and individuals (0.1% of all owners), but those properties account for 646,493 acres or 11.2% of Maryland’s total land area.

In terms of government and private ownership, Maryland is approximately 86.2% privately owned, followed by 7.1% state-owned, 4.6% city- and county-owned, and 2.1% federally owned. As size class increases, the percentage of property owned by federal and state governments increases. County and city government ownership is spread relatively evenly between larger size classes greater than 5 acres, with a peak in the 20–100-acre size class (Figure 4,

Table 3. Land area by size class and ownership entity (private, federal, state, and city/county) in Maryland, USA. Figure was created from all sizes of ownerships before merging, including parcels with unknown ownership.

Size Class (Acres)	Private Land Area (Acres)	Percent Private	State Land Area (Acres)	Percent State	County and City Land Area (Acres)	Percent County and City	Federal Land Area (Acres)	Percent Federal	Total Land Area (Acres)	Percent All Ownership
<1	415,416	98.20%	1270	0.30%	6089	1.40%	253	0.10%	423,029	6.80%
1–3	397,076	96.90%	1991	0.50%	10,209	2.50%	564	0.10%	409,841	6.60%
3–5	253,931	95.60%	2053	0.80%	9056	3.40%	476	0.20%	265,516	4.30%
5–20	734,390	91.10%	16,929	2.10%	51,789	6.40%	3024	0.40%	806,132	13.00%
20–100	1,463,698	89.40%	68,978	4.20%	91,988	5.60%	11,713	0.70%	1,636,377	26.40%
100–250	1,272,653	89.20%	82,048	5.80%	53,967	3.80%	17,358	1.20%	1,426,026	23.00%
>250	808,796	65.70%	267,483	21.70%	61,333	5.00%	94,179	7.60%	1,231,791	19.90%
Total	5,345,962	86.20%	440,752	7.10%	284,432	4.60%	127,567	2.10%	6,198,713	100.00%

).

3.2. Land Use/Land Cover Distribution

Our analysis revealed three patterns in the distribution of major land cover classes by size class: (1) a bimodal pattern for forests and tidal wetlands, (2) a unimodal pattern for crops and hay, and (3) a left-skewed distribution for turf grass (Figure 5). The bimodal distribution in the forest and tidal wetland land cover classes shows significant portions of the total acreage concentrated in two size categories: the 20–100-acre range and the greater than 1000-acre range. We found a unimodal pattern among cropland and pasture/hay categories, represented by a curve that is highest in the middle at the acreage class of 20–100 acres. The left-skewed distribution of turf grass shows a generally decreasing percentage of land cover as size class increases. This pattern is only exhibited by turf grass and shows that holdings of turf grass are predominantly small parcels less than 3 acres, but with sizable components held in larger size classes of up to 100 acres.

Across the state, forest is the dominant land cover, with over 2.39 million acres, followed by cropland at 1.07 million acres and pasture/hay at 583,013 acres (

Table 4. Acreage and ownership characteristics in Maryland, USA by land use/land cover. Data includes all ownerships greater than 0.5 acres merged by owner name and address, excluding unknown ownership.

Land Use Type	Total Acres	Number of Owners	Average Acres	25th Percentile	Median Acres	75th Percentile	95th Percentile	Largest Land Use Ownership	Percent of Land Cover in Largest Ownership
Forest	2,392,720	301,214	7.94	0.18	0.76	2.96	27.6	20,609	0.86
Cropland	1,070,741	54,819	19.53	0.03	0.96	11.72	105.75	2702	0.25
Pasture/Hay	583,013	74,140	7.86	0.08	1.27	4.31	39.43	2407	0.41
Turf Grass	351,547	479,797	0.73	0.18	0.38	0.74	2	1253	0.36
Tree Canopy over Turf Grass	260,084	494,036	0.53	0.18	0.34	0.6	1.37	894	0.34
Wetlands, Tidal Non-forested	177,403	26,105	6.8	0.02	0.12	0.86	14.71	8948	5.04
Impervious, Other	141,178	466,431	0.3	0.02	0.06	0.14	0.96	1402	0.99
Natural Succession	116,453	110,437	1.05	0.01	0.09	0.44	3.3	1177	1.01
Tree Canopy, Other	109,478	213,299	0.51	0.01	0.06	0.3	2.17	511	0.47
Impervious Structures	68,475	461,993	0.15	0.04	0.06	0.09	0.37	348	0.51
Pervious Developed, Other	67,884	78,548	0.86	0	0.01	0.14	2.45	1937	2.85
Water	63,889	49,683	1.29	0.01	0.06	0.35	2.63	7164	11.21
Tree Canopy over Impervious	25,075	426,915	0.06	0.01	0.02	0.06	0.2	117	0.47
Wetlands, Riverine Non-forested	25,055	21,283	1.18	0.02	0.11	0.66	4.52	345	1.38
Impervious Roads	18,125	152,810	0.12	0	0.01	0.05	0.37	408	2.25
Extractive	9325	488	19.11	0.22	3.22	15.74	103.51	435	4.66
Wetlands, Terrene Non-forested	3290	5838	0.56	0.01	0.09	0.44	2.47	38	1.16
Harvested Forest	1654	257	6.44	0.11	1.22	3.72	33.96	141	8.52

). The median forest ownership is 0.76 acres, with an average parcel size of 7.94 acres, and the largest single ownership holding 20,609 acres of forest. Similar trends are observed in cropland, where the median ownership is just under 1 acre, but the largest holding is 2702 acres. Despite its smaller footprint compared to forested areas, cropland has a far larger average acreage at 19.53 compared to both forest and pasture/hay (both around 8 mean acres), signifying a relatively low proportion of small cropland ownerships on the landscape.

3.3. Cluster Analysis

The clustering analysis reveals distinct patterns of ownership groups sharing common land use and land cover characteristics. Although almost all properties consist of a mix of land covers, they are each characterized by a dominant land cover: (1) forest, (2) turf grass, (3) developed, (4) hay/pasture, and (5) cropland. Each cluster has other land covers associated with them in differing proportions (Figure 6). These clusters, derived from hierarchical clustering of aggregated land cover data, reflect varying compositions, spatial distributions, property sizes, and ownership patterns (Figure 6, Figure 7 and Figure 8).

3.3.1. Cluster Composition and Homogeneity

Each cluster is defined by a primary land cover, with varying degrees of homogeneity (Figure 6). The turf grass cluster exhibits the highest homogeneity, with 77.4% of its area comprising turf grass or turf grass under tree canopy, reflecting suburban lawns, golf courses, and recreational and civic landscapes such as sports fields and cemeteries. The cropland cluster follows closely, with 75.1% cropland, indicative of specialized agricultural use on these properties. The forest cluster, averaging 66.4% forested area, and the developed cluster, with 60.9% impervious surfaces, show moderate homogeneity. The hay/pasture cluster is the most diverse, with 54.9% hay/pasture, 31.3% forest, and 13.8% other covers, revealing a blend of agricultural and natural landscapes within this cluster.

Notably, forest is a significant secondary component across all non-forest clusters, constituting 17.9% of the cropland cluster, 31.3% of the hay/pasture cluster, and 11.9% of the developed cluster, underscoring its pervasive role even in non-forest-dominated properties. Turf grass also appears as a secondary cover in the forest (likely due to small residential lawns in wooded areas) and developed clusters (19.0%, driven by landscaping in urban settings) (Figure 6).

3.3.2. Cluster Coverage and Spatial Distribution

The clusters reveal significant disparities in total area (

Table 5. Overview of cluster characteristics. Data includes all ownerships greater than 0.5 acres merged by owner name and address, excluding unknown ownership.

Cluster Description	Total Acres	Percent of Total Acres	Number of Owners	Percent of Total Owners	Average Acres	Median Acres
Forest	3,004,443	54.7	180,132	34.5	16.7	3
Turf Grass	376,370	6.9	250,651	48	1.5	1
Developed	226,595	4.1	35,845	6.9	6.3	1.7
Hay/Pasture	909,073	16.6	37,003	7.1	24.6	5.9
Crop	973,008	17.7	18,777	3.6	51.8	11.8

). The forest cluster predominates, covering 3,004,443 acres (54.7% of the state). Forest-dominant properties have notable concentrations in Western and Southern Maryland, the Northeast part of the state, and the lower Eastern Shore. Cropland and hay/pasture clusters occupy substantial but lesser areas, at 973,008 acres (17.7%) and 909,073 acres (16.6%), respectively. Cropland is concentrated in the Eastern region, particularly the northern portion, and it extends into Central and Southern Maryland’s rural zones. Hay/pasture is prominent in a central swath north of Washington, D.C., intermingling with cropland in Central Maryland and forest in the Western region, with smaller pockets adjacent to cropland in the Southern and Eastern areas. In contrast, the turf grass (376,370 acres, 6.9%) and developed (226,595 acres, 4.1%) clusters are the smallest, predominantly located in urban and suburban areas around Washington, D.C., Baltimore, and other cities like Frederick and Hagerstown.

3.3.3. Cluster Size Class Ownership Patterns

Property size and acreage distribution vary widely across landowner clusters, reflecting contrasting management scales (

Table 6. Percentage of land cover cluster in each size class. Data includes all ownerships greater than 0.5 acres merged by owner name and address, excluding unknown ownership.

Size Class (Acres)	Crop	Developed	Forest	Pasture/Hay	Turf Grass
<1	1.4%	8.4%	3.0%	1.6%	27.6%
1–3	4.5%	19.7%	9.4%	6.0%	42.6%
3–5	1.7%	8.8%	5.3%	4.7%	10.6%
5–20	8.4%	31.1%	20.2%	17.8%	11.3%
20–100	33.8%	17.4%	32.0%	35.6%	3.8%
100–250	38.9%	5.3%	15.0%	26.3%	2.8%
>250	11.3%	9.2%	15.1%	8.0%	1.4%

, Figure 9). The cropland cluster is highly concentrated in larger holdings, with a total of 973,008 acres, of which over 70% is held in parcels of 20 acres or larger. The largest portion is in the 100–250-acre size class (378,494 acres), followed by substantial areas in the 20–100-acre (329,101 acres) and >250-acre (109,806 acres) categories. Only 7.6% of cropland area falls under 5 acres, reinforcing cropland’s strong association with larger ownerships.

The pasture/hay cluster spans 909,073 acres and shows a more balanced size class distribution. While 35.6% lies in parcels over 100 acres, nearly one-third is held in the 20–100-acre class (323,681 acres), with meaningful portions in both smaller (5–20 acres: 161,516 acres) and very large properties (>250 acres: 73,161 acres). This pattern suggests a mix of mid-sized family operations and larger grazing/haying lands.

The forest cluster is the most spatially extensive, occupying high coverage across a broad range of parcel sizes. Its distribution is notably even, with 32% of forest acreage in the 20–100-acre range (961,459 acres), and substantial areas in both 100–250 acres (451,561 acres) and >250 acres (453,548 acres). Importantly, smaller holdings remain common: 28% of forestland lies in parcels between 1 and 20 acres, indicating the prevalence of forest ownership in all size class scales.

In contrast, the turf grass cluster is characterized by small residential holdings, totaling 376,379 acres, with over 69% of that area (264,030 acres) found in parcels smaller than 5 acres. The dominant size class is 1–3 acres (160,236 acres), followed by <1 acre (103,793 acres). Properties over 20 acres are rare (<8%), underscoring the cluster’s association with dense, suburban-style land ownership.

The developed cluster, covering 226,595 acres, shows a broader range of parcel sizes than turf grass but remains skewed toward small-to-mid-sized parcels. Nearly half of its acreage (70,579 acres) falls within the 5–20-acre range, and a significant amount lies in the 20–100-acre category (39,365 acres). The developed cluster blends suburban expansion with large-lot development or commercial land holdings.

Cropland is the most consolidated, turf grass is the most fragmented, and forest is the most evenly distributed across all size classes. Hay/pasture holdings occupy an intermediate position, and developed lands reflect the structure of built environments occupying small- to medium-size ownerships.

3.3.4. Connectivity of Clusters

We examined the spatial and structural characteristics of Maryland’s land cover clusters, and identified three key themes: the large, cohesive, and structurally connected landscapes of forest and cropland clusters; the fragmented and spatially isolated configurations of turf grass and developed clusters; and the moderately fragmented hay/pasture cluster.

The forest and cropland clusters remain among the most structurally connected landscape types in Maryland (

Table 7. Patch and edge characteristics for landowner clusters in Maryland. Data includes all ownerships greater than 0.5 acres merged by owner name and address, excluding unknown ownership.

Cluster	Patch Density (#/100 Acres)	Edge Density (m/Acres)	Largest Patch Index (%)	Clumpiness Index	Interspersion Juxtaposition Index (%)
Forest	0.59	9.02	5.50	0.87	94.0
Turf Grass	1.12	5.14	0.08	0.72	75.0
Developed	0.38	1.74	0.11	0.82	79.2
Hay and Pasture	0.23	5.02	0.42	0.91	78.0
Crop	0.13	3.60	0.68	0.93	78.9

and

Table 8. Connectivity, diversity, and variability characteristics for landowner clusters in Maryland. Data includes all ownerships greater than 0.5 acres merged by owner name and address, excluding unknown ownership.

Cluster	Mean Euclidean Nearest Neighbor (ENN) Distance (m)	Cohesion (%)	Proportional Landscape Adjacency (%)	Effective Mesh Size (Acres)	Standard Deviation of ENN (m)	Coefficient of Variation of ENN (m)
Forest	111.9	99.6	93.9	41,409.6	100.7	90.0
Turf Grass	132.7	92.7	73.7	16.1	156.2	117.8
Developed	207.6	95.3	82.7	31.1	364.0	175.4
Hay and Pasture	235.3	98.2	92.4	411.9	379.2	161.2
Crop	265.8	99.0	94.4	1,244.7	498.2	187.4

). Both exhibit extremely high cohesion (99.6% for forest, 99.0% for cropland), indicating the presence of large, internally connected land areas. The effective mesh size is especially high for the forest cluster (41,410 acres), with cropland cluster following at 1245 acres. These values suggest that properties associated with high forest cover provide by far the largest expanses of uninterrupted land cover, while cropland also supports substantial contiguous land areas.

Despite this structural connectedness, cropland patches are more isolated spatially. The cropland cluster has the highest mean Euclidean Nearest Neighbor (ENN) distance (265.8 m), as well as the highest variability (standard deviation = 498.2 m; coefficient of variation = 187.4), highlighting large and inconsistent spacing between crop cluster patches. In contrast, forest cluster patches are on average more proximate (mean ENN = 111.9 m) with lower variability (CV = 90.0), indicating a tighter and more predictable spatial configuration. This reinforces the role of forests as a spatially connected matrix, with the lowest variability of distances among any other cluster type.

Edge and shape complexity metrics further differentiate these clusters. The forest cluster has higher edge density (9.02 m/acre) and a higher interspersion and juxtaposition index (IJI = 94.0%) than cropland (edge density = 3.60 m/acre; IJI = 78.9%), reflecting greater boundary complexity and greater mixing with adjacent land cover types. Cropland, by contrast, exhibits simpler patch shapes and is more spatially segregated from other clusters.

The turf grass and developed clusters exhibit the most fragmented spatial configurations. The turf grass cluster has a relatively high patch density (1.12 per 100 acres) and the smallest effective mesh size (16.1 acres), underscoring its highly broken nature. Mean ENN distance is moderately high (132.7 m), and with a high coefficient of variation (117.8), patch spacing is both irregular and moderately dispersed. Turf grass also has the lowest clumpiness index (0.72), suggesting a more scattered spatial arrangement than other clusters.

The developed cluster shows the second-lowest patch density (0.38), but has the second-smallest mesh size (31.1 acres) and a relatively high mean ENN distance (207.6 m). Despite low edge density (1.74 m/acre), its cohesion remains relatively high (95.3%), indicating that while developed areas tend to form compact clusters, these are not well connected spatially.

The turf grass and developed clusters exhibit lower PLA values (73.7% for turf grass, 82.7% for developed) than other clusters, indicating fewer like adjacencies and greater mixing with diverse land cover types due to higher fragmentation in urban/suburban areas.

The hay/pasture cluster occupies a middle ground between the structural extremes of forest/cropland and turf/developed areas. With a moderate mean ENN distance (235.3 m) and relatively high cohesion (98.2%), this cluster exhibits intermediate connectivity. The mesh size (412 acres) supports this interpretation, indicating that hay/pasture parcels are moderately sized but not highly fragmented.

The clumpiness index (0.91) and PLA (92.4%) for hay/pasture indicate spatial cohesion, while the IJI (78.0%) reflects a notable level of intermixing with adjacent land cover types. Edge density (5.02 m/acre) is moderate, and patch density (0.23 per 100 acres) remains relatively low, suggesting that patches tend to be larger but less frequent.

Overall, forest and cropland clusters provide the core spatial anchors for landscape connectivity in Maryland, while hay/pasture serves as a structurally intermediate but ecologically significant matrix. Developed and turf clusters represent more fragmented zones on the landscape.

3.4. Cluster Prediction of Landowner Motivations and Behavior

To evaluate land cover clustering as a tool for understanding landowner behaviors, we analyzed survey responses against the 5-cluster land cover classification above. Due to small sample sizes, turf grass and developed clusters were excluded. The survey assessed percent income derived from the property, the mean interest, knowledge, and implementation scores for a variety of conservation practices, ownership motivations (e.g., hunting, income), wildlife management motivations (e.g., endangered species, biodiversity, wildlife viewing), and barriers to conservation (e.g., increasing wildlife damage, labor, equipment) (

Table 9. Regression results for landowner motivations and behaviors by cluster. Regression results examining the effects of cluster membership (crop, forest, with hay/pasture as the reference) and property size (acres) on landowner motivations and behaviors. The ‘economic activities: hunting’ model uses logistic regression (coefficients are log-odds), while all other models use OLS (linear regression). Turf grass and developed clusters are excluded due to small sample sizes.

Regression Results for Landowner Motivations and Behaviors by Cluster
	OLS	OLS	logistic	OLS	OLS	OLS	OLS	OLS	OLS	OLS	OLS	OLS
	Income (%)	Own Reason: Income	Econ Act: Hunting	Mean Interest	Own Reason: Hunting	Wildlife Reason: Hunting	Own Reason: Non-Hunting Recreation	Own Reason: Privacy	Barrier: Damage	Barrier: Priorities	Barrier: Labor	Barrier: Equipment
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)	(12)
Forest	−1.119	−0.309 **	0.463 *	0.028	0.617 ***	0.389 **	−0.762 ***	0.223 **	−0.209	−0.478 ***	−0.080	0.208
	(1.005)	(0.151)	(0.241)	(0.123)	(0.177)	(0.164)	(0.149)	(0.114)	(0.146)	(0.148)	(0.154)	(0.168)
Crop	7.750 ***	1.028 ***	1.088 ***	−0.595 ***	0.521 *	0.076	−1.062 ***	0.286	0.882 ***	0.232	−0.565 **	−0.259
	(1.651)	(0.249)	(0.314)	(0.202)	(0.290)	(0.269)	(0.244)	(0.187)	(0.239)	(0.243)	(0.253)	(0.276)
Acres	0.171 ***	0.013 ***	0.007 ***	−0.001	0.004 ***	0.002	−0.008 ***	−0.002	0.006 ***	0.005 ***	−0.002	−0.004 **
	(0.009)	(0.001)	(0.001)	(0.001)	(0.002)	(0.001)	(0.001)	(0.001)	(0.001)	(0.001)	(0.001)	(0.001)
Forest:Acres	−0.130 ***	−0.009 ***	−0.001	0.002 *	−0.0002	0.001	0.005 ***	0.0003	−0.006 ***	−0.004 ***	0.0004	0.001
	(0.010)	(0.001)	(0.002)	(0.001)	(0.002)	(0.002)	(0.001)	(0.001)	(0.001)	(0.001)	(0.002)	(0.002)
Crop:Acres	−0.053 ***	−0.005 ***	−0.003	0.003 **	−0.001	0.0005	0.003 *	−0.001	−0.002	−0.001	0.001	0.0004
	(0.012)	(0.002)	(0.002)	(0.001)	(0.002)	(0.002)	(0.002)	(0.001)	(0.002)	(0.002)	(0.002)	(0.002)
Constant	3.322 ***	4.848 ***	−3.009 ***	6.144 ***	5.627 ***	6.056 ***	8.142 ***	8.817 ***	3.199 ***	4.003 ***	6.624 ***	5.834 ***
	(0.854)	(0.129)	(0.212)	(0.105)	(0.150)	(0.139)	(0.126)	(0.097)	(0.124)	(0.126)	(0.131)	(0.143)
Observations	2024	2024	2024	2024	2024	2024	2024	2024	2024	2024	2024	2024
R2	0.339	0.139		0.011	0.029	0.019	0.059	0.012	0.058	0.040	0.008	0.021
Adjusted R2	0.337	0.137		0.008	0.027	0.017	0.057	0.009	0.056	0.037	0.006	0.018

* p < 0.1; ** p < 0.05; *** p < 0.01.

). Four variables—percent income of the property, conservation practice interest, hunting and non-hunting recreation as a reason for ownership—showed significant relationships with clusters, influenced by property size (Figure 10). However, clustering often failed to predict other motivations and behaviors, suggesting limitations in using land cover alone to capture complex landowner dynamics.

3.4.1. Property Income

Landowners in the crop cluster reported higher baseline income percentages from their properties than the hay/pasture reference group (β = 7.75, p < 0.001). Although the percentage of income increases with property size, the negative interaction term reveals diminishing returns at larger scales compared to the hay/pasture reference cluster (β = −0.053, p < 0.001). Forest landowners exhibited slower income growth with size (β = −0.130, p < 0.001). This model had the highest explanatory power (R² = 0.336) of all variables tested. Confirming their economic focus, crop landowners also valued income motives more strongly (β = 1.03, p < 0.001). Additionally, hunting as an economic activity was positive for crop (β = 0.081, p = 0.004) but not significant for forest (β = 0.031, p = 0.067) or hay/pasture clusters.

3.4.2. Conservation Interest

Landowners in the crop cluster showed lower baseline conservation interest than hay/pasture (β = −0.595, p = 0.003), but interest increased with property size, surpassing hay/pasture at approximately 200 acres (β = 0.003, p = 0.029). In contrast, hay/pasture exhibited a negative (but non-significant) slope with property size (β = −0.001, p = 0.473), an anomaly compared to crop and forest cluster positive trends (crop: β = 0.003, p = 0.029; forest: β = 0.002, p = 0.079), suggesting neutral to declining interest on larger hay/pasture properties in conservation. Conservation knowledge and implementation showed no significant cluster differences (R² < 0.02), indicating limited predictive power of land cover clusters.

3.4.3. Recreation Motivation

Landowners in the forest cluster prioritized hunting as a reason for ownership more than hay/pasture (β = 0.617, p < 0.001), an effect amplified by property size (β = 0.004, p = 0.007). They also expressed stronger hunting-related reasons for managing wildlife (β = 0.389, p = 0.018). Interests in biodiversity, endangered species, and wildlife watching showed no significant cluster differences (R² < 0.01).

Hay/pasture landowners showed higher non-hunting recreation motives, potentially linked to activities like horseback riding, compared to crop (β = −1.062, p < 0.001) and forest cluster landowners (β = −0.762, p < 0.001), indicating a stronger preference for non-hunting recreation in this cluster. However, this preference decreases with property size for hay/pasture landowners (β = −0.008, p < 0.001) more rapidly than for forest and crop clusters.

3.4.4. Other Motivations and Barriers

Forest cluster landowners valued privacy more than those in the hay/pasture cluster (β = 0.223, p = 0.0498), while motivations like family, beauty, and investment showed no significant variation (R² < 0.01). Wildlife management interests such as biodiversity, endangered species, and wildlife viewing were not significantly predicted by clusters (R² < 0.01). Crop cluster landowners reported greater concern for property damage from increased wildlife habitat (β = 0.882, p < 0.001), while forest cluster landowners perceived fewer barriers related to competing priorities (β = −0.478, p = 0.001). Crop cluster landowners viewed labor as less of a constraint (β = −0.565, p = 0.024). Equipment barriers decreased with property size across clusters (β = −0.005, p < 0.001), suggesting larger properties face fewer equipment constraints, while barriers like knowledge, time, and cost showed little significance by cluster (R² < 0.02).

4. Discussion

This analysis of land ownership by land cover informs our understanding of the landscape at the scale of where management decisions are made: the property owner. Understanding size class distributions, how land cover is distributed across property size, and uncovering distinct typologies of properties reveal characteristics that can inform landscape planning, conservation priorities, and educational efforts.

While parcel characteristics such as size and land cover alone offer only a partial view of landowner decision-making [7], they consistently serve as key explanatory factors in defining landowner typologies [7,29,37,38]. In the case of agriculture, these attributes can highlight economic drivers known to influence land use changes [39,40,41]. Our analytical approach confirms prior findings that land cover alone can be limited in its predictive capacity, but reveals that income, recreational motivations, and interest in conservation practices vary significantly by dominant land cover characteristics and property size. These findings, albeit somewhat limited, are nonetheless useful as they reveal trends that can be extrapolated across the entire population of landowners using public datasets without the time, expense, and limited scale of landowner surveys.

With increasingly accurate and accessible land ownership polygons and ever-higher-resolution land cover data, this analysis provides a model for aligning landscape-scale conservation and planning with the needs and behaviors of distinct landowner groups, enhancing the effectiveness of conservation and land use planning strategies.

4.1. Land Ownership Distribution

This analysis reveals a land ownership distribution pattern fundamental to landscape-scale conservation consistent with findings in other parts of the United States: a minority of large-acreage property owners comprise the majority of land area, juxtaposed by a large majority of small landowners owning a minority of land area [41,42,43,44]. In Maryland, where 86.1% of land is privately owned, this pattern underscores the critical role of private landowners in shaping the landscape. Although government ownership (federal, state, and county/city) increases in larger size classes—peaking in the 20–100-acre class for county/city and >250 acres for federal and state holdings (Figure 4, Table 3)—private land still constitutes the vast majority, even among these larger parcels, accounting for 82.5% of land in size classes above 20 acres.

The high proportion of Maryland’s landscape found in the 20–100-acre size class suggests a trend toward market forces that favor medium-sized farms or estates. It may also reveal Maryland’s status in a process moving toward an increasingly fragmented ownership landscape. With the generational transfer of property, larger properties are often subdivided between children, leading to increasingly smaller properties over time when land is split up and sold to divide assets upon inheritance [45,46]. Although aggregation of property into larger properties has occurred in certain regions of the U.S. [47], in Maryland, the trend in recent decades is towards an increasingly fragmented landscape [48,49]. Understanding the current status of land ownership fragmentation allows for enhanced targeting of education or policy initiatives that might reduce fragmentation or allow improved educational targeting for particular types of owners.

Our results show that property size influences landowner income, with larger properties showing increased percentages of their income from property (Figure 10), confirming the literature’s observation that larger properties often prioritize production-oriented activities like agriculture or forestry [38,50]. Similarly, our finding that landowners in all clusters demonstrate increased hunting motivation with property size lends support to research findings that larger properties engage in wildlife habitat enhancement, as hunters seek to develop abundant game on their properties [26,37,51].

Conservation interest also varies with property size, increasing for landowners in the crop and forest clusters but not increasing for those in the hay/pasture cluster, highlighting differing priorities by land cover as properties grow larger. This partially aligns with the literature’s finding that larger landowners are more likely to participate in environmental improvements, such as habitat enhancement or water quality management, but underscores how some properties may face competing demands like non-hunting recreation or livestock production [37,51]. Although conservation interest aligns with previous research on the implementation of environmental improvements, our results detected no trends in the knowledge levels or implementation of conservation practices with property size across clusters, suggesting that factors beyond size may play a larger role in implementation practices.

Equipment barriers decrease with property size across all clusters, confirming that smaller properties face greater resource constraints, which may limit their conservation actions [37,51,52]. These dynamics highlight that conservation efforts must prioritize unique constraints of private landowners across size classes, tailoring strategies to leverage landowner differences. One key finding is that practical guides and policy incentives can be oriented to the implementation of best management practices on large properties, but broad-based educational efforts about the benefits of various management practices can inform the wider population who own small properties about the importance of land management practices (such as prescribed fire or the use of herbicides to control invasive vegetation) and inform policy development. One policy example targeting residential landowners is a 2021 Maryland law (House Bill 322) that prevents homeowners’ associations (HOAs) from prohibiting “low-impact landscaping” that includes pollinator gardens, native plants, and rain gardens, enabling the expansion of wildlife and pollinator habitats into more developed areas [53]. Public education and outreach programs can use analyses like this to target efforts regionally to land uses and landowner size classes that will yield the greatest impact with limited resources.

4.2. Land Use/Land Cover Ownership Distribution

While property size is an important explanatory variable, incorporating land use and land cover characteristics by ownership size reveals more nuanced strategies for conservation, policy, and educational efforts. The bimodal pattern of ownership by size class for forested and tidal wetland parcels (Figure 5a) suggests a two-pronged approach targeting very large properties (>1000 acres), and medium-sized (20–100-acre) properties might yield improved efficiency in changing management. The unimodal pattern of cropland and pasture/hay land suggests focused efforts targeted to properties between 5 and 100 acres may be more relevant and reach desired audiences more directly, with further distinction of an orientation to smaller landowners for hay/pasture (due to its smaller mean size of 7.8 acres), and larger landowners for cropland education (due to its average size of 19.5 acres).

Although turf grass and turf grass under tree canopy comprise just under 10% of the land area in the state (Table 4), approximately 48% of landowners (Figure 6) fall into the turf grass-dominant cluster, the largest percentage of any other land cover category. Even though they are small in overall area and patch size, research has found that when managed properly, small patches of habitat have special wildlife benefits, including serving as a foraging or breeding habitat for a variety of species, both vertebrates and invertebrates, and facilitating wildlife movement through fragmented landscapes [54,55,56]. These findings show that smaller landowners can still play an important role in providing conservation benefits on the landscape.

The conversion of turf grass to habitats like native meadows and wildflowers could benefit a suite of wildlife and pollinator species in the more developed areas of the state by providing increased vegetation diversity and habitat connectivity [23,57,58,59]. To further encourage and advance these benefits, future research could explore how urban and suburban turf grass patches, such as lawns, parks, and road verges and islands, can be managed to improve pollinator and wildlife habitats and support biodiversity in fragmented landscapes. For example, studies could investigate practices like planting native wildflowers or reducing mowing frequency in urban settings to increase pollinator habitat and populations, complementing our findings on turf grass dominance and landowner motivations. Due to the high number of owners with turf grass, educational efforts can also help foster a broader public understanding of best practices for conservation. Targeted extension programs and financial incentives can effectively promote behavioral change and encourage the adoption of ecologically beneficial turf grass management practices [26,60].

4.3. Cluster Analysis

The cluster analysis reveals patterns of land ownership by land cover at the state scale. Because nearly all properties in Maryland contain a variety of land covers, categorizing landowners into discrete categories composed of similar land cover characteristics assists with understanding land use and land cover in the state and how it can be best managed.

The clustering analysis reveals distinct spatial patterns among landowners in Maryland, revealing differences in land cover mixes, landscape structure, fragmentation, and connectivity. The forest and cropland clusters, which together make up 72.4% of the state’s land area, exhibit contrasting spatial arrangements. Forest-dominant parcels on average are smaller than crop-dominated properties (mean 16.7 acres and 51.8 acres, respectively) and are more evenly distributed across the state. Conversely, properties in the crop cluster have relatively high connectivity, but greater isolation into core regional hubs of activity (ENN = 265.8 m). This differentiation has important implications for forest and agricultural management as well as conservation planning [28]. In particular, it suggests that forests are dispersed more widely, making planning for forest management a statewide effort, while cropland areas tend to be more regional, suggesting that cropland management work can be focused on regions with high agricultural activity.

The prevalence of forests within the other clusters—accounting for 31% of the average land cover in hay/pasture, 18% in cropland, and 12% in developed—further underscores the need for targeted forest management education across a diverse array of Maryland landowners, each with distinct goals and primary land cover types. These landowners, ranging from agricultural producers to suburban property managers, play a pivotal role in maintaining a forest habitat that supports declining forest-dependent species such as wood thrush (Hylocichla mustelina), Eastern whip-poor-wills (Antrostomus vociferus), American woodcock (Scolopax minor), and ruffed grouse (Bonasa umbellus) in Western Maryland, whose populations have declined due to habitat loss and fragmentation [61,62,63].

Encouraging landowners to implement forest management practices that promote structural diversity, reduce invasive species, and establish a diversity of age classes within forests can significantly enhance habitat quality, connectivity, metapopulation persistence, and biodiversity [28,64,65,66]. For instance, creating early successional vegetation types through practices such as timber harvesting and prescribed burning can be crucial for species like the ruffed grouse and American woodcock, which depend on young forests for parts of their life cycle. These efforts can also complement efforts to conserve grassland birds like the northern bobwhite (Colinus virginianus), which can utilize young regenerating forests as habitat [67,68]. Invasive plant control is also essential, as these species can reduce native biodiversity through competition and disrupt habitat structure [69]. Additionally, maintaining or creating age and species diversity in forests builds forest stand resilience to stressors that specific age classes or species may face and can provide increased ecosystem services for the surrounding landscape [70,71]. Educating landowners on these strategies can help them make informed decisions that benefit the forest landscape of Maryland.

The developed and turf grass clusters are the most fragmented and dispersed, reflecting the patchwork nature of suburban and urban landscapes and development. Higher patch densities (0.38 and 1.12) and lower clumpiness (0.82 and 0.72) values indicate that these landscapes are composed of small, scattered parcels and are heavily intermixed with other land covers. Their lower connectivity and small effective mesh sizes (31.1 and 16.1 acres) suggest they are less contiguous across the landscape, but they can still support highly localized ecological processes [54]. Given their coverage of ~10% of state area and high abundance of landowners, these areas provide high opportunities for developing a culture of conservation and management among the broader population.

The hay/pasture cluster falls between the extremes of the cropland/forest and turf grass/developed clusters, exhibiting moderate connectivity and fragmentation. From a conservation perspective, properties in this cluster are particularly important for conservation outreach due to the strong negative impacts that hay cutting can have on wildlife [72,73]. Hay cutting often coincides with the breeding season of many wildlife, which leads to destroyed nests and mortalities of herpetofauna, and deer fawns caught in cutting blades [74,75]. For wildlife that survive cutting, they are left vulnerable to predation due to the loss of vegetative cover, as predators are known to follow behind hay-cutting machinery, preying on exposed wildlife [73,76,77,78].

Representing 16.6% of Maryland’s landscape, improved management of hayfields to wildlife-friendly practices can yield benefits for breeding grassland birds, which are declining faster than any other bird guild [23,78]. Delaying hay cuts on hay and pasture fields between July 1st and 15th allows the majority of grassland bird species to complete their nesting cycles, thereby reducing nest destruction and increasing fledgling survival rates [76,79]. Gruntorad et al. 2021 [80] found that most private landowners in the Great Plains region would be willing to delay hay cutting for bird conservation, showing that this strategy could also be successful among landowners in Maryland. Additionally, transitioning hay operations to native warm-season grasses such as big bluestem (Andropogon gerardi) and Indiangrass (Sorghastrum nutans), as well as maintaining the residual height of this vegetation at 18–24 inches, can provide structurally suitable habitats that fulfill the nesting and foraging requirements for various grassland bird species while retaining higher haying value in a later harvest scenario [81,82]. Implementing these management techniques could turn hay and pasture fields into population sources rather than sinks, enhancing the habitat for species like Eastern Meadowlarks (Sturnella magna), Bobolinks (Dolichonyx oryzivorus), Grasshopper Sparrows (Ammodramus savannarum), Northern Bobwhites (Colinus virginianus), and Henslow’s Sparrows (Centronyx henslowii) [77].

4.4. Landowner Survey by Cluster

The distinct profiles of landowners in the crop, hay/pasture, and forest clusters reveal unique motivations and barriers that necessitate tailored conservation outreach strategies. Below, each cluster is discussed individually to highlight their specific characteristics and similarities, followed by an integrated discussion of cross-cluster strategies to enhance conservation adoption.

Landowners in the crop cluster exhibit a strong income orientation that increases with property size, alongside a positive trend in conservation interest as property size grows (β = 0.003, p = 0.029). A significant barrier to conservation adoption is their concern about property damage from wildlife, primarily deer (β = 0.882, p < 0.001) [83], highlighting a trade-off between economic motivations like crop income and conservation goals. To address this, economically neutral practices, financial incentives, and wildlife damage mitigation tools can balance these priorities, effectively encouraging conservation adoption among crop cluster landowners. The crop cluster shares its size-related income increase with the hay/pasture cluster, but its increasing conservation interest with property size suggests that increasing interest levels with property size may offset the need for economic incentives.

Hay/pasture landowners exhibit increasing percentage of income with property size, similar to the crop cluster, but their interest in conservation practices does not rise with larger properties (β = −0.001, p = 0.473), diverging from trends in the crop and forest clusters where conservation interest grows (β = 0.003, p = 0.029; β = 0.002, p = 0.079, respectively). These landowners also place a higher priority on non-hunting recreational activities, such as horseback riding, compared to the crop (β = −1.062, p < 0.001) and forest clusters (β = −0.762, p < 0.001), though this preference decreases with property size (β = −0.008, p < 0.001), suggesting that larger hay/pasture properties may prioritize other uses like grazing or hay farming over recreation. Conversely, non-hunting recreation motives do not decrease as rapidly with property size for crop (β = 0.003, p < 0.1) and forest clusters (β = 0.005, p < 0.001) (Figure 10). This non-hunting recreational focus in smaller hay/pasture properties may compete with conservation priorities, underscoring the need for targeted outreach that highlights opportunities for integrating wildlife-friendly practices with non-hunting recreation, such as creating trails that support both horseback riding and conservation.

Landowners in the forest cluster are more motivated by hunting and privacy (β = 0.617, p < 0.001), distinguishing them from the income-oriented crop and hay/pasture clusters. Their conservation interest increases with property size, similar to the crop cluster, making hunting-compatible conservation programs particularly effective. The forest cluster’s hunting priority suggests opportunities for education about habitat improvements to increase game species abundance that can also benefit non-game wildlife.

The diverse motivations and barriers across the crop, hay/pasture, and forest clusters highlight trade-offs that shape conservation outcomes. Crop and hay/pasture landowners balance economic priorities like crop, haying, or grazing income against conservation goals, often hindered by barriers such as wildlife damage. Forest landowners navigate trade-offs between privacy preferences and conservation efforts that could benefit from collaboration. By tailoring strategies to each cluster’s unique profile—addressing economic concerns for crop and hay/pasture clusters through incentives and mitigation tools, integrating recreation with conservation for hay/pasture, and leveraging hunting motivations for forest—conservation programs can maximize engagement and adoption across Maryland’s diverse landowner groups.

4.5. Limitations to Study

Although we believe this study offers the best analysis of land cover in the state to date, the land use/land cover dataset nonetheless includes classification errors. The accuracy assessment of the database found that Maryland was mapped with an average of approximately 90% [84]. Certain land covers are more prone to misclassification than others. Wetlands had the most significant land cover error, with an accuracy assessment of 80%. We also found some examples in the dataset where golf courses and other larger areas of turf grass were sometimes misclassified as pasture/hay or natural succession. Because of the similarities between these land cover types, they are more prone to misclassification in the dataset than other land covers, such as forests, which are more distinguishable. As a result, we believe our findings are accurate to approximately 80–90%. Despite these inaccuracies, we maintain that this analysis represents a reliable summary of land cover and its interaction with land ownership.

The connectivity analysis required the conversion of polygon data into a raster format with a 30 m resolution, as finer resolutions (e.g., <30 m) exceeded the processing capacity on available hardware and software, including cloud computing tools (see Methods). This approach may lead to errors in connectivity values on the landscape.

Although the clustering algorithm provided useful insights into grouping landowners, it grouped properties into clusters that may not be dominated by a particular land cover type but exhibited enough similar composition to other properties to warrant grouping with a particular cluster. Results should be interpreted with the understanding that a diversity of land covers exist within a cluster and may diverge from the averages presented (Figure 7).

Additional limitations arise from data processing choices detailed in the Methods section. To avoid aggregating properties’ missing ownership information into a large mega-parcel, we removed landowners lacking owner name and address data, reducing the total land area by 7.20% to 5.751 million acres and the number of unique parcels by 3.67% to 2.151 million. For the cluster analysis, we further excluded parcels smaller than 0.5 acres due to high computational costs associated with processing numerous small urban parcels that contributed minimally to the landscape, reducing the total land area by an additional 4.5% to 5.489 million acres and the number of owners by 71.9% to 522,412. The exclusion of missing ownership data primarily affected large forested rural parcels—often government-owned—in Western Maryland. The exclusion of parcels less than 0.5 acres removed many small urban/suburban parcels. Including these areas would likely slightly increase the representation of forested land in large rural properties and increased impervious surfaces and turf grass in small properties, potentially altering cluster composition slightly, though we expect no significant changes to the overall findings.

Our approach to handling missing data—excluding respondents with less than 50% of survey answers completed and imputing median values for the remaining respondents with more complete responses—may influence results, though we minimized this impact by preserving a robust sample size and restricting imputation to variables with high completion rates.

5. Conclusions

The findings indicate that private landowners, regardless of property size, significantly shape Maryland’s landscape, as 86.1% of the state is privately owned. The high proportion of small landowners (<1 acre, 80% of owners) indicates a need for outreach and education tailored to this demographic to inform the broader public of priorities related to landscape conservation. Although these landholders collectively occupy a relatively small percentage of the total acreage (6.8%), their cumulative impact on conservation policy is substantial due to the high proportion of the population.

This study reinforces the importance of integrating ownership data with land cover analyses so that policymakers and conservation organizations can better understand the cultural and economic motivators of different landowner groups, which can lead to more effective engagement strategies.

This study reveals that education and conservation organizations like the Cooperative Extension Services and NRCS can help enhance wildlife habitat and ecosystem resilience through tailored conservation and educational strategies that align with the distinct economic, recreational, and privacy priorities of landowners to promote sustainable land management across a diverse landscape.