The Challenge of Wildlife Conservation from Its Biogeographical Distribution Perspectives, with Implications for Integrated Management in Peru ^†

Abstract

Biodiversity is an indispensable resource and contributes to the balance of ecosystems, being of great importance for the development of a society and its culture through good management of natural spaces. However, the reduction in and fragmentation of habitats, trafficking, and illegal trade in wild animals affect the great diversity of wild flora and fauna that characterize Peru. Considering this problem, we modeled the biogeographic distribution of five species of wildlife categorized as threatened by Peruvian legislation and included in the red list of threatened species of the International Union for the Conservation of Nature (IUCN): critically endangered (CR) Lagothrix flavicauda, endangered (EN) Aotus miconax, in vulnerable-status (VU) Tremarctos ornatus and Lagothrix cana, and in the near-threatened category (NT) Panthera onca. Our study aimed to identify their current potential distribution in the Peruvian territory which is legally protected by the conservation areas of national, regional, or private administration. In this regard, we used a maximum-entropy approach (MaxEnt), integrating 14 variables (7 bioclimatic variables, 3 topographic, 3 variables of vegetation cover, and relative humidity). It was observed that 3.6% (46,225.50 km²) of the Peruvian territory presented a high probability (>0.6) of distribution of the evaluated species and 10.7% (136,918.28 km²) of moderate distribution (0.4–0.6). Based on this, our study allowed us to identify the geographical spaces for threatened species on which conservation actions should focus, through the formulation of strategies, plans, policies, and participatory management in the Peruvian territory.

1. Introduction

Peru is one of the 17 megadiverse countries in the world [1,2,3], and to preserve this biodiversity in recent decades, national, regional, and private conservation areas have been created, with the aim of conserving species of high diversity and endemism, marine biodiversity, or particular groups of organisms [4,5,6]. However, over time, anthropic pressures, such as deforestation, installation of agricultural crops, extensive livestock, illegal mining, and forest fires, among others, have reduced the habitat of native and endemic species, which has led to the reduction in their populations [7]. Based on this, Peruvian legislation [8] and the International Union for Conservation of Nature (IUCN) red list of threatened species of wildlife have categorized the species in our study as described: critically endangered (CR) Lagothrix flavicauda, endangered (EN) Aotus miconax, vulnerable-status (VU) Tremarctos ornatus and Lagothrix cana, and near-threatened category (NT) Panthera once [9]. This situation describes the importance of knowing the territorial spaces in which to manage and implement plans for the survival of local populations, mitigating poaching and illegal trade [7,10].

The species considered in this study have been identified as those requiring urgent conservation measures by the International Union for Conservation of Nature, the Convention on International Trade in Endangered Species, and the International Primatological Society, in addition to being protected by the Peruvian legislation [11]. However, the National Service of Natural Areas Protected by the State reports only 22,645,810.51 hectares of protected natural areas for conservation, equivalent to only 17.62% of the Peruvian territory [6]. In addition, the nature protection offices in the country work with extremely small budgets [11]; therefore, it is crucial to develop international, strategic alliances for habitat conservation.

Therefore, species distribution models (SDM) are important tools in conservation approaches [12], allowing the identification of geographical spaces with similar topographic characteristics, bioclimatic features, and records of presence [13]. SDMs have been widely applied in the identification of potential wildlife distribution in large mammals [14] and flora species, the prediction of deforestation and forest fires [15,16,17,18], as well as the assessment of the impact of anthropogenic land-use change in protected areas [19]. The maximum-entropy algorithm (MaxEnt) is the one that presents reliable, optimal, and defensible results and surpasses other SDM algorithms [13,20,21,22]. In this study, using MaxEnt, we identified the biographical distribution under current conditions for an integrated management of the wild fauna of the species L. flavicauda, A. miconax, T. ornatus, L. cana, and P. onca in the Peruvian territory.

2. Study Area

This study is located between the parallels 0°03′00′’ and 18°30′00” south and the meridians 68°30′00” and 81°30′00′’ to the west, covering the Peruvian territory, in an area of approx. 1,300,000 km², with a rugged area consisting of geographical regions of coast, mountains, and jungle and altitudinal gradients from 0 m to 6800 m above sea level (m.a.s.l) (Figure 1).

3. Material and Methods

3.1. Datase

The presence data of the species were obtained from the register of CITES species by the Ministry of the Environment of Peru—MINAM [23], in combination with spatial information from the Global Biodiversity Information Service—GBIF [24] and “Species Explorer”, collected through the noncommercial software QGIS. The data were exported in a comma-delimited format (.csv) for integration into the maximum-entropy (MaxEnt) software ver. 3.4.1 [25]. To perform the spatial modeling of the species, initially, 28 variables were included (Table S1) and rescaled to a spatial resolution of 250 m. Likewise, in order to minimize the multiple multicollinearities of these variables, they were filtered using the Pearson correlation coefficient through the R 3.6 software (The R Foundation, Vienna, Austria) and r = ±0.8 was established as the cut-off value for the highly correlated variables [26,27,28]. Finally, the 14 variables (

Table 1. Bioclimatic, topographic, and vegetation cover variables used in modelling.

Variable		Units	Symbol
Bioclimatic	Annual mean temperature	°C	bio01
	Min. temperature of coldest month	°C	bio06
	Mean temperature of warmest quarter	°C	bio10
	Precipitation of driest month	mm	bio14
	Precipitation seasonality	mm	bio15
	Precipitation of wettest quarter	mm	bio16
	Precipitation of coldest quarter	mm	bio19
	Relative humidity	%	rhm
Topographic	Elevation above mean sea level	m.a.s.l	dem
	Slope of the terrain	°	slope
	Distance to hydrography	m	d_water
Vegetation cover	Ecosystem	Type	Ecosystem
	Tree height	m	Tree_h
	Land use and land cover	Type	LULC

) were chosen for the final modeling: 7 bioclimatic variables were included, in addition to relative humidity from WorlClim [29] and 3 topographic variables derived from the digital elevation model (DEM), available on the United States Geological Survey (USGS) portal [30]. The variables of vegetation cover of ecosystems were from the MINAM study [23,31], tree altitude [32] and land use/land cover (LULC) [33].

3.2. Methods

Figure 2 summarizes the methodological design of our research; based on the spatial standardization of cartographic variables and their trimming at the level of the Peruvian territory, the biogeographic modeling of the 05 species was carried out using the MaxEnt software [25,34]. We used 75% of the presence data for training and 25% for validation [34], using 5000 iterations and 10 replicas with random partitions (cross-validation method); other settings were maintained by default. The validation of the models was carried out according to the area under the curve (AUC) and differentiated performance in five levels: invalid (<0.6), bad (0.6–0.7), accepted (0.7–0.8), good (0.8–0.9), and excellent (>0.9) [34,35,36]; in the same way, the contribution of each of the variables in the model was obtained. Finally, the resulting raster of distribution was reclassified into four potential habitat ranges (“high”, >0.6; “moderate”, 0.4–0.6; “low”, 0.2–0.4; and “non-potential”, <0.2) [15,16,17,37] and converted to vector cartographic data to perform the surface calculation.

4. Result and Discussion

4.1. Results

It was observed that, from integrating the high potential individual distribution of the 05 species (L. flavicauda, A. miconax, T. ornatus, L. cana, and P. onca), 3.6% (46,225.50 km²) of the Peruvian territory presented a high probability of distribution (>0.6) and 10.7% (136,918.28 km²) a moderate distribution (0.4–0.6) (Figure 3f). Correspondingly, L. flavicauda had a high potential distribution of 3354.74 km² (Figure 3a) and A. miconax had a high distribution of 2,324.96 km² (Figure 3b). T. ornatus presented the largest area of high potential distribution in 23,179.96 km² (Figure 3c); finally, the high potential distribution of L. cana covers 5833.33 km² (Figure 3d) and 11,532.50 km² for P. onca, respectively (Figure 3e).

Modeling showed an average performance of AUC = 0.97, considered excellent (AUC > 0.9). Likewise, the bioclimatic variables with the greatest contribution to modeling were the precipitation of the driest month (Bio14) and relative humidity (rhm); in the same way, the variables such as altitude (DEM), type of ecosystem (ecosystem), and slope (slope) had a high contribution in the modeling of each species (Figure 4).

4.2. Discussion

SDMs are a statistical tool [38] widely used in studies of rare and endangered groups, as well as the environmental variables that affect them [39]. SDMs have contributed significantly to the challenge of wildlife conservation from a biogeographical approach [40,41,42,43,44,45]. Our study is the first to integrate the high potential individual distribution of 05 species of threatened wildlife—in critically endangered condition (L. flavicauda),” endangered (A. miconax), vulnerable (T. ornatus and L. cana) and near-threatened (P. onca)—within the Peruvian territory. The model required strong performance values of AUC = 0.97 [45,46]. Of the 14 integrated variables, the precipitation of the driest month (Bio14) and relative humidity (rhm) were the most representative to predict integrated areas for the five species, according to the potential habitat ranges; however, the topographic variables (altitude), the type of ecosystem (ecosystem), and the slope (slope), that also contributed significantly during the modeling, were not dismissed. Our study validates the restricted range of endemic species (L. miconax and L. Favicauda) [7]; it is necessary to indicate that it is possible to find P. onca in other territories of native communities in the Peruvian Amazon, in which recent studies are documenting and reporting it [47]. Thus, it is necessary to carry out subsequent studies to improve the performance of the model with a greater amount of presence data and other variables; this will allow identifying territories’ suitable conservation methods, avoiding the reduction in their population by hunting and habitat loss, as is happening with L. cana [7]. The different methods used to select the variables and, therefore, the different variables introduced in the models, contributed to differentiating their contribution [45,48].

Thus, from the identification of potential areas, it is possible to establish measures to mitigate the reduction and fragmentation of the habitats of these five species, in the Peruvian territory [45]. So, 46,225.50 km² of the territory is within the 33 geographical spaces suggested in this study, for the threatened species on which conservation actions should be focused [49], through the formulation of strategies, plans, policies, and participatory management in the Peruvian territory. New studies will allow the evaluation of the distribution in future conditions of climate change, in an integrated way, for these five species, as long as this is carried out considering the qualities of the species to adapt to new conditions of persistence and survival [44]. Finally, modeling will allow having a support for the adequate management of the territory to ensure the survival of the species.

5. Conclusions

Modeling through maximum entropy (MaxEnt) obtained a performance considered excellent, with an area under the curve (AUC) of 0.97. From this, under the current conditions, the biogeographic distribution of the 05 species (L. flavicauda, A. miconax, T. ornatus, L. cana, and P. onca) covers 3.6% (46,225.50 km²) of the Peruvian territory; this area presents a high probability of distribution. Added to this, 136,918.28 km² (10.7%) was identified with a moderate probability of distribution. Finally, the bioclimatic variables with the greatest contribution to modeling are the precipitation of the driest month (Bio14) and relative humidity (rhm), as well as the topographic variables (altitude), the type of ecosystem (ecosystem) and the slope (slope), which contributed significantly during the modeling.