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

Learn More →

Global land use intensity and the endangerment status of mammal species

Global land use intensity and the endangerment status of mammal species Introduction The globe is experiencing a biodiversity crisis in which humans play a primary role (Novacek & Cleland, ; Hoekstra et al ., ). Of all human activities, land use changes pose the greatest threat to biodiversity currently (Wilcove et al ., ; Czech et al ., ; Fischer & Lindenmayer, ) and over the next century (Sala et al ., ). Land use change in the form of urbanization raises particular concern for two reasons. First, human population growth rate is significantly greater in biodiversity hotspots compared to the global average (Cincotta et al ., ). Second, urban expansion proceeds at several times the rate of population growth (Pijanowski & Robinson, ; Seto et al ., ) owing to migration of people from rural and other non‐urban settlements to cities. Population growth also creates greater demand for food leading to conversion of land to agricultural use (Grau et al ., ), mostly in the form of high‐intensity crop production (Foley et al ., ). These changes indicate that not only are anthropogenic lands expanding, they are increasing in intensity, particularly where the world's biodiversity is greatest. Thus, to preserve the planet's biodiversity in the coming decades, it is necessary to understand the relationship between human land use intensity and species endangerment at a global scale. Species endangerment has been linked with a variety of human activities across the United States (Flather et al ., ; Wilcove et al ., ; Kirkland & Ostfeld, ; Czech et al ., ; Brown & Laband, ), Canada (Kerr & Cihlar, ; Kerr & Deguise, ), the Mediterranean biome (Underwood et al ., ) and the globe (Kerr & Currie, ; Mckinney, ; Davies et al ., ; Lenzen et al ., ). While these findings suggest that species are threatened by high levels of human activity as well as loss of suitable habitat, the role of land uses of different intensities in driving global biodiversity loss is largely unknown. The response of species to agricultural land use in particular is highly dependent on production intensity (Tscharntke et al ., ; Kleijn et al ., ). Evidence from North America (Kerr & Cihlar, ; Gibbs et al ., ) and Europe (Dormann et al ., ; Kleijn et al ., ) suggests that high levels of pesticide and fertilizer use associated with industrialized agriculture pose the greatest threat to species, whereas a landscape dominated by a patchwork of low‐intensity farming and grazing may help conserve biodiversity (Hayes & Holl, ; Zamora et al ., ; Sirami et al ., ). However, it has also been argued that reducing the intensity of agricultural lands may not be beneficial to species if it results in conversion of natural habitat to agriculture to meet the world's food demand (Green et al ., ). Urbanization, in the form of residential, commercial and industrial development, also poses a significant threat to species (Wilcove et al ., ; Czech et al ., ; Mckinney, ). Sub‐urban sprawl in particular is strongly linked with habitat fragmentation and reduced species diversity in North America (Maestas et al ., , ; Schwartz et al ., ) and Europe (Vermaat et al ., ). However, in the developing world, a much larger proportion of the human population resides in ‘villages’ than in cities or urban settlements (Ellis & Ramankutty, ). Anthropogenic causes of species endangerment and biodiversity loss have mainly been studied in North America (Flather et al ., ; Wilcove et al ., ; Kirkland & Ostfeld, ; Czech et al ., ; Kerr & Cihlar, ; Brown & Laband, ) and Europe (Dormann et al ., ; Vermaat et al ., ), and studies that incorporated developing parts of the world used socio‐economic factors not directly related to land use (Kerr & Currie, ; Mckinney, ; Davies et al ., ) or broad land use classes that do not reflect intensity (Lenzen et al ., ; Underwood et al ., ). Consequently, little is known regarding the relative impacts of urban settlements versus lower density settlements such as ‘villages’ on species diversity and endangerment. Because land use intensity encompasses many human activities, it is difficult to directly measure intensity at a global scale. The global anthropogenic biomes dataset integrates land use type with human population density to provide a proxy for the intensity of human‐environment interaction (Ellis & Ramankutty, ), and may reflect intensification in any one or several human activities including increased use of chemicals and fossil fuels and pressure from hunting and harvesting. Using this dataset, we assessed the relationship between species endangerment and the area of human settlements, croplands, rangelands and forestlands with variable population densities reflective of land use intensity. Previous studies have related species endangerment to land use and/or human activity using political units such as states (Kirkland & Ostfeld, ; Czech et al ., ; Brown & Laband, ) and countries (Kerr & Currie, ; Mckinney, ; Lenzen et al ., ), or other spatial units (Kerr & Deguise, ; Davies et al ., ; Underwood et al ., ). However, species distribution ranges are generally not restricted by human‐defined boundaries. Consequently, the endangerment status of species often reflects the level of human activity in several units (Lenzen et al ., ), which confounds analyses based on variation across spatial units. We related species endangerment directly to land use intensity by calculating the proportion of different land use classes within the full distribution range of all of the world's terrestrial mammal species and subspecies with a known endangerment status as defined by the IUCN Red List. We then tested the effect of each land use intensity class on the probability of having a more endangered status for all mammal species together as well as major mammal groups separately. In doing so, we sought to assess the role of different human land use and land management strategies in global biodiversity conservation. Methods Land use data The global anthropogenic biomes dataset is composed of six broad land use types including dense settlements, ‘village’ settlements, croplands, rangelands, forests and wildlands (Ellis & Ramankutty, ). The first five land use types are further broken down into classes with a specific population density class identified as dense (> 100 people km −2 ), residential (10–100 people km −2 ), populated (1–10 people km −2 ) or remote (< 1 person km −2 ) at a 0.08333 degree decimal resolution (Ellis & Ramankutty, ). Differences in population density within the same land use category are likely to reflect differences in socio‐economic factors that are associated with a particular intensity of use or management of the landscape (Ellis & Ramankutty, ). Thus, anthropogenic biome classes are highly useful for assessing land use impacts on the environment and the species that inhabit it. Wildlands (forested, sparse and barren) had no human population (Ellis & Ramankutty, ). Hence, they represent land cover with no direct human impact and are not likely to have an endangering effect on species. Consequently, the three classes of wildlands were summed into a single land use class for this study. A full list of all land use classes used in this study is provided in Table . The mean ( M ) and standard error ( SE ) of the proportion (%) of different land use classes within the distribution range of mammal species with different levels of endangerment, that is, least concern ( LC ), vulnerable ( VU ), endangered ( EN ) and critically endangered ( CE ) as defined by the IUCN R ed List. The M and SE of the area (10,000 × km 2 ) of distribution ranges ( distribution size ) is also shown for mammal species with different levels of endangerment LC NT VU EN CE n = 3166 n = 339 n = 526 n = 494 n = 209 M SE M SE M SE M SE M SE Dense settlements Urban 0.93 0.51 0.95 0.42 1.26 0.71 1.17 1.09 2.20 2.51 Dense 1.45 0.64 2.28 0.96 2.90 1.58 2.73 1.57 2.44 2.05 Village settlements Rice 0.66 0.59 0.60 0.50 1.02 1.16 1.12 1.20 0.49 0.86 Irrigated 0.63 0.50 0.78 0.54 0.49 0.47 0.37 0.42 0.04 0.07 Cropped 0.46 0.46 0.25 0.20 0.24 0.29 0.44 0.61 0.43 0.70 Pastoral 1.43 0.74 1.37 0.61 1.05 0.64 1.19 1.00 0.75 0.73 Rainfed 2.35 1.32 2.92 1.65 2.64 1.45 2.48 1.65 1.35 1.62 Mosaic 5.51 2.17 8.88 3.21 12.18 4.32 14.68 5.50 10.48 5.20 Croplands Res. irrigated 3.69 1.22 5.64 2.08 5.34 2.17 5.59 2.23 4.53 2.80 Res. rainfed 0.35 0.16 0.47 0.17 0.98 1.40 1.49 1.68 6.98 5.31 Pop. irrigated 1.08 0.45 1.39 0.58 1.33 0.84 1.60 1.20 1.69 1.95 Pop. rainfed 7.86 1.74 9.33 2.41 9.74 2.59 11.11 3.73 10.50 4.12 Remote 0.94 0.52 1.02 0.81 0.81 0.50 1.12 0.85 0.59 0.56 Rangelands Residential 8.09 2.60 6.48 2.53 5.71 2.71 6.58 3.40 6.06 3.74 Populated 9.68 2.75 8.13 2.64 5.85 2.71 5.46 2.84 5.78 3.69 Remote 11.30 4.29 8.83 4.37 5.74 3.58 5.11 3.40 6.74 4.35 Forests Populated 20.67 4.27 23.78 4.84 26.46 5.47 25.96 6.17 28.05 7.40 Remote 15.18 3.81 13.09 3.75 12.35 3.87 9.78 3.68 8.04 3.93 Wildlands All 7.73 3.45 3.82 2.46 3.92 2.80 2.01 1.94 2.85 2.74 Distribution size 206.89 98.31 95.22 61.04 49.53 68.91 15.22 14.97 6.26 6.60 *The number of species ( n ) is given for each endangerment group. Mean and standard error for the endangerment level with the smallest mean for a given land use class is shown in bold. For croplands, res. = residential and pop. = populated Mammal species data Digital distribution ranges of the world's currently extant (as of 2008) terrestrial mammal species and subspecies were obtained as a single shapefile from the IUCN Red List database (IUCN, ). The distribution range of each species/subspecies in the original shapefile was represented by single or multiple polygons of varying sizes depending on the size and connectedness of the species’ distribution range. We merged all the polygons belonging to the same species/subspecies using the dissolve tool in ArcGIS 10.1 (ESRI, ) to obtain a single polygon representing the full distribution range of a given species or subspecies. The status of some of the mapped species/subspecies was either missing, extinct in the wild or data deficient leaving a total of 4734 with a status of least concern (LC), vulnerable (VU), endangered (EN) or critically endangered (CE). Only species/subspecies (from here on referred to as species) displaying one of these classifications were used in further analyses. All placental mammal species were separated into major groups based on similarities in body size, taxonomy and/or feeding habits. ‘Primates’ (i.e. all species in the order Primates), ‘carnivores’ (i.e. all species in the order Carnivora) and ‘bats’ (i.e. all species in the order Chiroptera) were given their own groups. Even‐toed ungulates (i.e. all species in the order Artiodactyla), odd‐toed ungulates (i.e. all species in the order Perissodactyla) and elephant species (Elephantidae) were combined as ‘megaherbivores’. Rodents and small herbivores (i.e. all species in the orders Rodentia, Lagomorpha, Hyracoidea and Scandentia) were combined as ‘omniherbivores’. Mammal species that primarily feed on insects (i.e. all species in the orders Eulipotyphla, Afrosoricida, Cingulata, Erinaceomorpha, Macroscelidea, Pholidota, Pilosa, Soricomorpha, Tubulidentata) were combined as ‘insectivores’. All non‐placental mammals, that is, marsupials and monotremes, were placed into a single group ‘marsupials’. Determining land use proportions within species distribution range Each row of the species distribution shapefile was saved as a unique shapefile representing the full distribution area and extent of a single species. The proportion of different land use classes within the distribution range of each species could then be calculated. The reiterate raster tool was used along with the zonal statistics tool in ArcMap model builder (ESRI, ) to output as a table the number (count) of pixels of the particular land use class as well as the total number of pixels and area (km 2 ) within each species distribution range. The tables for all species were then merged and the proportion of each land use class was determined by dividing the number of pixels of each land use class by the total pixel count. Statistical analyses All statistical analyses were conducted using the open source software r (R Development Core Team, ). Descriptive statistics (mean and standard error) of land use (anthropogenic biome class) proportions within species distribution ranges with different levels of endangerment, that is, LC, NT, VU, EN, CE, and species distribution range size were calculated for each endangerment level to assess the correlation between land use and species endangerment. An ordered logistic regression (OLR) (Pyle et al ., ) was used to further test the effect of land use classes on endangerment status of all mammal species together as well as major mammal groups separately. Red List categories were ordered such that they reflected increasing endangerment with LC reflecting lowest and CE highest level of endangerment. OLR was then used to predict the probability of endangerment as a function of the independent variables (land use classes) with polr in the r package ‘MASS’ (Venables & Ripley, ). Accordingly, the OLR output indicates whether or not the endangerment status of a species is likely to increase as the proportion of a particular land use class within that species’ distribution range increases or decreases. The size of the coefficients associated with each land use class in the OLR output indicates the probability that the endangerment status of a species will increase as the proportion of that land use class within its distribution range changes. Remote forests and wildlands were input as a single variable because they represent natural land covers with no or insignificant human impact, hence they are both likely to have a ‘conserving’ effect on species. Variable selection was performed using a combination of Bayesian information criterion (BIC) and Akaike's information criterion (AIC). Initially, an exhaustive search was conducted using all land use classes with the function regsubsets in the ‘leaps’ package to determine subsets of land use classes with the lowest BIC. Where subsets had equally low BIC values, AIC was used to select the best fit model. All land use variables were standardized prior to analyses so that coefficients from the different OLR analyses (i.e. coefficients from model output for different mammal groups) could be compared. To illustrate the relative effect of the land use classes on mammal species endangerment, the probability of endangerment (for all mammal species together and species within major mammal groups) with increasing proportion of each land use class, as indicated by their coefficients in the OLR output, was summarized in Fig. 2. Results Variation in land use proportions and distribution range size with R ed L ist status Dense settlement classes comprised < 3% of species distribution ranges across all endangerment groups (Table ). However, the proportion of both urban and dense settlements was lowest within the distribution range of species with a Red List status of LC (Table ), and the proportion of urban settlements was 2.4 times higher and dense settlements 1.7 times higher within the distribution range of critically endangered species compared to LC species. In contrast, the proportions of ‘village’ classes were generally lowest within the distribution range of critically endangered species (Table ). The exception was rainfed mosaic ‘villages’, the proportion of which was 1.9 times greater within the distribution range of critically endangered species compared to species with a LC status (Table ). Species with a more endangered status also generally had the highest proportion of croplands relative to species with a less endangered status, with the exception of remote croplands which had the lowest proportion for critically endangered species (Table ). Proportions of populated and remote rangelands were lowest within the distribution range of endangered species, and residential rangelands within the distribution range of VU species (Table ). The proportion of populated forests was lowest within the distribution range of LC species, whereas the proportion of remote forest within the distribution range of LC species was nearly twice the proportion of remote forest within the distribution range of critically endangered species (Table ). The proportion of wildlands within the distribution range of LC species was also 2.7 and 3.8 times higher than the proportion of wildlands within the distribution range of critically endangered and endangered species, respectively (Table ). More endangered species had much smaller distribution ranges on average than less endangered species (Table ). Variable selection for OLR models The combination or subset of variables that produced the lowest BIC also had the lowest AIC for all mammal species together and for most of the mammal groups (Fig. ). The exceptions were carnivores (Fig. c) and megaherbivores (Fig. e). In these cases, BIC was used to select best fit model as the subset with the lowest BIC represented a more parsimonious model than the subset with the lowest AIC (Fig. c,e). The subset of variables that best explained the endangerment status of all mammal species included irrigated, pastoral, rainfed and mosaic ‘villages’, residential rainfed and remote croplands, residential, populated and remote rangelands, and wildlands and remote forests (Fig. a). While dense settlement classes and populated forests were not included in the best fit model for all species together, they were included in the best fit model for several mammal groups (Fig. ). Rice ‘villages’ were not included in the selected model for any of the mammal groups (Fig. ). Results of variable selection for regression models for all mammal species together and for each mammal group separately. Bayesian information criterion ( BIC ) and A kaike's information criterion ( AIC ) are shown for the five variable combinations with the lowest BIC for each group of mammals. The best fit combination used in final ordered logistic regression analysis is shown on top for each mammal group. For croplands, res. = residential and pop. = populated. Land use classes and the probability of having a more endangered status Human settlements were variably correlated with mammal endangerment status (Fig. ; Table ). Of all settlement classes, the area of mosaic ‘villages’ was most strongly correlated with the endangerment status of the greatest number of mammal groups. The probability of having a more endangered status for all mammal species together as well as megaherbivores, bats, omniherbivores, insectivores and marsupials separately increased with the proportion of mosaic ‘village’ settlements (Fig. ; Table ). Probability of having a more endangered status also increased for bats, omniherbivores and insectivores with area of urban settlements, for carnivores with proportion of dense settlements and for bats with proportion of cropped ‘villages’ (Fig. ; Table ). In contrast, there was a decrease in the probability of having a more endangered status for all mammal species together with the proportion of irrigated, cropped and rainfed ‘villages’ (Fig. ; Table ). Probability of having a more endangered status also decreased with the proportion of irrigated ‘villages’ for omniherbivores, pastoral ‘villages’ for carnivores, bats and omniherbivores, and rainfed ‘villages’ for primates and bats (Fig. ; Table ). Similarly, urban and dense settlements were negatively correlated with the probability of having a more endangered status for marsupials and megaherbivores, respectively (Fig. ; Table ). Summary of land use effects on mammal endangerment based on ordered logistic regression results. Land use classes with a positive effect on the probability of having a more endangered status are indicated by a + sign. Land use classes with a negative effect on the probability of having a more endangered status are indicated by a − sign. Darker shades indicate stronger effect based on the regression coefficients. For croplands, res. = residential and pop. = populated. Coefficients of land use classes included in ordered logistic in regression ( OLR ) model for all mammal species together as well as mammal groups separately All species Primates Carnivores Megaherbivores Bats Omniherbivores Insectivores Marsupials n = 4735 n = 533 n = 238 n = 315 n = 937 n = 1936 n = 465 n = 310 Est. SE Est. SE Est. SE Est. SE Est. SE Est. SE Est. SE Est. SE Dense settlements Urban – – – – – – – – 0.32 0.07 0.09 0.04 0.23 0.12 −3.35 1.12 Dense – – – – 0.57 0.15 −1.12 0.31 – – – – – – – – Village settlements Rice – – – – – – – – – – – – – – – – Irrigated −0.16 0.05 – – – – – – – – −0.40 0.11 – – – – Cropped – – – – – – – – 0.22 0.11 – – – – – – Pastoral −0.27 0.04 – – −0.61 0.28 – – −0.32 0.14 −0.55 0.10 – – – – Rainfed −0.15 0.04 −0.29 0.08 – – – – −0.34 0.12 – – – – – – Mosaic 0.13 0.03 – – – – 0.40 0.17 0.57 0.09 0.12 0.05 0.45 0.09 0.53 0.22 Croplands Res. irrigated – – – – – – 0.33 0.11 0.23 0.09 – – −1.13 0.26 – – Res. rainfed 0.22 0.04 – – – – – – 1.58 0.55 0.31 0.08 0.29 0.07 2.12 1.08 Pop. irrigated – – 0.19 0.09 – – 0.12 0.05 −0.22 0.14 – – −1.93 0.52 – – Pop. rainfed – – 0.19 0.07 – – – – – – – – 0.71 0.12 – – Remote −0.08 0.03 – – – – – – – – −0.18 0.06 – – – – Rangelands Residential −0.14 0.04 – – – – – – – – −0.18 0.08 – – – – Populated −0.32 0.05 −0.86 0.14 – – – – – – −0.35 0.08 −0.69 0.19 – – Remote −0.25 0.04 – – – – – – – – −0.17 0.06 – – – – Forests Populated – – – – 0.60 0.15 – – 0.55 0.10 – – – – 0.36 0.10 Wildlands & remote forests All −0.61 0.05 −0.79 0.09 – – – – −0.46 0.15 −0.71 0.09 −0.71 0.22 – – The estimate (Est.) and standard error (SE) are given for land use classes included in each of the models. Land use classes with a negative effect on species endangerment are shown in bold. Significance level of each variable/class is indicated by asterisks where (0.05 ≥ P > 0.01), (0.01 ≥ P > 0.005) and (0.005 ≥ P ). Number of species ( n ) within each mammal group is also given. For croplands, res. = residential and pop. = populated. Cropland classes, except remote croplands, were generally positively correlated with mammal species endangerment (Fig. ; Table ). The probability of having a more endangered status for all mammal species together increased only with the proportion of residential rainfed croplands (Fig. ; Table ). However, the probability of having a more endangered status increased with proportion of (1) residential irrigated croplands for megaherbivores and bats, (2) residential rainfed croplands for bats, omniherbivores, insectivores and marsupials, (3) populated irrigated croplands for primates and megaherbivores, and (4) populated rainfed croplands for primates and insectivores (Fig. ; Table ). In contrast, the probability of having a more endangered status decreased with proportion of (1) residential irrigated croplands for insectivores, (2) populated irrigated croplands for bats and insectivores, and (3) remote croplands for omniherbivores (Fig. ; Table ). Remote croplands were also negatively correlated with the probability of having a more endangered status for all mammal species together (Fig. ; Table ). In contrast to croplands and human settlement classes, rangelands only displayed a negative correlation with endangerment (Fig. ; Table ). Probability of having a more endangered status decreased for all mammal species together as the proportion of all three rangeland classes increased (Fig. ; Table ). The probability of having a more endangered status also decreased with increasing proportion of (1) residential rangelands for omniherbivores, (2) populated rangelands for primates, omniherbivores and marsupials, and (3) remote rangelands for omniherbivores (Fig. ; Table ). As expected, the probability of having a more endangered status decreased as the combined proportion of wildlands and remote forests within species distribution ranges increased for all mammal species together and all mammal groups except megaherbivores, bats and marsupials (Fig. ; Table ). In contrast, the probability of having a more endangered status increased for carnivores, bats and marsupials as the proportion of populated forests within their distribution ranges increased (Fig. ; Table ). Discussion Kerr & Currie ( ), Mckinney ( ) and Davies et al . ( ) showed that the proportion of endangered species increases with human population density across the globe, suggesting that biodiversity loss is driven by the intensity of human activity. However, ours is the first global study to show that the probability of species endangerment increases with the proportion of some types of human settlement while decreasing with others. Several studies have shown that species endangerment increases with urbanization at national or regional scales (Wilcove et al ., ; Czech et al ., ; Underwood et al ., ). In contrast, at a global scale, Lenzen et al . ( ) found that built‐up areas had a ‘conserving influence’ on mammal and bird species. Because their study did not distinguish between different intensities of land use, the endangering effect of urban areas may have been masked where certain lower‐intensity settlement classes dominate the ‘built‐up’ or urban landscape. However, our results also suggest that some low‐intensity settlements, such as mosaic ‘villages’, may pose as great of a threat to mammal species as high‐intensity urban settlements. Urban settlements and mosaic ‘villages’ together comprise a large majority of human settlement area in the developed world such as North America, whereas irrigated, pastoral and rainfed ‘villages’ are several times more extensive in the developing world (Ellis & Ramankutty, ). Thus, the decrease in endangerment probability observed with increasing proportion of low‐intensity ‘villages’ in our study may reflect human activities associated with a less industrialized or a more traditional lifestyle. However, our results suggest that croplands, particularly high‐intensity croplands, may pose an even greater threat to mammal species than human settlements. Species endangerment has previously been shown to be correlated with cropland area at an ecoregion scale across Canada (Gibbs et al ., ) and at a national scale across the globe (Lenzen et al ., ). Analyses of criteria for Red‐Listing species have also revealed that agricultural production is among the most important causes of species endangerment for mammals globally (Hayward, ) and for a variety of species in North America (Czech et al ., ). While the role of agriculture in endangering species is relatively well established, our results further suggest that high‐intensity croplands pose a much greater threat to mammal species than lower‐intensity cropland classes. Very low‐intensity croplands may actually help maintain biodiversity at a global scale as evidenced by the decreasing endangerment probability, especially of omniherbivores, observed in our study with increasing area of remote croplands. Residential croplands are likely to use more fertilizers and chemicals than cropland classes with lower population densities, such as populated or remote croplands, under similar environmental conditions (Ellis & Ramankutty, ). Industrialized agriculture with high chemical inputs is a strong driver of species endangerment in North America (Kerr & Cihlar, ; Gibbs et al ., ). Several studies from Europe have also shown that high levels of pesticide and fertilizer use results in loss of species at a regional scale (Dormann et al ., ; Kleijn et al ., ). Thus, the strong endangering effect of residential croplands observed in our study may be a consequence of extensive use of large quantities of agricultural chemicals. However, other factors associated either with industrialized agriculture, for example, mechanization and the use of fossil fuels, or pressures more directly associated with a high rural human population, for example, hunting, trapping or poisoning of animals, may also play a role. Residential croplands constitute 70% of the world's cropland area, while populated and remote croplands constitute 26% and 4%, respectively (Ellis & Ramankutty, ). Thus, agricultural industrialization poses an especially big challenge for global biodiversity conservation. Our finding that the endangerment probability of mammals decreases with the proportion of rangelands is also unique. While Lenzen et al . ( ) found a slight positive correlation between species endangerment and pastureland globally, rangelands have been shown to support greater number of species, particularly native species, than urban lands at regional scales (Maestas et al ., , ). It has also been suggested that grazing by domestic animals can help maintain rare species (Zamora et al ., ; Sirami et al ., ) in the landscape which would otherwise become locally extinct if the land was not managed by humans. Thus, our findings regarding rangelands may be explained by the fact that rangelands provide better habitat for wildlife than most other anthropogenic land uses as well as former pasture or farm land that has been abandoned and converted into forest or other vegetation cover. Given that rangelands constitute approximately 30% of the earth's land area (Ellis & Ramankutty, ), they are likely to have an especially important role in maintaining biodiversity at a global scale. Loss of wild or undisturbed habitat may also be an important driver of species endangerment as, according to the OLR results, the endangerment probability of most mammal groups decreases as the proportion of wildlands and remote forests within their distribution ranges increases. Furthermore, total distribution range size of highly endangered species was much lower than the range size of less endangered species (Table ) suggesting that a smaller area of potential habitat makes species more prone to endangerment. Thus, a reduction in the intensity of current land uses, such as croplands, may not benefit species if it results in conversion of natural habitat to agriculture. Species endangerment has previously been shown to depend on area of natural habitat as well as species distribution range size (Czech et al ., ; Kerr & Deguise, ). However, our results suggest that not all land with natural vegetation cover provides good habitat for species as the endangerment probability of several mammal groups increases as the proportion of populated forests within their distribution ranges increases. Thus, species may not necessarily benefit from having more area with natural vegetation cover in their distribution range if these areas are over utilized or not managed properly. There has recently been great interest in identifying areas of exceptionally high biodiversity value and/or threat (Olson & Dinerstein, ; Myers et al ., ) and determining changes in land use patterns within them (Scharlemann et al ., ; Hoekstra et al ., ; Grau et al ., ). However, few attempts have been made to explore the links between socio‐economic drivers, human land use or management strategies, and biodiversity loss at regional or global scales. Over the next several decades, there will be a huge increase in the world's urban population (Seto et al ., ), mostly owing to urbanization in developing parts of the world such as south Asia, Africa and Latin America, where natural habitat loss is already highly extensive (Hoekstra et al ., ). This increase in urban population will put greater pressure on the world's species not only by causing conversion of natural land to human uses, but also through increases in the intensity of land use on existing anthropogenic lands. Consequently, combining knowledge regarding the socio‐economic drivers of land use change and the role of different land use types and intensities in endangering or maintaining species diversity may be the key to making successful land use policy/mitigation decisions for conserving the world's biodiversity. In summary, 75% of the earth's ice‐free terrestrial surface has been altered and is used by humans to some extent (Ellis & Ramankutty, ), and the area of wild habitat remaining in the world's biodiversity hotspots is especially small (Hoekstra et al ., ). Thus, determining the potential of human land uses in providing habitat for species is highly important for global biodiversity conservation (Novacek & Cleland, ). To this end, we found that some human land use types, for example, rangelands, are better habitat for a larger proportion of mammal species than other human land use types, for example, croplands or human settlements, and even some natural lands with high human activity, such as, populated forests. Furthermore, a low‐intensity land use, for example, remote croplands, provides better habitat for species than the same type of land use with a higher intensity, for example, residential and populated croplands. These results together demonstrate that some human settlement and food production strategies are better than others and may actually help maintain biodiversity at a global scale. Acknowledgements We thank J. Doucette and J. Plourde at the Department of Forestry and Natural Resources at Purdue University for their GIS help, and J. Datta and B. Craig at the Department of Statistics at Purdue University for their advice regarding the statistical methods used in this study. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Diversity and Distributions Wiley

Global land use intensity and the endangerment status of mammal species

Download PDF
10 pages

Loading next page...
 
/lp/wiley/global-land-use-intensity-and-the-endangerment-status-of-mammal-dbPt5LVtRo

References (46)

Publisher
Wiley
Copyright
Copyright © 2012 Blackwell Publishing Ltd
ISSN
1366-9516
eISSN
1472-4642
DOI
10.1111/j.1472-4642.2012.00928.x
Publisher site
See Article on Publisher Site

Abstract

Introduction The globe is experiencing a biodiversity crisis in which humans play a primary role (Novacek & Cleland, ; Hoekstra et al ., ). Of all human activities, land use changes pose the greatest threat to biodiversity currently (Wilcove et al ., ; Czech et al ., ; Fischer & Lindenmayer, ) and over the next century (Sala et al ., ). Land use change in the form of urbanization raises particular concern for two reasons. First, human population growth rate is significantly greater in biodiversity hotspots compared to the global average (Cincotta et al ., ). Second, urban expansion proceeds at several times the rate of population growth (Pijanowski & Robinson, ; Seto et al ., ) owing to migration of people from rural and other non‐urban settlements to cities. Population growth also creates greater demand for food leading to conversion of land to agricultural use (Grau et al ., ), mostly in the form of high‐intensity crop production (Foley et al ., ). These changes indicate that not only are anthropogenic lands expanding, they are increasing in intensity, particularly where the world's biodiversity is greatest. Thus, to preserve the planet's biodiversity in the coming decades, it is necessary to understand the relationship between human land use intensity and species endangerment at a global scale. Species endangerment has been linked with a variety of human activities across the United States (Flather et al ., ; Wilcove et al ., ; Kirkland & Ostfeld, ; Czech et al ., ; Brown & Laband, ), Canada (Kerr & Cihlar, ; Kerr & Deguise, ), the Mediterranean biome (Underwood et al ., ) and the globe (Kerr & Currie, ; Mckinney, ; Davies et al ., ; Lenzen et al ., ). While these findings suggest that species are threatened by high levels of human activity as well as loss of suitable habitat, the role of land uses of different intensities in driving global biodiversity loss is largely unknown. The response of species to agricultural land use in particular is highly dependent on production intensity (Tscharntke et al ., ; Kleijn et al ., ). Evidence from North America (Kerr & Cihlar, ; Gibbs et al ., ) and Europe (Dormann et al ., ; Kleijn et al ., ) suggests that high levels of pesticide and fertilizer use associated with industrialized agriculture pose the greatest threat to species, whereas a landscape dominated by a patchwork of low‐intensity farming and grazing may help conserve biodiversity (Hayes & Holl, ; Zamora et al ., ; Sirami et al ., ). However, it has also been argued that reducing the intensity of agricultural lands may not be beneficial to species if it results in conversion of natural habitat to agriculture to meet the world's food demand (Green et al ., ). Urbanization, in the form of residential, commercial and industrial development, also poses a significant threat to species (Wilcove et al ., ; Czech et al ., ; Mckinney, ). Sub‐urban sprawl in particular is strongly linked with habitat fragmentation and reduced species diversity in North America (Maestas et al ., , ; Schwartz et al ., ) and Europe (Vermaat et al ., ). However, in the developing world, a much larger proportion of the human population resides in ‘villages’ than in cities or urban settlements (Ellis & Ramankutty, ). Anthropogenic causes of species endangerment and biodiversity loss have mainly been studied in North America (Flather et al ., ; Wilcove et al ., ; Kirkland & Ostfeld, ; Czech et al ., ; Kerr & Cihlar, ; Brown & Laband, ) and Europe (Dormann et al ., ; Vermaat et al ., ), and studies that incorporated developing parts of the world used socio‐economic factors not directly related to land use (Kerr & Currie, ; Mckinney, ; Davies et al ., ) or broad land use classes that do not reflect intensity (Lenzen et al ., ; Underwood et al ., ). Consequently, little is known regarding the relative impacts of urban settlements versus lower density settlements such as ‘villages’ on species diversity and endangerment. Because land use intensity encompasses many human activities, it is difficult to directly measure intensity at a global scale. The global anthropogenic biomes dataset integrates land use type with human population density to provide a proxy for the intensity of human‐environment interaction (Ellis & Ramankutty, ), and may reflect intensification in any one or several human activities including increased use of chemicals and fossil fuels and pressure from hunting and harvesting. Using this dataset, we assessed the relationship between species endangerment and the area of human settlements, croplands, rangelands and forestlands with variable population densities reflective of land use intensity. Previous studies have related species endangerment to land use and/or human activity using political units such as states (Kirkland & Ostfeld, ; Czech et al ., ; Brown & Laband, ) and countries (Kerr & Currie, ; Mckinney, ; Lenzen et al ., ), or other spatial units (Kerr & Deguise, ; Davies et al ., ; Underwood et al ., ). However, species distribution ranges are generally not restricted by human‐defined boundaries. Consequently, the endangerment status of species often reflects the level of human activity in several units (Lenzen et al ., ), which confounds analyses based on variation across spatial units. We related species endangerment directly to land use intensity by calculating the proportion of different land use classes within the full distribution range of all of the world's terrestrial mammal species and subspecies with a known endangerment status as defined by the IUCN Red List. We then tested the effect of each land use intensity class on the probability of having a more endangered status for all mammal species together as well as major mammal groups separately. In doing so, we sought to assess the role of different human land use and land management strategies in global biodiversity conservation. Methods Land use data The global anthropogenic biomes dataset is composed of six broad land use types including dense settlements, ‘village’ settlements, croplands, rangelands, forests and wildlands (Ellis & Ramankutty, ). The first five land use types are further broken down into classes with a specific population density class identified as dense (> 100 people km −2 ), residential (10–100 people km −2 ), populated (1–10 people km −2 ) or remote (< 1 person km −2 ) at a 0.08333 degree decimal resolution (Ellis & Ramankutty, ). Differences in population density within the same land use category are likely to reflect differences in socio‐economic factors that are associated with a particular intensity of use or management of the landscape (Ellis & Ramankutty, ). Thus, anthropogenic biome classes are highly useful for assessing land use impacts on the environment and the species that inhabit it. Wildlands (forested, sparse and barren) had no human population (Ellis & Ramankutty, ). Hence, they represent land cover with no direct human impact and are not likely to have an endangering effect on species. Consequently, the three classes of wildlands were summed into a single land use class for this study. A full list of all land use classes used in this study is provided in Table . The mean ( M ) and standard error ( SE ) of the proportion (%) of different land use classes within the distribution range of mammal species with different levels of endangerment, that is, least concern ( LC ), vulnerable ( VU ), endangered ( EN ) and critically endangered ( CE ) as defined by the IUCN R ed List. The M and SE of the area (10,000 × km 2 ) of distribution ranges ( distribution size ) is also shown for mammal species with different levels of endangerment LC NT VU EN CE n = 3166 n = 339 n = 526 n = 494 n = 209 M SE M SE M SE M SE M SE Dense settlements Urban 0.93 0.51 0.95 0.42 1.26 0.71 1.17 1.09 2.20 2.51 Dense 1.45 0.64 2.28 0.96 2.90 1.58 2.73 1.57 2.44 2.05 Village settlements Rice 0.66 0.59 0.60 0.50 1.02 1.16 1.12 1.20 0.49 0.86 Irrigated 0.63 0.50 0.78 0.54 0.49 0.47 0.37 0.42 0.04 0.07 Cropped 0.46 0.46 0.25 0.20 0.24 0.29 0.44 0.61 0.43 0.70 Pastoral 1.43 0.74 1.37 0.61 1.05 0.64 1.19 1.00 0.75 0.73 Rainfed 2.35 1.32 2.92 1.65 2.64 1.45 2.48 1.65 1.35 1.62 Mosaic 5.51 2.17 8.88 3.21 12.18 4.32 14.68 5.50 10.48 5.20 Croplands Res. irrigated 3.69 1.22 5.64 2.08 5.34 2.17 5.59 2.23 4.53 2.80 Res. rainfed 0.35 0.16 0.47 0.17 0.98 1.40 1.49 1.68 6.98 5.31 Pop. irrigated 1.08 0.45 1.39 0.58 1.33 0.84 1.60 1.20 1.69 1.95 Pop. rainfed 7.86 1.74 9.33 2.41 9.74 2.59 11.11 3.73 10.50 4.12 Remote 0.94 0.52 1.02 0.81 0.81 0.50 1.12 0.85 0.59 0.56 Rangelands Residential 8.09 2.60 6.48 2.53 5.71 2.71 6.58 3.40 6.06 3.74 Populated 9.68 2.75 8.13 2.64 5.85 2.71 5.46 2.84 5.78 3.69 Remote 11.30 4.29 8.83 4.37 5.74 3.58 5.11 3.40 6.74 4.35 Forests Populated 20.67 4.27 23.78 4.84 26.46 5.47 25.96 6.17 28.05 7.40 Remote 15.18 3.81 13.09 3.75 12.35 3.87 9.78 3.68 8.04 3.93 Wildlands All 7.73 3.45 3.82 2.46 3.92 2.80 2.01 1.94 2.85 2.74 Distribution size 206.89 98.31 95.22 61.04 49.53 68.91 15.22 14.97 6.26 6.60 *The number of species ( n ) is given for each endangerment group. Mean and standard error for the endangerment level with the smallest mean for a given land use class is shown in bold. For croplands, res. = residential and pop. = populated Mammal species data Digital distribution ranges of the world's currently extant (as of 2008) terrestrial mammal species and subspecies were obtained as a single shapefile from the IUCN Red List database (IUCN, ). The distribution range of each species/subspecies in the original shapefile was represented by single or multiple polygons of varying sizes depending on the size and connectedness of the species’ distribution range. We merged all the polygons belonging to the same species/subspecies using the dissolve tool in ArcGIS 10.1 (ESRI, ) to obtain a single polygon representing the full distribution range of a given species or subspecies. The status of some of the mapped species/subspecies was either missing, extinct in the wild or data deficient leaving a total of 4734 with a status of least concern (LC), vulnerable (VU), endangered (EN) or critically endangered (CE). Only species/subspecies (from here on referred to as species) displaying one of these classifications were used in further analyses. All placental mammal species were separated into major groups based on similarities in body size, taxonomy and/or feeding habits. ‘Primates’ (i.e. all species in the order Primates), ‘carnivores’ (i.e. all species in the order Carnivora) and ‘bats’ (i.e. all species in the order Chiroptera) were given their own groups. Even‐toed ungulates (i.e. all species in the order Artiodactyla), odd‐toed ungulates (i.e. all species in the order Perissodactyla) and elephant species (Elephantidae) were combined as ‘megaherbivores’. Rodents and small herbivores (i.e. all species in the orders Rodentia, Lagomorpha, Hyracoidea and Scandentia) were combined as ‘omniherbivores’. Mammal species that primarily feed on insects (i.e. all species in the orders Eulipotyphla, Afrosoricida, Cingulata, Erinaceomorpha, Macroscelidea, Pholidota, Pilosa, Soricomorpha, Tubulidentata) were combined as ‘insectivores’. All non‐placental mammals, that is, marsupials and monotremes, were placed into a single group ‘marsupials’. Determining land use proportions within species distribution range Each row of the species distribution shapefile was saved as a unique shapefile representing the full distribution area and extent of a single species. The proportion of different land use classes within the distribution range of each species could then be calculated. The reiterate raster tool was used along with the zonal statistics tool in ArcMap model builder (ESRI, ) to output as a table the number (count) of pixels of the particular land use class as well as the total number of pixels and area (km 2 ) within each species distribution range. The tables for all species were then merged and the proportion of each land use class was determined by dividing the number of pixels of each land use class by the total pixel count. Statistical analyses All statistical analyses were conducted using the open source software r (R Development Core Team, ). Descriptive statistics (mean and standard error) of land use (anthropogenic biome class) proportions within species distribution ranges with different levels of endangerment, that is, LC, NT, VU, EN, CE, and species distribution range size were calculated for each endangerment level to assess the correlation between land use and species endangerment. An ordered logistic regression (OLR) (Pyle et al ., ) was used to further test the effect of land use classes on endangerment status of all mammal species together as well as major mammal groups separately. Red List categories were ordered such that they reflected increasing endangerment with LC reflecting lowest and CE highest level of endangerment. OLR was then used to predict the probability of endangerment as a function of the independent variables (land use classes) with polr in the r package ‘MASS’ (Venables & Ripley, ). Accordingly, the OLR output indicates whether or not the endangerment status of a species is likely to increase as the proportion of a particular land use class within that species’ distribution range increases or decreases. The size of the coefficients associated with each land use class in the OLR output indicates the probability that the endangerment status of a species will increase as the proportion of that land use class within its distribution range changes. Remote forests and wildlands were input as a single variable because they represent natural land covers with no or insignificant human impact, hence they are both likely to have a ‘conserving’ effect on species. Variable selection was performed using a combination of Bayesian information criterion (BIC) and Akaike's information criterion (AIC). Initially, an exhaustive search was conducted using all land use classes with the function regsubsets in the ‘leaps’ package to determine subsets of land use classes with the lowest BIC. Where subsets had equally low BIC values, AIC was used to select the best fit model. All land use variables were standardized prior to analyses so that coefficients from the different OLR analyses (i.e. coefficients from model output for different mammal groups) could be compared. To illustrate the relative effect of the land use classes on mammal species endangerment, the probability of endangerment (for all mammal species together and species within major mammal groups) with increasing proportion of each land use class, as indicated by their coefficients in the OLR output, was summarized in Fig. 2. Results Variation in land use proportions and distribution range size with R ed L ist status Dense settlement classes comprised < 3% of species distribution ranges across all endangerment groups (Table ). However, the proportion of both urban and dense settlements was lowest within the distribution range of species with a Red List status of LC (Table ), and the proportion of urban settlements was 2.4 times higher and dense settlements 1.7 times higher within the distribution range of critically endangered species compared to LC species. In contrast, the proportions of ‘village’ classes were generally lowest within the distribution range of critically endangered species (Table ). The exception was rainfed mosaic ‘villages’, the proportion of which was 1.9 times greater within the distribution range of critically endangered species compared to species with a LC status (Table ). Species with a more endangered status also generally had the highest proportion of croplands relative to species with a less endangered status, with the exception of remote croplands which had the lowest proportion for critically endangered species (Table ). Proportions of populated and remote rangelands were lowest within the distribution range of endangered species, and residential rangelands within the distribution range of VU species (Table ). The proportion of populated forests was lowest within the distribution range of LC species, whereas the proportion of remote forest within the distribution range of LC species was nearly twice the proportion of remote forest within the distribution range of critically endangered species (Table ). The proportion of wildlands within the distribution range of LC species was also 2.7 and 3.8 times higher than the proportion of wildlands within the distribution range of critically endangered and endangered species, respectively (Table ). More endangered species had much smaller distribution ranges on average than less endangered species (Table ). Variable selection for OLR models The combination or subset of variables that produced the lowest BIC also had the lowest AIC for all mammal species together and for most of the mammal groups (Fig. ). The exceptions were carnivores (Fig. c) and megaherbivores (Fig. e). In these cases, BIC was used to select best fit model as the subset with the lowest BIC represented a more parsimonious model than the subset with the lowest AIC (Fig. c,e). The subset of variables that best explained the endangerment status of all mammal species included irrigated, pastoral, rainfed and mosaic ‘villages’, residential rainfed and remote croplands, residential, populated and remote rangelands, and wildlands and remote forests (Fig. a). While dense settlement classes and populated forests were not included in the best fit model for all species together, they were included in the best fit model for several mammal groups (Fig. ). Rice ‘villages’ were not included in the selected model for any of the mammal groups (Fig. ). Results of variable selection for regression models for all mammal species together and for each mammal group separately. Bayesian information criterion ( BIC ) and A kaike's information criterion ( AIC ) are shown for the five variable combinations with the lowest BIC for each group of mammals. The best fit combination used in final ordered logistic regression analysis is shown on top for each mammal group. For croplands, res. = residential and pop. = populated. Land use classes and the probability of having a more endangered status Human settlements were variably correlated with mammal endangerment status (Fig. ; Table ). Of all settlement classes, the area of mosaic ‘villages’ was most strongly correlated with the endangerment status of the greatest number of mammal groups. The probability of having a more endangered status for all mammal species together as well as megaherbivores, bats, omniherbivores, insectivores and marsupials separately increased with the proportion of mosaic ‘village’ settlements (Fig. ; Table ). Probability of having a more endangered status also increased for bats, omniherbivores and insectivores with area of urban settlements, for carnivores with proportion of dense settlements and for bats with proportion of cropped ‘villages’ (Fig. ; Table ). In contrast, there was a decrease in the probability of having a more endangered status for all mammal species together with the proportion of irrigated, cropped and rainfed ‘villages’ (Fig. ; Table ). Probability of having a more endangered status also decreased with the proportion of irrigated ‘villages’ for omniherbivores, pastoral ‘villages’ for carnivores, bats and omniherbivores, and rainfed ‘villages’ for primates and bats (Fig. ; Table ). Similarly, urban and dense settlements were negatively correlated with the probability of having a more endangered status for marsupials and megaherbivores, respectively (Fig. ; Table ). Summary of land use effects on mammal endangerment based on ordered logistic regression results. Land use classes with a positive effect on the probability of having a more endangered status are indicated by a + sign. Land use classes with a negative effect on the probability of having a more endangered status are indicated by a − sign. Darker shades indicate stronger effect based on the regression coefficients. For croplands, res. = residential and pop. = populated. Coefficients of land use classes included in ordered logistic in regression ( OLR ) model for all mammal species together as well as mammal groups separately All species Primates Carnivores Megaherbivores Bats Omniherbivores Insectivores Marsupials n = 4735 n = 533 n = 238 n = 315 n = 937 n = 1936 n = 465 n = 310 Est. SE Est. SE Est. SE Est. SE Est. SE Est. SE Est. SE Est. SE Dense settlements Urban – – – – – – – – 0.32 0.07 0.09 0.04 0.23 0.12 −3.35 1.12 Dense – – – – 0.57 0.15 −1.12 0.31 – – – – – – – – Village settlements Rice – – – – – – – – – – – – – – – – Irrigated −0.16 0.05 – – – – – – – – −0.40 0.11 – – – – Cropped – – – – – – – – 0.22 0.11 – – – – – – Pastoral −0.27 0.04 – – −0.61 0.28 – – −0.32 0.14 −0.55 0.10 – – – – Rainfed −0.15 0.04 −0.29 0.08 – – – – −0.34 0.12 – – – – – – Mosaic 0.13 0.03 – – – – 0.40 0.17 0.57 0.09 0.12 0.05 0.45 0.09 0.53 0.22 Croplands Res. irrigated – – – – – – 0.33 0.11 0.23 0.09 – – −1.13 0.26 – – Res. rainfed 0.22 0.04 – – – – – – 1.58 0.55 0.31 0.08 0.29 0.07 2.12 1.08 Pop. irrigated – – 0.19 0.09 – – 0.12 0.05 −0.22 0.14 – – −1.93 0.52 – – Pop. rainfed – – 0.19 0.07 – – – – – – – – 0.71 0.12 – – Remote −0.08 0.03 – – – – – – – – −0.18 0.06 – – – – Rangelands Residential −0.14 0.04 – – – – – – – – −0.18 0.08 – – – – Populated −0.32 0.05 −0.86 0.14 – – – – – – −0.35 0.08 −0.69 0.19 – – Remote −0.25 0.04 – – – – – – – – −0.17 0.06 – – – – Forests Populated – – – – 0.60 0.15 – – 0.55 0.10 – – – – 0.36 0.10 Wildlands & remote forests All −0.61 0.05 −0.79 0.09 – – – – −0.46 0.15 −0.71 0.09 −0.71 0.22 – – The estimate (Est.) and standard error (SE) are given for land use classes included in each of the models. Land use classes with a negative effect on species endangerment are shown in bold. Significance level of each variable/class is indicated by asterisks where (0.05 ≥ P > 0.01), (0.01 ≥ P > 0.005) and (0.005 ≥ P ). Number of species ( n ) within each mammal group is also given. For croplands, res. = residential and pop. = populated. Cropland classes, except remote croplands, were generally positively correlated with mammal species endangerment (Fig. ; Table ). The probability of having a more endangered status for all mammal species together increased only with the proportion of residential rainfed croplands (Fig. ; Table ). However, the probability of having a more endangered status increased with proportion of (1) residential irrigated croplands for megaherbivores and bats, (2) residential rainfed croplands for bats, omniherbivores, insectivores and marsupials, (3) populated irrigated croplands for primates and megaherbivores, and (4) populated rainfed croplands for primates and insectivores (Fig. ; Table ). In contrast, the probability of having a more endangered status decreased with proportion of (1) residential irrigated croplands for insectivores, (2) populated irrigated croplands for bats and insectivores, and (3) remote croplands for omniherbivores (Fig. ; Table ). Remote croplands were also negatively correlated with the probability of having a more endangered status for all mammal species together (Fig. ; Table ). In contrast to croplands and human settlement classes, rangelands only displayed a negative correlation with endangerment (Fig. ; Table ). Probability of having a more endangered status decreased for all mammal species together as the proportion of all three rangeland classes increased (Fig. ; Table ). The probability of having a more endangered status also decreased with increasing proportion of (1) residential rangelands for omniherbivores, (2) populated rangelands for primates, omniherbivores and marsupials, and (3) remote rangelands for omniherbivores (Fig. ; Table ). As expected, the probability of having a more endangered status decreased as the combined proportion of wildlands and remote forests within species distribution ranges increased for all mammal species together and all mammal groups except megaherbivores, bats and marsupials (Fig. ; Table ). In contrast, the probability of having a more endangered status increased for carnivores, bats and marsupials as the proportion of populated forests within their distribution ranges increased (Fig. ; Table ). Discussion Kerr & Currie ( ), Mckinney ( ) and Davies et al . ( ) showed that the proportion of endangered species increases with human population density across the globe, suggesting that biodiversity loss is driven by the intensity of human activity. However, ours is the first global study to show that the probability of species endangerment increases with the proportion of some types of human settlement while decreasing with others. Several studies have shown that species endangerment increases with urbanization at national or regional scales (Wilcove et al ., ; Czech et al ., ; Underwood et al ., ). In contrast, at a global scale, Lenzen et al . ( ) found that built‐up areas had a ‘conserving influence’ on mammal and bird species. Because their study did not distinguish between different intensities of land use, the endangering effect of urban areas may have been masked where certain lower‐intensity settlement classes dominate the ‘built‐up’ or urban landscape. However, our results also suggest that some low‐intensity settlements, such as mosaic ‘villages’, may pose as great of a threat to mammal species as high‐intensity urban settlements. Urban settlements and mosaic ‘villages’ together comprise a large majority of human settlement area in the developed world such as North America, whereas irrigated, pastoral and rainfed ‘villages’ are several times more extensive in the developing world (Ellis & Ramankutty, ). Thus, the decrease in endangerment probability observed with increasing proportion of low‐intensity ‘villages’ in our study may reflect human activities associated with a less industrialized or a more traditional lifestyle. However, our results suggest that croplands, particularly high‐intensity croplands, may pose an even greater threat to mammal species than human settlements. Species endangerment has previously been shown to be correlated with cropland area at an ecoregion scale across Canada (Gibbs et al ., ) and at a national scale across the globe (Lenzen et al ., ). Analyses of criteria for Red‐Listing species have also revealed that agricultural production is among the most important causes of species endangerment for mammals globally (Hayward, ) and for a variety of species in North America (Czech et al ., ). While the role of agriculture in endangering species is relatively well established, our results further suggest that high‐intensity croplands pose a much greater threat to mammal species than lower‐intensity cropland classes. Very low‐intensity croplands may actually help maintain biodiversity at a global scale as evidenced by the decreasing endangerment probability, especially of omniherbivores, observed in our study with increasing area of remote croplands. Residential croplands are likely to use more fertilizers and chemicals than cropland classes with lower population densities, such as populated or remote croplands, under similar environmental conditions (Ellis & Ramankutty, ). Industrialized agriculture with high chemical inputs is a strong driver of species endangerment in North America (Kerr & Cihlar, ; Gibbs et al ., ). Several studies from Europe have also shown that high levels of pesticide and fertilizer use results in loss of species at a regional scale (Dormann et al ., ; Kleijn et al ., ). Thus, the strong endangering effect of residential croplands observed in our study may be a consequence of extensive use of large quantities of agricultural chemicals. However, other factors associated either with industrialized agriculture, for example, mechanization and the use of fossil fuels, or pressures more directly associated with a high rural human population, for example, hunting, trapping or poisoning of animals, may also play a role. Residential croplands constitute 70% of the world's cropland area, while populated and remote croplands constitute 26% and 4%, respectively (Ellis & Ramankutty, ). Thus, agricultural industrialization poses an especially big challenge for global biodiversity conservation. Our finding that the endangerment probability of mammals decreases with the proportion of rangelands is also unique. While Lenzen et al . ( ) found a slight positive correlation between species endangerment and pastureland globally, rangelands have been shown to support greater number of species, particularly native species, than urban lands at regional scales (Maestas et al ., , ). It has also been suggested that grazing by domestic animals can help maintain rare species (Zamora et al ., ; Sirami et al ., ) in the landscape which would otherwise become locally extinct if the land was not managed by humans. Thus, our findings regarding rangelands may be explained by the fact that rangelands provide better habitat for wildlife than most other anthropogenic land uses as well as former pasture or farm land that has been abandoned and converted into forest or other vegetation cover. Given that rangelands constitute approximately 30% of the earth's land area (Ellis & Ramankutty, ), they are likely to have an especially important role in maintaining biodiversity at a global scale. Loss of wild or undisturbed habitat may also be an important driver of species endangerment as, according to the OLR results, the endangerment probability of most mammal groups decreases as the proportion of wildlands and remote forests within their distribution ranges increases. Furthermore, total distribution range size of highly endangered species was much lower than the range size of less endangered species (Table ) suggesting that a smaller area of potential habitat makes species more prone to endangerment. Thus, a reduction in the intensity of current land uses, such as croplands, may not benefit species if it results in conversion of natural habitat to agriculture. Species endangerment has previously been shown to depend on area of natural habitat as well as species distribution range size (Czech et al ., ; Kerr & Deguise, ). However, our results suggest that not all land with natural vegetation cover provides good habitat for species as the endangerment probability of several mammal groups increases as the proportion of populated forests within their distribution ranges increases. Thus, species may not necessarily benefit from having more area with natural vegetation cover in their distribution range if these areas are over utilized or not managed properly. There has recently been great interest in identifying areas of exceptionally high biodiversity value and/or threat (Olson & Dinerstein, ; Myers et al ., ) and determining changes in land use patterns within them (Scharlemann et al ., ; Hoekstra et al ., ; Grau et al ., ). However, few attempts have been made to explore the links between socio‐economic drivers, human land use or management strategies, and biodiversity loss at regional or global scales. Over the next several decades, there will be a huge increase in the world's urban population (Seto et al ., ), mostly owing to urbanization in developing parts of the world such as south Asia, Africa and Latin America, where natural habitat loss is already highly extensive (Hoekstra et al ., ). This increase in urban population will put greater pressure on the world's species not only by causing conversion of natural land to human uses, but also through increases in the intensity of land use on existing anthropogenic lands. Consequently, combining knowledge regarding the socio‐economic drivers of land use change and the role of different land use types and intensities in endangering or maintaining species diversity may be the key to making successful land use policy/mitigation decisions for conserving the world's biodiversity. In summary, 75% of the earth's ice‐free terrestrial surface has been altered and is used by humans to some extent (Ellis & Ramankutty, ), and the area of wild habitat remaining in the world's biodiversity hotspots is especially small (Hoekstra et al ., ). Thus, determining the potential of human land uses in providing habitat for species is highly important for global biodiversity conservation (Novacek & Cleland, ). To this end, we found that some human land use types, for example, rangelands, are better habitat for a larger proportion of mammal species than other human land use types, for example, croplands or human settlements, and even some natural lands with high human activity, such as, populated forests. Furthermore, a low‐intensity land use, for example, remote croplands, provides better habitat for species than the same type of land use with a higher intensity, for example, residential and populated croplands. These results together demonstrate that some human settlement and food production strategies are better than others and may actually help maintain biodiversity at a global scale. Acknowledgements We thank J. Doucette and J. Plourde at the Department of Forestry and Natural Resources at Purdue University for their GIS help, and J. Datta and B. Craig at the Department of Statistics at Purdue University for their advice regarding the statistical methods used in this study.

Journal

Diversity and Distributions – Wiley

Published: Sep 1, 2012

There are no references for this article.