Environment Systems and Decisions

, Volume 36, Issue 3, pp 310–328

Ownership property size, landscape structure, and spatial relationships in the Edwards Plateau of Texas (USA): landscape scale habitat management implications

  • Edith González Afanador
  • Michael E. Kjelland
  • X. Ben Wu
  • Neal Wilkins
  • William E. Grant
Article

DOI: 10.1007/s10669-016-9604-7

Cite this article as:
González Afanador, E., Kjelland, M.E., Wu, X.B. et al. Environ Syst Decis (2016) 36: 310. doi:10.1007/s10669-016-9604-7

Abstract

The present research focused on using spatial analysis to determine relationships among land ownership property sizes and landscape structure, with a focus on conservation management implications. Indices and metrics of ownership property sizes and landscape structure were calculated for 20 km buffer areas around 31 North American Breeding Bird Survey transects, 12 located within the Edwards Plateau ecoregion and 18 in contiguous ecoregions. The number of bird species observed at each transect provided a measure of avian species richness associated with land cover classes for each respective transect (González in Urban influence on diversity of avifauna in the Edwards Plateau of Texas: effect of property sizes on rural landscape structure, Texas A&M University, 2005). Spatial correlations were calculated between each pair of the landscape indices. Spatial analysis identified a “threshold of habitat fragmentation” for the 500 acre (ac) ownership property size. Significant spatial correlations among variables showed that property sizes lower than 500 ac produced habitat fragmentation represented by a decrease in mean patch size (MN) and proximity among habitat patches (Index PROX). Spatial analysis also made possible the prioritization of ecological sub-regions of the Edwards Plateau for conservation or restoration. The Live Oak-Mesquite Savannah showed the highest average ownership property size (7305 ac) and the highest values of patch richness. Based on the results, management in the Live Oak-Mesquite Savannah sub-region should focus on the conservation of land mosaic diversity to assure native avian species turnover (Whittaker 1972). In Balcones Canyon Lands, 64 % of land was covered by farms smaller than 500 ac and the overall average ownership property size was above the threshold of fragmentation (1440 ac), implying that management policies there should focus both on habitat conservation and on restoration. In contrast, 71 % of land in the Lampasas Cut Plains was covered by farms smaller than 500 ac, and average ownership property size was very close to the fragmentation threshold (625 ac). Consequently, the results indicate that management in the Lampasas Cut Plains sub-region should focus on habitat restoration (e.g., corridors that connect isolated habitat patches). In general, the threshold of ownership property size, 500 ac, is important for conservation planning because below that threshold of property size, habitat patch size begins to decrease and the distance between equivalent patches of habitat increases. Isolated patches act as islands within a sea of less suitable habitat which produce negative effects on biodiversity. Identifying the spatial characteristics indicative of habitat fragmentation, or the likelihood thereof, is an important issue for conservation planning in places with urban sprawl influence.

Keywords

Landscape ecology Landscape structure Land fragmentation Habitat fragmentation 

1 Introduction

Human-induced landscape transformations have important implications for the maintenance of biodiversity. Although private land conservation efforts have grown rapidly in recent years, the total area of developed land in the USA is still 10 times that of privately conserved lands, and land is being converted to residential and urban development at twice the rate that it is being protected (Smith et al. 2012). Urbanization is a major driver of biodiversity loss, with repercussions for the provision of ecosystem services, agricultural production, and human health (Cardinale et al. 2012; Foley et al. 2005; Grimm et al. 2008; McKinney 2002; Vitousek et al. 1997; Feinberg et al. 2015). As a landscape urbanizes, a biotic homogenization occurs, i.e., the first species to disappear are native ones which are gradually replaced by common ones (Reed et al. 2012). Further, Smith et al. (2012) noted that exurban development is increasing more rapidly in areas of high conservation value (Suarez-Rubio et al. 2011), such as within forests (Radeloff et al. 2005a), adjacent to protected lands (Rasker and Hansen 2000), and along lakeshores (Radeloff et al. 2001). Exurban development, characterized as low-density development on large lots (5–40 acres), has disproportionate effects on wildlife due to the amount of land consumed and fragmentation of land ownership and management (Smith et al. 2012).

Ecological processes are related not only to land use, but also to landscape structure, that is, to the spatial arrangement of land elements (Zonneveld and Forman 1989; Baudry 1993). Among the more pervasive landscape changes whose impact on biodiversity is potentially great, yet remains poorly understood, is that associated with changes in ownership property sizes along urban–rural gradients (Brooks et al. 2002; Fischer and Lindenmayer 2007). Urban sprawl generates economic pressure extending well past city limits into the rural landscape, which leads to a reduction in ownership property sizes (Costanza et al. 1997; Adger and Luttrel 2000; Antrop 2000; Swenson and Franklin 2000; Luck and Wu 2002; Radeloff et al. 2005b). Reduction in property sizes leads to changes in landscape structure (Stanfield et al. 2002; Wilkins et al. 2003; Kjelland et al. 2007), which can lead to changes in biodiversity (Donovan and Flather 2002; Lovett-Doust et al. 2003; Blair 2004; Gavier-Pizarro et al. 2010; Shochat et al. 2010; Kowarik 2011; Fontana et al. 2011; Hamilton et al. 2013; Martinuzzi et al. 2013). As Trousdale and Gregory (2004) point out, one of the most important and challenging aspects of biodiversity conservation is identifying priority lands for protection from development or other incompatible uses. Between 1982 and 2007, about 11 million acres of agricultural land in the USA were lost to development, essentially an irreversible loss (NRIS 2013). Between 1987 and 1992, Texas lost 234,300 acres of prime farmland with another 332,800 acres lost between 1992 and 1997, an increase of 42 % between the two, five-year periods (Dodds-Weir and Dykstra 2003). Further, between 1992 and 1997, just over 504,095 acres of palustrine and estuarine wetlands were lost in the USA; 75 % of these losses were attributed to either development (49 %) or agriculture (26 %) (US Department of Agriculture 2000; Gutzwiller and Flather 2011).

In Texas, the fragmentation of large, family-owned farms and ranches has been identified as the greatest threat to wildlife habitat within the state (Shackelford and Shackelford 2003; Wilkins et al. 2003; González 2005; López 2014). Many rural areas in the 1990s experienced a large increase in residential development; the southern and western portions of the state (the Trans Pecos, Edwards Plateau, South Texas Brush Country, and Coastal Sand Plains ecoregions) have been losing more than 235,000 acres (ac) annually that were in large ownerships (>2000 ac), thus dramatically shifting the size-class distribution of farms within these regions (Wilkins et al. 2003). In a case study within a single county (Bastrop), Wilkins et al. (2003) found that landscape characteristics such as number of patches per unit area and average patch size of native rangeland were influenced significantly by the subdivision of farms and ranches, that is, by land ownership fragmentation. More recently, López (2014) stated, “This dramatic loss and fragmentation of privately owned farms, ranches and forests—also known as working lands—is affecting the State’s rural economies. The conservation of water and other natural resources is also being affected, as is the nation’s national security and food security.”

As a contribution to understanding the problem of land fragmentation, the objective of the present study was to investigate specific linkages between ownership property size and landscape structure within the Edwards Plateau of Texas. The conceptual framework of landscape ecology (Forman 1984; Zonneveld and Forman 1989; Turner 1989; McGarigal and Marks 1995; Turner et al. 2001; Gergel and Turner 2002; Naveh 2007; Musacchio 2011; Wu 2013) and the quantitative tools of spatial statistics (Ludwig and Reynolds 1988; Isaaks and Srivastava 1989; Legendre and Fortin 1989; Fortin and Gurevitch 1993; Dutilleul et al. 1993; Dale and Fortin 2002; Fortin and Payette 2002) were both utilized to achieve the aforementioned objective. Specifically, tests were conducted to determine whether statistically significant spatial correlations exist among six indices of landscape structure and four categories of ownership property sizes. An evaluation of the potential implications of the spatial analysis results was made with regard to conservation planning. Conservation development techniques and policies can be used to protect biodiversity in the face of residential land use change (Smith et al. 2012). Conservation development refers to a set of land development techniques aimed at minimizing impacts on natural resources (Feinberg et al. 2015).

2 Study area

The study area was bounded by a circle with a radius of 300 km located at the geographical center point of the Edwards Plateau Ecoregion, which includes all of the Edwards Plateau and portions of five other ecoregions: South Texas Brush, Blackland Prairie, Llano Uplift, Rolling Plains, and Oak Woods. For purposes of the present research, eight Texas ecological sub-regions (Gould (1975) adapted by Wu et al. (2002)) were used and included: Balcones Canyon Lands, Black Land Prairies, Brush Country, Lampasas Cut Plains, Live Oak-Mesquite Savannah, Mesquite Savannah, Mesquite Plains, and Western Cross Timbers (Fig. 1).
Fig. 1

Geographic location of the study area. The study area included the eight ecological sub-regions referred to in this study area. Locations of the 31 North American Breeding Bird Survey (BBS) transects used by González (2005) to calculate avian diversity in the year 1992. Numbers correspond to the BBS code, of which the first two digits (83) signify Texas and the final three digits (015-238) correspond to the transect number

The climate in Texas ranges from subtropical steppe to subtropical sub-humid, with mean annual precipitation ranging from 375 mm in the west to 750 mm in the east, about three-fourths of which falls during the growing season, April through mid-November (Wilkins et al. 2003). The area is predominantly shrub land grazed by cattle, sheep, and goats, but local tracts are cultivated for domestic pasture and hay. Cotten and grain sorghum are also grown locally on irrigated land, and there are some pecan orchards on flood plains.

Many rural areas are experiencing greatly increased residential development, especially in the eastern portion of the region, due in large part to the influence of large cities such as San Antonio and Austin (Wilkins et al. 2003). In the Blackland Prairie ecological region of Texas, of the approximately 10.5 million acres that existed historically (Collins et al. 1975), less than 1 % of the original vegetation remains and consists of scattered parcels (Smeins and Diamond 1983). The potentially negative impact of the fragmentation of Texas rural lands on many species of flora and fauna is of particular concern because of the State’s high biodiversity and because there are growing urban centers in all 11 of the State’s ecological regions. In terms of biodiversity, Texas ranks second in the nation with 6273 species of plants and animals, with the Edwards Plateau containing some of the rarest species in the nation (Stein 2002).

Identifying and managing changing farm size distributions is important because general trends in land use in Texas are associated with changes in ownership size and include: (1) area in large farms and ranches (>2000 acres) is more likely to remain as native rangeland; (2) area in medium-sized farms and ranches (500 to 2000 acres) is more likely to remain as cropland; and (3) area in small farms and ranches (<500 acres) is more likely to be converted to non-native improved pastures (Wilkins et al. 2003). Improved pasture offers little in the way of habitat or food for native birds (Shackelford and Shackelford 2003). The trend of native rangeland being converted to non-native improved pastures eventually destroys the habitat value for wildlife species that require relatively large patches of native rangeland habitat (Engle 2002).

With regard to the shift from non-developed to developed land, it should be mentioned that the ramifications for wildlife are not equal among different species. For example, Glennon and Kretser (2013) found that occupancy rates of human-adapted and human-sensitive bird species were different (36 % higher and 26 % lower, respectively) at points near homes versus those in surrounding forest, and with similar effects in structurally different eastern and western North American landscapes (Glennon and Kretser 2013).

3 Methods

In this section, first a description is given of the databases and the calculation of the indices used to represent ownership property size and landscape structure. Next, the method used to test for spatial autocorrelation for each variable (an index) is described. A description of the tests for spatial correlations between each pair of variables is also presented. Finally, the test for correlation between the values of each pair of variables, corrected for the effects of spatial correlation, is described.

3.1 Ownership property sizes

We obtained data on ownership property sizes in 1992 for each of the 27 Texas counties that contained one or more of the 31 North American Breeding Bird Survey (BBS) transects (Sauer et al. 2003), used by González (2005), to calculate the avian diversity for the year 1992 as a historical reference baseline for calculating the biotic novelty in the future (Radeloff et al. 2015), within the study area (Fig. 1). These data included average size of rural (farm and ranch) property in acres (USDA 1992) and the proportions of rural acreage in each of four size classes: <50 (S1), 50–99 (S2), 100–500 (S3), >500 (S4) acres [unpublished data, Land Information Systems Laboratory, Texas A&M University; data summary available in Wilkins et al. (2003)]. The mean and standard deviation of the proportions of acreage in each of the four size classes were also calculated for each of the five ecological sub-regions.

Importantly, the time period of this study was selected based on the important findings of another study, Wilkins et al. (2003), which found that the most notable land use trend in Texas from 1992 to 2001 was the conversion of native rangelands and croplands to non-native “improved pastures.” Conversion from native rangeland to non-native vegetation, e.g., Bermuda grass, represents a significant loss of important wildlife habitat, especially in the central and eastern portions of Texas. The aforementioned trend is likely to continue according to a study by Martinuzzi et al. (2013) which predicted habitat loss within areas of biodiversity significance in the USA for the period 2001–2051, by ecoregions, and under different scenarios of future land use change. Given the recent land fragmentation trends in Texas, some areas of the State’s landscape are likely to be even less suitable as wildlife habitat for some species in the not too distant future.

3.2 Landscape structure indices

Land cover data were obtained for the Edwards Plateau Ecoregion for 1992 from the National Land Cover Dataset (NLCD 1992) of the United States Geological Service (USGS). The spatial resolution of the data was 30 m, mapped in the Albers Conic Equal Area projection, NAD 83. First, the 21-class land cover classification scheme of the NLCD was regrouped into 10 land cover classes, using the Spatial Analyst tool of Arc 8 GIS (ESRI 2000). The 10 land cover classes correspond to ecological systems found in the study area (NatureServe 2009). We then used the more aggregated land cover classes (woodland, shrubland, grassland, wetland, and urban), that provide habitat for the avian species observed on the BBS transects included in this study, identified by González (2005).

Next, using Arc 8 GIS (ESRI 2000), buffer scenes of 5, 10, and 20 km were created around the 31 BBS transects (buffer shapefiles). Subsequently, an identical area was “cut” on the NLCD. Within each of the 31 buffer scenes, six indices of landscape structure, two at the landscape mosaic level (patch richness, PR; and Shannon Diversity index, SHDI), and four at the land cover class level (percent of land PL; patch density PD (number of patches/100 acres); mean patch size MN (acres); proximity index PROX were calculated. At the landscape mosaic level, patch richness (PR), which is the number of patches (units of relatively homogeneous areas that differ from its surroundings) of the land cover classes present in the landscape of each buffer scene, was calculated while excluding the landscape border if present. The SHDI was also utilized for measuring diversity:
$${\text{SHDI}} = - \sum\limits_{i = 1}^{m} {P_{i} \times \ln P_{i} }$$
(1)
where Pi is the proportion of the landscape of each buffer scene occupied by land cover class i.
At the land cover class level, and for each of the land cover classes, percent of land (PL), patch density (PD) (number of patches/100 ac), mean patch size MN (ac), and the proximity index (PROX) were calculated using FRAGSTATS 2.0 (McGarigal and Marks 1995; Stanfield et al. 2002).
$${\text{PROX}} = \sum\limits_{s = 1}^{n} {\frac{{a_{ijs} }}{{h_{ijs}^{2} }}}$$
(2)
where PROX represents the proximity index for focal patch i, aijs is the area (m2) of patch ijs within a specified neighborhood (m) of patch ij, and hijs is the distance (m) between patch ijs and patch ijs, based on patch edge-to-edge distance, computed from cell center to cell center. A 100-m search radius was selected under the assumption that this was within the daily range of movements of all of the avian species included in the study area. Low index values indicate patches that are relatively isolated from other patches within the specified buffer distance, and high values indicate patches that are relatively connected to other patches (Turner et al. 2001).

PROX was used as an indicator of wildlife habitat fragmentation, high values indicating less fragmentation and low values indicating more fragmentation (Mortberg 2001; Brooks et al. 2002). The mean and standard deviation of PR, SHDI, PL, PD, MN, and PROX of the buffer scenes in each of the five ecological sub-regions were also calculated.

3.3 Spatial autocorrelation of variables

In order to be sure of the independence of the data and identify whether the variables were correlated with themselves, spatial autocorrelation (Felizola et al. 2003) was tested for each of the ownership property sizes and for each of the indices of landscape structure using a Mantel Test (r) (Fortin and Gurevitch 1993), i.e.,
$$r = \frac{{\sum {\sum {{\text{std}}A_{ij} \times {\text{std}}B_{ij} } } }}{n - 1}({\text{sum from }}i{\text{ to }}n{\text{ and sum from }}j{\text{ to }}n ,\quad {\text{for}}\;i \ne j)$$
(3)
where n is the number of sample locations, i and j identify the matrix element, Bij is the Euclidian distance matrix of location points, and Aij is the dissimilarity matrix of the variable of interest (S1, S2, S3, S4, PR, SHDI, PL, PD, MN, or PROX).

3.4 Spatial correlations between ownership property sizes and landscape structure indices

The spatial correlation between variables was conducted to try and identify patterns across the landscape. One essentially tests whether the relationship between ownership property size and landscape structure characteristic is correlated, and if so, positively or negatively. By identifying landscape patterns for the variables of interest, one can elucidate the mechanism involved and attempt to manage or mitigate for an undesired change across space and time.

To determine whether ownership property sizes and the indices of landscape structure are spatially correlated, a Cross Mantel Test (r) (Fortin and Gurevitch 1993) was conducted between each pair of ownership property sizes and landscape structure indices, i.e.,
$$r = \frac{{\sum {\sum {{\text{std}}A_{ij} \times {\text{std}}C_{ij} } } }}{n - 1}({\text{sum from }}i{\text{ to }}n{\text{ and sum from }}j{\text{ to }}n ,\quad {\text{for}}\;i \ne j)$$
(4)
where n is the number of sample locations, i and j identify the matrix element, Aij is the dissimilarity matrix of one of the variables of interest, that is an index of landscape structure (PR, SHDI, PL, PD, MN, PROX), and Cij is the dissimilarity matrix of the other variable of interest, that is, an ownership property size (S1, S2, S3, S4).

3.5 Correlation between values of ownership property sizes and landscape structure indices

A Pearson’s pair-wise correlation (ρ) between each pair of ownership property sizes and landscape structure indices was carried out, i.e.,
$$\rho = \frac{{\sum\nolimits_{i = 1} {(u_{i} - m_{u} )(v_{i} - m_{v} )} }}{{S_{u} S_{v} }}$$
(5)
where u and v are two variables (u is one of the six landscape structure indices and v is one of the four ownership property sizes), mu and mv are their respective means, and Su and Sv are their respective standard deviations.
The modified t test for autocorrelation (CHR) was also used, correcting the degrees of freedom based on the amount of autocorrelation in the data, to assess the correlation between each pair of spatially correlated variables (Clifford et al. 1989; Dutilleul et al. 1993), i.e.,
$$n^{{\prime }} (R) = \frac{{n^{2} }}{{\sum {\sum {{\text{cor}}(u_{i} u_{j} )} } }}$$
(6)
where R is the autocorrelation matrix, n is the number of observations, and ui and uj are the observations of the two variables. The corrected degrees of freedom (n′ − 2) were then used to test the significance of the correlation. The PASSAGE program (Rosenberg and Anderson 2011) was utilized to perform the calculations of Mantel Test, spatial correlations, and modified t test.

4 Results

4.1 Ownership property sizes

Ownership property sizes increased along an east–west gradient in the study area (Fig. 2). With respect to ecological sub-region, the largest average ownership property sizes occurred in Live Oak-Mesquite Savannah, followed by Mesquite Plains and Balcones Canyon Lands, Mesquite Savannah, Brush Country, Lampasas Cut Plains, and Western Cross Timber (Table 1). At county level, mean ownership property sizes ranged from 276 ac (Caldwell County) to 12,746 ac (Crocket County) (Table 2).
Fig. 2

Average ownership property size in the ecological sub-regions of Texas

Table 1

Mean, maximum, and minimum ownership property sizes (OPS) in acres, the range of sizes and the proportions of rural acreage in each of four size classes: <50 (S1), 50–99 (S2), 100–500 (S3), >500 (S4) in acres

Ecological sub-region

Data

Mean OPS

S1

S2

S3

S4

Balcones Canyon Lands

Mean

1441

20

10

34

36

 

Max

4232

28

17

41

69

 

Min

298

8

2

21

17

Blackland Prairie

Mean

387

20

16

45

20

 

Max

609

28

19

52

28

 

Min

276

12

9

38

14

Brush Country

Mean

965

17

12

43

28

 

Max

3288

25

18

55

55

 

Min

281

10

6

29

11

Lampasas Cut Plains

Mean

625

17

10

46

29

 

Max

636

22

11

52

31

 

Min

609

12

9

42

25

Live Oak-Mesquite Savannah

Mean

7305

7

4

14

75

 

Max

12,746

13

6

19

90

 

Min

2964

3

2

4

66

Mesquite Plains

Mean

1530

13

9

39

40

 

Max

2391

22

13

46

56

 

Min

706

8

6

30

26

Mesquite Savannah

Mean

1089

9

4

37

50

 

Max

1278

9

4

41

52

 

Min

995

9

4

35

46

Western Cross Timbers

Mean

349

25

17

42

16

 

Max

355

29

19

47

18

 

Min

343

21

14

37

14

Calculations were made for counties in each of the eight ecological sub-regions included in this study

Table 2

Mean of rural (farm and ranch) ownership property size (USDA 1992), and the proportions of rural acreage in each of four size classes: <50 (S1), 50–99 (S2), 100–500 (S3), >500 (S4) in acres

BBS transect code

Ecoregion

Ecological sub-region

County

Mean OPS

S1

S2

S3

S4

83027

Blackland Prairie

Lampasas Cut Plains

Caldwell

276

22

17

47

14

83029

South Texas Brush

Brush Country

Wilson

281

25

18

45

11

83139

Oak Woods

Lampasas Cut Plains

Williamson

298

28

17

38

17

83238

Edwards Plateau

Balcones Canyon Lands

Williamson

298

28

17

38

17

83062

Oak Woods

Mesquite Plains

Hood

343

29

19

37

14

83064

Oak Woods

Mesquite Plains

Erath

355

21

14

47

18

83015

South Texas Brush

Brush Country

Karnes

364

12

13

55

19

83048

Blackland Prairie

Lampasas Cut Plains

Falls

366

19

19

43

19

83030

South Texas Brush

Brush Country

Medina

451

23

13

40

25

83028

Edwards Plateau

Balcones Canyon Lands

Kendall

528

22

10

41

27

83050

Edwards Plateau

Lampasas Cut Plains

Coryell

609

12

9

52

28

83051

Edwards Plateau

Lampasas Cut Plains

Coryell

609

12

9

52

28

83017

South Texas Brush

Brush Country

Atascosa

618

20

14

41

25

83052

Edwards Plateau

Lampasas Cut Plains

Lampasas

628

16

9

44

31

83053

Edwards Plateau

Lampasas Cut Plains

Lampasas

628

16

9

44

31

83066

Edwards Plateau

Mesquite Plains

Callahan

636

22

11

42

25

83140

Edwards Plateau

Balcones Canyon Lands

Hays

658

27

15

39

19

83065

Rolling Plains

Mesquite Plains

Palo Pinto

706

22

13

40

26

83016

South Texas Brush

Brush Country

Live Oak

790

13

8

46

34

83076

Rolling Plains

Mesquite Plains

Young

834

12

9

46

33

83042

Llano Uplift

Live Oak-Mesquite Savannah

Mason

995

9

4

35

52

83043

Llano Uplift

Live Oak-Mesquite Savannah

Mason

995

9

4

35

52

83054

Llano Uplift

Live Oak-Mesquite Savannah

Mcculloch

1278

9

4

41

46

83031

Edwards Plateau

Balcones Canyon Lands

Uvalde

1487

13

8

31

48

83092

Rolling Plains

Live Oak-Mesquite Savannah

Garza

2190

8

6

30

56

83067

Rolling Plains

Lampasas Cut Plains

Shackelford

2391

9

7

40

44

83113

Edwards Plateau

Live Oak-Mesquite Savannah

Schleicher

2964

5

6

19

70

83018

South Texas Brush

Live Oak-Mesquite Savannah

Dimmitt

3288

10

6

29

55

83114

Edwards Plateau

Balcones Canyon Lands

Edwards

4232

8

2

21

69

83086

Edwards Plateau

Live Oak-Mesquite Savannah

Upton

6205

13

2

19

66

83112

Edwards Plateau

Live Oak-Mesquite Savannah

Crocket

12,746

3

3

4

90

Calculations were made in the 27 Texas counties associated with the 31 North American Breeding Bird Survey (BBS) transects included in the study area. The ecoregion and ecological sub-region (Fig. 1) to which each transect belongs are presented

Although a relatively high percentage of the rural acreage was in the largest (S4, >500 ac; from 16 to 75 %) and next-to-largest (S3, 100–500 ac; from 14 to 45 %) size classes in all eight ecological sub-regions, seven of the eight sub-regions also had roughly one-fifth of the rural acreage in the smallest (S1, <50 ac; from 7 to 25 %) size class, the exception being Live Oak-Mesquite Savannah (only 7 % in S1) (Table 1; Fig. 3).
Fig. 3

Mean percent of rural acreage in each of four size classes [<50 (S1), 50–99 (S2), 100–500 (S3), >500 (S4) acres]. Data are given for counties within each of the eight ecological sub-regions included in this study

4.2 Landscape structure indices

Of the 10 land cover classes that resulted from aggregation of the 21 NLCD classes, five (woodland, scrubland, grassland, wetland, and urban) provide habitat for the avian species included in this study area (González 2005) (Table 3; Fig. 4).
Table 3

Five land cover classes used in this study and their relation to the 21 land cover classes identified in the National Land Cover Data (NLCD)

NLCD code

NLCD land cover class

Reclassification code

Land cover reclassification

Classes providing habitat for birds in this study

11

Water

1

Water

 

12

Water

1

Water

 

21

Developed

2

Urban

Urban

22

Developed

2

Urban

 

23

Developed

2

Urban

 

31

Barren

3

Barren

 

32

Barren

3

Barren

 

33

Barren

3

Barren

 

41

Forest upland

4

Woodland

Woodland

42

Forest upland

4

Woodland

 

43

Forest upland

4

Woodland

 

51

Shrubland

5

Shrubland

Shrubland

61

Non-Natural Woody

6

Non-Natural Woody

 

71

Herbaceous Upland Natural/Semi-Natural Vegetation

7

Grassland

Grassland

81

Herbaceous planted/Cultivated

8

Pasture

 

82

Herbaceous planted/Cultivated

9

Herbaceous planted/Cultivated

 

83

Herbaceous planted/Cultivated

9

Herbaceous planted/Cultivated

 

84

Herbaceous planted/Cultivated

9

Herbaceous planted/Cultivated

 

85

Herbaceous planted/Cultivated

2

Urban

Urban

91

Wetland

10

Wetland

Wetland

92

Wetland

10

Wetland

 

The 21 NLCD classes reclassified into 10 classes, as described in the text, and then those classes that provide habitat for the bird species included in the present study were identified. From González (2005)

Fig. 4

Habitat preferences for 92 avian species registered by the North American Breeding Bird Survey (BBS) in 1992 on the 31 transects within the study area (calculated from data in González 2005)

At the landscape mosaic level, patch richness (PR) ranged from eight to ten and Shannon Diversity index (SHDI) ranged from 0.5 to 1.7 within the 31 buffer scenes (Table 4). In the eight ecological sub-regions, at the landscape mosaic level, Western Cross Timber, Black Land Prairie, and Lampasas Cut Plains had the highest mean PR values, whereas Black Land Prairie had the highest mean value of SHDI (Table 5). At the land cover class level, Balcones Canyon and Live Oak-Mesquite Savannah had almost 95 % of area covered by native habitats (woodland, scrubland, and grassland). Balcones Canyon Lands had the highest value in woodland cover, and Live Oak–Mesquite Savannah the highest in scrubland. Mesquite Savannah, Lampasas Cut Plains, and Western Cross Timber had almost 85 % of area covered in native habitats, especially in scrubland and grassland. Black Land Prairie, Brush Country, and Mesquite Plains had lower values in area covered by natural habitats, around 65 % (Fig. 5). Balcones Canyon Lands were characterized by woodlands (PL = 45), with low PD (2.8/100 ac), highest value of MN (19 ac), and high values of PROX (503.23), indicating relatively little fragmentation of woodlands (Tables 5, 6; Figs. 5, 6). Live Oak-Mesquite Savannah and Brush Country both were characterized by scrubland, but the former has the best habitat quality. Live Oak-Mesquite Savannah was more than half covered by scrubland (PL = 67.7), with big patches (MN = 120 ac) which are in close proximity to each other, as indicated by PROX (301.948) (Tables 5, 6; Figs. 5, 6). Mesquite Plains and Lampasas Cut Plains were characterized by grassland (PL = 45 and 36, respectively), but the Mesquite Plains had less fragmented habitat, MN (19) and PROX (68,458) compared to Lampasas Cut Plains, MN (11 ac) and PROX (3155) (Tables 5, 6; Figs. 5, 6).
Table 4

Landscape structure indices at the landscape mosaic level (patch richness, PR; and Shannon Diversity index, SHDI) and the land cover class level [percent of land, PL; patch density, PD (number of patches/100 acres); mean patch size, MN (acres); proximity index, PROX]

BBS

Ecoregion

Ecological sub-region

Landscape mosaic indices

Transect code

PR

SHDI

83015

South Texas Brush

Brush Country

9

1.7

83016

South Texas Brush

Brush Country

9

1.4

83017

South Texas Brush

Brush Country

9

1.5

83018

South Texas Brush

Live Oak-Mesquite Savannah

10

1.2

83027

Blackland Prairie

Lampasas Cut Plains

9

1.7

83028

Edwards Plateau

Balcones Canyon Lands

9

1.2

83029

South Texas Brush

Brush Country

9

1.7

83030

South Texas Brush

Brush Country

10

1.5

83031

Edwards Plateau

Balcones Canyon Lands

10

1.0

83042

Llano Uplift

Live Oak-Mesquite Savannah

9

1.2

83043

Llano Uplift

Live Oak-Mesquite Savannah

9

1.2

83048

Blackland Prairie

Lampasas Cut Plains

10

1.7

83050

Edwards Plateau

Lampasas Cut Plains

9

1.7

83051

Edwards Plateau

Lampasas Cut Plains

9

1.4

83052

Edwards Plateau

Lampasas Cut Plains

9

1.3

83053

Edwards Plateau

Lampasas Cut Plains

10

1.3

83054

Llano Uplift

Live Oak-Mesquite Savannah

9

1.3

83062

Oak Woods

Mesquite Plains

10

1.5

83064

Oak Woods

Mesquite Plains

10

1.4

83065

Rolling Plains

Mesquite Plains

9

1.6

83066

Edwards Plateau

Mesquite Plains

10

1.4

83067

Rolling Plains

Lampasas Cut Plains

9

1.1

83076

Rolling Plains

Mesquite Plains

9

1.5

83086

Edwards Plateau

Live Oak-Mesquite Savannah

10

0.9

83092

Rolling Plains

Live Oak-Mesquite Savannah

9

1.2

83112

Edwards Plateau

Live Oak-Mesquite Savannah

8

0.7

83113

Edwards Plateau

Live Oak-Mesquite Savannah

10

0.5

83114

Edwards Plateau

Balcones Canyon Lands

9

1.0

83139

Oak Woods

Lampasas Cut Plains

10

1.6

83140

Edwards Plateau

Balcones Canyon Lands

9

1.1

83238

Edwards Plateau

Balcones Canyon Lands

9

1.4

BBS

Land cover class indices

Transect code

Woodland

Scrubland

Grassland

Wetland

Urban

PL

PD

MN

PROX

PL

PD

MN

PROX

PL

PD

MN

PROX

PL

PD

MN

PROX

PL

PD

MN

PROX

83015

17

4.5

4.9

166

24

6.9

2.5

222

16

6.1

2.5

68

1.1

0.9

1.2

3

1

0.1

7.4

37

83016

13

4.5

2.5

88

55

2.0

27.2

45,153

8

2.4

2.5

22

0.6

0.3

1.7

13

1

0.2

2.5

9

83017

16

5.3

2.5

74

48

3.2

14.8

52,866

8

2.4

2.5

18

0.9

0.6

1.5

7

0

0.0

4.9

6

83018

3

1.6

2.5

88

60

2.4

27.2

100,000

24

4.5

4.9

690

0.2

0.1

2.2

9

0

0.0

2.5

2

83027

27

3.2

7.4

3412

12

7.3

2.5

9

23

4.0

4.9

103

0.0

0.1

0.2

0

2

0.2

12.4

118

83028

46

3.6

12.4

50,323

7

10.5

0.0

30

40

3.2

12.4

4372

0.0

0.1

0.2

0

2

0.2

12.4

43

83029

16

4.0

4.9

405

22

6.1

2.5

70

14

4.9

2.5

27

0.2

0.4

0.7

1

5

0.2

27.2

2160

83030

14

4.0

2.5

597

39

2.8

12.4

7799

6

2.4

2.5

5

0.1

0.1

0.5

0

1

0.1

4.9

24

83031

69

1.2

49.4

100,000

13

4.0

2.5

81

8

3.2

2.5

14

0.0

0.0

0.2

0

0

0.1

2.5

4

83042

37

4.9

7.4

3160

43

3.6

12.4

31,159

16

4.0

4.9

80

0.0

0.0

0.2

0

0

0.1

4.9

7

83043

15

5.7

2.5

78

53

2.4

22.2

84,367

25

3.6

7.4

847

0.0

0.0

0.5

0

1

0.1

7.4

15

83048

24

3.2

7.4

621

16

6.1

2.5

40

12

5.7

2.5

21

0.2

0.3

0.7

1

0

0.2

2.5

7

83050

29

3.6

7.4

1039

13

6.5

2.5

17

26

3.2

9.9

886

0.1

0.2

0.2

0

3

0.2

14.8

89

83051

23

4.0

4.9

435

19

6.5

2.5

453

46

2.4

19.8

8356

0.0

0.1

0.2

0

4

0.1

37.1

511

83052

19

4.0

4.9

303

37

5.3

7.4

3957

35

4.0

9.9

1892

0.0

0.1

0.2

0

2

0.1

19.8

165

83053

13

4.9

2.5

62

46

3.6

12.4

44,837

30

4.5

7.4

269

0.0

0.0

0.2

0

0

0.1

4.9

12

83054

6

4.0

2.5

94

43

2.0

19.8

45,072

20

4.0

4.9

174

0.0

0.1

0.2

0

1

0.1

4.9

26

83062

25

3.2

7.4

219

12

5.7

2.5

588

44

3.2

14.8

20,418

0.1

0.2

0.2

0

2

0.4

4.9

16

83064

18

4.5

4.9

100

43

3.6

12.4

14,981

26

5.3

4.9

606

0.0

0.1

0.2

0

1

0.3

4.9

41

83065

28

2.8

9.9

847

14

7.7

2.5

13

33

3.2

9.9

3911

0.1

0.0

3.0

35

1

0.1

4.9

56

83066

13

4.0

2.5

29

40

3.6

12.4

29,115

32

4.5

7.4

2104

0.0

0.0

0.5

0

1

0.4

2.5

7

83067

1

1.2

2.5

4

21

4.9

4.9

270

64

2.0

29.7

138,471

0.1

0.2

0.5

0

0

0.0

4.9

41

83076

4

2.4

2.5

11

14

5.7

2.5

52

43

3.2

14.8

28,348

0.1

0.2

0.5

0

0

0.0

22.2

95

83086

0

0.0

0.0

0

52

3.6

14.8

73,598

42

3.6

9.9

18,424

0.0

0.0

1.2

1

0

0.3

0.0

3

83092

0

0.0

0.0

0

14

6.9

2.5

207

40

1.6

22.2

103,102

0.1

0.1

1.0

1

0

0.1

2.5

20

83112

1

0.4

2.5

4

64

2.4

29.7

300,000

35

5.7

7.4

3426

0.0

0.0

0.2

0

0

0.2

0.0

2

83113

1

0.8

2.5

2

87

0.4

316.3

500,000

10

7.3

2.5

8

0.0

0.0

0.2

0

0

0.1

2.5

12

83114

15

2.8

4.9

165

64

1.2

49.4

200,000

20

4.0

4.9

394

0.0

0.0

0.2

0

0

0.0

7.4

20

83139

13

3.6

2.5

272

13

6.5

2.5

40

19

4.9

4.9

172

0.0

0.1

0.2

0

2

0.2

12.4

91

83140

56

2.8

19.8

40,323

8

10.1

0.0

22

31

4.0

7.4

360

0.0

0.1

0.2

0

3

0.2

17.3

72

83238

40

3.6

9.9

8122

14

8.5

2.5

78

36

3.6

9.9

5843

0.1

0.1

0.5

0

6

0.2

22.2

1265

The calculation were made for the buffer scenes around the 31 North American Breeding Bird Survey (BBS) transects from which data were drawn for the present study

Table 5

Mean (±1SD) landscape structure indices at the landscape mosaic level (patch richness, PR; and Shannon Diversity index, SHDI) and the land cover class level (percent of land, PL; patch density, PD (number of patches/100 acres); mean patch size, MN (acres); proximity index, PROX)

Ecological sub-regions

Balcones Canyon Lands

Blackland Prairie

Brush Country

Lampasas Cut Plains

Live Oak-Mesquite Savannah

Mesquite Plains

Mesquite Savannah

Western Cross Timbers

Land mosaic level

 PR

9.2 (0.4)

9.5 (0.6)

9.3 (0.5)

9.5 (0.6)

9.3 (1.2)

9 (0)

9 (0)

10 (0)

 SHDI

1.1 (0.2)

1.7 (0.1)

1.5 (0.2)

1.4 (0.1)

0.7 (0.2)

1.4 (0.2)

1.2 (0.1)

1.5 (0.1)

Class level

 Grassland

  MN

7 (1)

6 (3)

3 (1)

11 (6)

7 (4)

19 (9)

6 (1)

10 (7)

  PD

3.6 (0.4)

4.5 (1.1)

3.8 (1.6)

3.8 (1.0)

5.5 (1.8)

2.5 (0.8)

3.9 (0.2)

4.2 (1.4)

  PL

27 (13)

20 (6.1)

13 (6.8)

36 (7.1)

29 (16.8)

45 (13.3)

20 (4.5)

35 (12.7)

  PROX

2197 (2711.8)

296 (398.5)

138 (271.1)

3155 (3562.7)

7286 (9796)

68,458 (62,921)

367 (418.3)

10,512 (14,009.2)

 Scrubland

  MN

11 (21)

2 (0)

14 (11)

9 (5)

120 (170)

3 (1)

18 (5)

7 (7)

  PD

6.9 (4.1)

6.6 (0.5)

3.9 (2.0)

4.8 (1.4)

2.1 (1.6)

6.3 (1.3)

2.7 (0.8)

4.7 (1.4)

  PL

21.2 (24)

13.5 (1.7)

41.3 (16)

35.5 (11.6)

67.7 (17.8)

15.8 (3.5)

46.3 (5.8)

27.5 (21.9)

  PROX

47,629.6 (106,385)

26.5 (15.9)

37,505.3 (46,037)

19,590.5 (21,124.6)

301,947.7 (215,326)

135.5 (122.7)

53,532.7 (27,594.6)

7784.5 (10,177)

 Urban

  MN

12 (8)

11 (5)

8 (9)

16 (16)

1 (1)

9 (9)

6 (1)

5 (0)

  PD

0.2 (0.08)

0.2 (0.04)

0.1 ((0.08)

0.2 (0.12)

0.2 (0.08)

0.1 (0.04)

0.1 (0.04)

0.4 (0.04)

  PL

2.2 (2.5)

1.8 (1.3)

1.3 (1.9)

1.8 (1.7)

1 (0)

0.3 (0.5)

0.7 (0.6)

1.5 (0.7)

  PROX

280.8 (550.8)

76.3 (48)

373 (875.5)

173.8 (236.5)

5.7 (5.5)

53 (31.7)

16 (9.5)

28.5 (17.7)

 Wetland

  MN

0.2 (0.0)

0.5 (0.2)

1.2 (0.7)

0.2 (0.2)

0.5 (0.5)

1.2 (1.2)

0.2 (0.2)

0.2 (0.0)

  PD

0.08 (0.04)

0.16 (0.12)

0.40 (0.32)

0.08 (0.04)

0.00 (0.00)

0.16 (0.08)

0.04 (0.04)

0.16 (0.04)

  PL

0 (0)

0.1 (0.1)

0.5 (0.4)

0 (0)

0 (0)

0.1 (0)

0 (0)

0.1 (0.1)

  PROX

0 (0)

0.3 (0.5)

5.5 (5)

0 (0)

0.3 (0.6)

9 (17.3)

0 (0)

0 (0)

 Woodland

  MN

19.3 (17.8)

6.2 (2.5)

3.2 (1.2)

3.7 (1.5)

1.7 (1.5)

3.7 (4.2)

4.2 (3.0)

6.2 (1.7)

  PD

2.8 (1.0)

3.4 (0.2)

4.0 (1.3)

4.2 (0.4)

0.4 (0.0)

1.6 (1.3)

4.9 (0.8)

3.8 (0.8)

  PL

45.2 (20.1)

23.3 (7.1)

13.2 (5.2)

17 (4.9)

0.7 (0.6)

8.3 (13.3)

19.3 (15.9)

21.5 (4.9)

  PROX

50,323 (57,297)

1336 (1419)

236 (216)

207 (195)

2 (2)

216 (421)

1111 (1775)

160 (84)

The calculation was made for the buffer scenes in each of the five ecological sub-regions

Fig. 5

Mean percent of land in each of the five land cover classes in each of the eight ecological sub-regions included in the present study

Table 6

Proximity index of land cover classes (woodland, scrubland, grassland, wetland, and urban) in the ecological sub-regions

Ecological sub-regions

Grassland

Scrubland

Urban

Wetland

Woodland

Balcones Canyon Lands

2197

47,630

281

0.0

50,323

Blackland Prairie

296

27

76

0.3

1336

Brush Country

138

37,505

373

5.5

236

Lampasas Cut Plains

3155

19,591

174

0.0

207

Live Oak-Mesquite Savannah

7286

301,948

6

0.3

2

Mesquite Plains

68,458

136

53

9.0

216

Mesquite Savannah

367

53,533

16

0.0

1111

Western Cross Timbers

10,512

7785

29

0.0

160

Fig. 6

Mean patch size (MN) of land cover classes (woodland, scrubland, grassland, wetland, and urban) in the ecological sub-regions

Balcones Canyon Lands, Lampasas Cut Plains, Black Land Prairie, and Brush Country were the most urbanized sub-regions (PL = 2.2, 1.8, 1.8, and 1.3, respectively, and mean patch size MN = 12, 16, 11, and 8 ac, respectively). Proximity index (PROX) was highest in Brush Country (373), followed by Balcones Canyon Lands (281), Lampasas Cut Plains (174), and Black Land Prairie (76), indicating that urban zones were more compact in Brush Country, Balcones Canyon Lands, and Lampasas Cut Plains (Tables 5, 6; Figs. 5, 6).

4.3 Spatial autocorrelation of variables

Approximately half of the indices showed significant levels of positive spatial autocorrelation, including S2, S3, and S4 (P = 0.0035, 0.0013, and 0.0003, respectively), but not S1 (P = 0.0977) among ownership property size classes and SHDI (P = 0.0072), and not PR (P = 0.2406) among landscape structure indices at the landscape mosaic level (Table 7).
Table 7

Degree of spatial autocorrelation in each of four ownership property sizes [<50 (S1), 50–99 (S2), 100–500 (S3), >500 (S4) acres] and each of six indices of landscape structure (two at the landscape mosaic level [patch richness (PR) and Shannon Diversity index (SHDI)] and four at the land cover class level [percent of land (PL), patch density (PD), mean patch size (MN), proximity (PROX)], as indicated by Mantel’s r (Fortin and Gurevitch 1993)

 

Indices

Mantel’s r

P

Ownership property size

S1

0.0932

0.0977

S2

0.2140

0.0035

S3

0.3418

0.0013

S4

0.3809

0.0003

Landscape structure

 Landscape mosaic level

PR

0.0498

0.2406

SHDI

0.2643

0.0072

Land cover class level

 Woodland

PL

0.1009

0.1517

PD

0.4162

0.0001

MN

−0.0248

0.5400

PROX

−0.1049

0.8365

 Scrubland

PL

0.2216

0.0071

PD

0.1303

0.0521

MN

0.0570

0.2952

PROX

0.1726

0.0884

Grassland

PL

0.3811

0.0001

PD

0.1742

0.0416

MN

0.3734

0.0005

PROX

0.3140

0.0070

 Wetland

PL

0.2164

0.0295

PD

0.1861

0.0603

MN

0.2970

0.0063

PROX

0.1889

0.0605

 Urban

PL

−0.0770

0.7763

PD

0.0696

0.1807

MN

−0.0643

0.7308

PROX

−0.0859

0.7581

Landscape structure indices at the land cover class level (PL, PD, MN, and PROX) tended to be more significantly spatially autocorrelated (P < 0.1) in scrublands, grasslands, and wetlands than in woodlands and urban areas (P > 0.1), although MN was not significantly autocorrelated in scrublands (P = 0.2952) and PD was significantly autocorrelated in woodlands (P = 0.0001).

4.4 Spatial correlations between ownership property sizes and landscape structure indices

All four ownership property size classes (S1–S4) showed significant positive spatial correlation with both landscape structure indices at the landscape mosaic level (PR and SHDI) (P < 0.053), with one exception: S2 was not significantly spatially correlated with PR (P = 0.2226) (Table 8). At the land cover class level, the four ownership property sizes were more spatially correlated with the four landscape structure indices in scrublands (13 of the 16 pairs showed significant correlations; P < 0.05) than in other ecological sub-regions (4/16 in urban areas, 3/16 in woodlands, 2/16 in grasslands, and 0/16 in wetlands); all significant correlations were positive.
Table 8

Degree of spatial correlation between pairs of ownership property sizes [<50 (S1), 50–99 (S2), 100–500 (S3), >500 (S4) acres] and landscape structure [richness of patches (PR), Shannon Diversity index (SHDI), percent of land (PL), patch density (PD), mean patch size (MN), proximity (PROX)] indices, as indicated by Cross Mantel’s r (Fortin and Gurevitch 1993)

Landscape structure

Indices

Ownership Property Size (OPS)

  

S1

S2

S3

S4

Mean OPS

Cross Mantel’s r

P

Cross Mantel’s r

P

Cross Mantel’s r

P

Cross Mantel’s r

P

Cross Mantel’s r

P

Landscape mosaic level

SHDI

0.2780

0.0014

0.3733

0.0002

0.6573

0.0001

0.6773

0.0001

0.5083

0.0059

 

PR

0.1116

0.0523

0.0473

0.2226

0.2720

0.0062

0.2115

0.0163

0.3753

0.0043

Land cover class level

 Woodland

PL

0.1504

0.0372

0.0402

0.3017

0.0932

0.1761

0.1641

0.0681

0.1316

0.11939

 

PD

0.1032

0.0934

0.1163

0.0673

0.4467

0.0003

0.4598

0.0001

0.4628

0.0002

 

MN

−0.0217

0.4472

−0.0501

0.5944

−0.0313

0.4190

−0.0028

0.3445

−0.0393

0.40886

 

PROX

−0.0045

0.4010

−0.0847

0.7697

−0.0680

0.5591

−0.0565

0.5926

−0.0788

0.56014

 Scrubland

PL

0.2417

0.0017

0.2532

0.0018

0.4313

0.0005

0.4959

0.0002

0.3683

0.0244

 

PD

0.2709

0.0004

0.2377

0.0017

0.1810

0.0385

0.3286

0.0008

0.1549

0.07949

 

MN

0.1638

0.1053

0.0082

0.4125

0.3120

0.0279

0.3039

0.0309

0.1272

0.08589

 

PROX

0.2695

0.0051

0.1526

0.0690

0.6415

0.0003

0.6150

0.0004

0.5492

0.0305

 Grassland

PL

−0.0546

0.7653

−0.0739

0.8519

−0.1292

0.9420

−0.0829

0.8487

−0.0696

0.69273

 

PD

0.0845

0.1476

−0.0277

0.6315

0.2731

0.0309

0.2366

0.0325

0.226

0.05889

 

MN

−0.0285

0.5778

−0.0833

0.8239

−0.0921

0.7548

−0.0446

0.6045

−0.0388

0.46485

 

PROX

0.0015

0.4485

−0.0595

0.7077

−0.0707

0.6207

0.0030

0.3971

0.0199

0.24848

 Wetland

PL

−0.1208

0.9687

−0.0459

0.6252

0.0565

0.2312

−0.0829

0.7294

−0.0863

0.59854

 

PD

−0.0842

0.8347

0.0418

0.2868

0.1591

0.0937

0.0144

0.2902

−0.0311

0.35716

 

MN

−0.1083

0.9387

−0.0759

0.7818

−0.0445

0.5402

−0.0740

0.7174

−0.0147

0.34807

 

PROX

−0.1083

0.9425

−0.0950

0.8484

−0.1105

0.8027

−0.1154

0.9213

−0.078

0.59304

 Urban

PL

0.2799

0.0015

0.2057

0.0103

0.0087

0.3601

0.1296

0.1080

−0.0309

0.38526

 

PD

0.2029

0.0094

0.1573

0.0256

−0.0488

0.6452

0.0678

0.2354

0.0263

0.32517

 

MN

0.0362

0.2950

0.0199

0.3717

0.1019

0.1644

0.0358

0.2735

0.006

0.31827

 

PROX

0.1367

0.0950

0.1610

0.0571

−0.0815

0.6235

0.0337

0.2387

−0.0762

0.55214

Within scrublands, all four ownership property sizes were significantly correlated (P < 0.05) with PL and PD; all but S2 were significantly correlated with PROX; and S3 and S4 were significantly correlated with MN. Within urban areas both S2 and S3 were significantly correlated with PL and PD; within woodlands S1 was significantly correlated with PL, and S3 and S4 were significantly correlated with PD; and within grasslands both S3 and S4 were significantly correlated with PD.

4.5 Correlation between values of ownership property sizes and landscape structure indices

After appropriate adjustments for spatial autocorrelation, values of all four ownership property size classes showed significant correlation (P < 0.05) with values of SHDI at the landscape mosaic level. Correlations were positive for the three smallest ownership property size classes (S1–S3) and negative for the largest property size class (S4) and average ownership property size. Only the values of S3 showed a significant (positive) correlation with values of PR (Table 9).
Table 9

Degree of correlation between values of pairs of ownership property sizes [<50 (S1), 50–99 (S2), 100–500 (S3), >500 (S4) acres] and landscape structure [patch richness (PR), Shannon Diversity index (SHDI), percent of land (PL), patch density (PD), mean patch size (MN), proximity (PROX)] indices, as indicated by the modified t test for autocorrelation (CHR) (Clifford et al. 1989; Dutilleul et al. 1993)

Landscape structure

Indices

Ownership Property Size (OPS)

  

S1

S2

S3

S4

Mean OPS

CHR

P

CHR

P

CHR

P

CHR

P

GRH

P

Landscape mosaic level

SHDI

0.5682

0.0403

0.6987

0.0134

0.8149

0.0049

−0.8392

0.0053

−0.6721

0.0241

 

PR

0.2971

0.0685

0.2297

0.1525

0.0228

0.8798

−0.1850

0.2000

−0.287

0.0655

Land cover class level

 Woodland

PL

0.4637

0.0543

0.3635

0.1664

0.2745

0.3425

−0.4153

0.1621

−0.4171

0.1129

 

PD

0.3805

0.1602

0.3494

0.2556

0.6558

0.0299

−0.5897

0.0382

−0.6212

0.0244

 

MN

0.1937

0.3357

0.1397

0.4894

0.0179

0.9327

−0.1175

0.5813

−0.1634

0.4179

 

PROX

0.1774

0.3466

0.0837

0.6672

0.2180

0.1787

−0.0155

0.9385

−0.0672

0.7308

 Scrubland

PL

−0.5721

0.0227

−0.617

0.0206

−0.5953

0.0444

0.7009

0.0186

−0.337

0.2027

 

PD

0.5965

0.0145

0.5997

0.0214

0.4881

0.0992

−0.6464

0.0280

0.0007

0.9975

 

MN

−0.3755

0.0718

−0.2665

0.1996

−0.4366

0.0462

0.4479

0.0456

−0.3596

0.153

 

PROX

−0.5454

0.0280

−0.4802

0.0711

0.7553

0.0246

0.7283

0.0102

−0.1756

0.3564

 Grassland

PL

−0.0383

0.8998

−0.1263

0.7330

0.0324

0.9333

0.0278

0.9462

0.109

0.752

 

PD

−0.1023

0.6094

0.0462

0.7959

−0.3057

0.1161

0.1837

0.3516

0.3087

0.1045

 

MN

−0.1672

0.5472

−0.1857

0.5682

0.0262

0.9393

0.0976

0.7855

0.0345

0.9105

 

PROX

−0.2464

0.3158

−0.1816

0.5124

−0.0682

0.8173

0.1748

0.5654

0.099

0.7142

 Wetland

PL

−0.0039

0.9889

0.2235

0.4635

0.3487

0.2681

−0.2532

0.4479

−0.1801

0.5377

 

PD

0.1029

0.6914

0.3833

0.1696

0.5046

0.0705

−0.4226

0.1529

−0.2948

0.2668

 

MN

−0.0243

0.9171

−0.0129

0.9547

−0.0201

0.9392

0.0290

0.9150

0.0363

0.886

 

PROX

0.0872

0.6632

0.0666

0.7225

0.0663

0.7502

−0.0763

0.7160

−0.0679

0.7429

 Urban

PL

0.5671

0.0173

0.5379

0.0314

0.3655

0.1987

−0.5540

0.0476

−0.337

0.2027

 

PD

0.4994

0.0149

0.3812

0.0356

0.0125

0.9540

−0.2988

0.1610

0.0007

0.9975

 

MN

0.2892

0.2468

0.3132

0.1986

0.4785

0.0701

−0.4498

0.0963

−0.3596

0.153

 

PROX

0.3502

0.0507

0.3823

0.0382

0.1852

0.3334

−0.3385

0.0711

−0.1756

0.3564

At the land cover class level, the values of the four ownership property sizes were more correlated with the values of the four landscape structure indices in scrublands (14 of the 16 pairs showed significant correlations) than in other ecological sub-regions (7/16 in urban areas, 3/16 in woodlands, 0/16 in grasslands, and 0/16 in wetlands). Within scrublands, values of all four ownership property sizes were significantly correlated with values of PL and PROX; values of all but S4 were significantly correlated with values of PD; and values of all but S3 were significantly correlated with values of MN. All significant correlations with values of landscape indices were negative for values of S1; all were positive for values of S2; and all were mixed for values of S3 and S4.

Within urban areas, values of both of the two smallest ownership property sizes (S1 and S2) showed significant positive correlations with values of PL, PD, and PROX; the only other significant correlation (between values of S4 and values of PL) was negative. Within woodlands, values of S1 showed significant positive correlations with values of PL; values of S3 showed significant positive correlations with values of PD; and values of S4 showed significant negative correlations with values of PD.

5 Discussion

Landscape structure is spatially correlated with ownership property sizes in the study area. The spatial correlation was found at land mosaic and class levels. At the class level, results showed that there is a threshold of ownership property size of 500 ac, below which habitat fragmentation is evident.

At the land mosaic level, there was positive and significant correlation between Shannon Diversity index (SHDI) and ownership property sizes smaller than 500 ac, but negative correlation with ownership property sizes larger than 500 ac. The aforementioned result may be interpreted as meaning that in landscapes with large property sizes, landscape diversity is low; however, it increases when land is divided into smaller properties. In the ecological sub-regions, the Black Land Prairie had the highest SHDI value (1.7) followed by the Western Cross Timbers (1.5) and notably the average ownership property size (387 and 349 ac, respectively) was below the threshold of 500 ac (Table 1). Perhaps the smaller average ownership property sizes fall below the 500 ac threshold because these ecological sub-regions are very close to urban areas where, in addition to natural vegetation patches, there is ornamental, non-native vegetation which can contribute to the landscape diversity.

At the class level, spatial correlation detected with the Cross Mantel is clearly evident in scrubland and urban classes of land cover. In scrubland, all indices of land structure showed positive spatial correlation with all ownership property sizes. Correlation identified with Pearson correlation and modified t test is significantly negative between percent of land (PL) and mean patch size (MN) with ownership property sizes smaller than 500 ac. The proximity index (PROX) became positive with ownership property sizes larger than 100 ac. These results can be interpreted as meaning that patches of habitat begin to decrease when land is dominated by ownership property sizes smaller than 500 ac, but the number of scrubland patches with a distance of 100 m in between them decreases if land becomes dominated by ownership property sizes smaller than 100 ac. With regard to urban PL and patch density (PD), a positive spatial correlation was shown to exist with ownership property sizes smaller than 100 ac, which is just opposite of the correlations with structure indices of natural land cover classes (scrubland). At both land mosaic and class levels, the tendency of the correlation changed at 500 ac ownership property sizes; therefore, this seems to correspond to an ownership size threshold below which the landscape fragmentation increases. This result represents an increase in diversity of land mosaic and habitat fragmentation at class level, which is indicated by a decrease in mean patch size and greater proximity among patches.

Ownership property size of 500 ac is a threshold too, for consolidation of urban or suburban landscape. It is indicated by the positive and significant correlation among all the urban structure indices and percentage of land occupied by ownership properties smaller than 100 ac, and negative over 500 ac (Table 8). The inverse tendency of the correlations with respect to those with class structure indices of natural habitats such as scrubland can indicate that natural habitats have been fragmented by the increase in urbanization. In the study area, there are seven counties with an average ownership property size below of the threshold of 500 ac, with six counties being very close to it. Counties with ownership property sizes below the threshold are: Karnes (364 ac) and Wilson (281 ac) in the Brush Country sub-ecoregion; Caldwell (276 ac) and Falls (366 ac) in the Lampasas Cut Plains; and Hood (343 ac) in the Mesquite Plains (Table 2). All of the aforementioned counties have greater than 75 % of the area covered by ownership property sizes smaller than 500 ac (Fig. 5).

Identification of the 500 ac threshold is important for conservation planning because a decrease in habitat patch size and proximity among patches are factors that are directly related to fauna meta-population management. It is known that isolated patches act as habitat islands and colonization or repopulation depends upon the patch area and proximity among patches.

6 Conclusions

The present study can serve as a model of the spatial analysis technique with important implications for conservation planning, and as such provides a template that may be used with more recent datasets in future studies. By identifying which areas are most likely to be in jeopardy of land fragmentation in the future, one can maximize efficiency with regard to the allocation of resources to protect the areas with habitat and/or wildlife most at risk. Importantly, if connectivity for one process or organism is restored, then other processes and organisms will likely benefit (Noss et al. 2006; Baldwin et al. 2012).

Landscape structure was spatially correlated with ownership property sizes in the study area. The correlation existed both at the land mosaic level and at the class level. Significant spatial correlations detected by Cross Mantel and significant correlations detected by Pearson correlation and modified t test identified a threshold of ownership property size at 500 ac; below 500 ac land diversity increased at the land mosaic level, but habitat fragmentation increased at the class level. The latter effect is represented by a decrease in mean patch size and an increase in the distance between equivalent patches.

The fragmentation of rural properties continues and conservation planning should incorporate the linkages between factors such as ownership property sizes and specific elements of landscape structure. Studies such as Kjelland et al. (2007) identified NAV of land and farm size as important land fragmentation indicators, i.e., variables for projections of land fragmentation in Texas. With today’s technology and ever increasing understanding of the specific linkages between ownership property size and landscape structure, spatially explicit modeling can provide risk-based conservation planning projections of areas and species that are most likely to be in jeopardy.

With regard to future research, it would be interesting to examine the spatial correlation of property size and land use change on individual bird species. Some species will do better than others depending on property size and associated habitat conditions, i.e., larger undeveloped properties, e.g., Northern Bobwhite, versus smaller property sizes or in more developed areas, e.g., Common Grackle.

Acknowledgments

We thank the Texas Cooperative Extension, the Texas A&M University System, Dr. Terri Morgan from Partnership for Environment (PFE) for partial economic support, and Dr. Maria Chakerian for her unconditional assistance with FRAGSTATS analysis.

Copyright information

© Springer Science+Business Media New York 2016

Authors and Affiliations

  • Edith González Afanador
    • 1
  • Michael E. Kjelland
    • 2
    • 3
    • 4
  • X. Ben Wu
    • 5
  • Neal Wilkins
    • 2
  • William E. Grant
    • 2
  1. 1.Universidad Nacional de ColombiaBogotáColombia
  2. 2.Department of Wildlife and Fisheries SciencesTexas A&M UniversityCollege StationUSA
  3. 3.Conservation, Genetics & Biotech, LLCValley CityUSA
  4. 4.U.S. Army Engineer Research and Development CenterVicksburgUSA
  5. 5.Department of Ecosystems Science and ManagementTexas A&M UniversityCollege StationUSA

Personalised recommendations