Qinghai Province is situated in the northeast of the Qinghai-Tibet plateau, which is the “water tower” of China and Asia (Huang 2013). Its total area is 7.2 × 10⁴ km², one thirteenth of China’s total area. It comprises the headwaters of several major Asian rivers, including the Yellow, Yangtze, Mekong, Salween and Yarlung Tsangpo (Brahmaputra) rivers, and thus contributes significantly to the livelihood and wellbeing of nearly 40 percent of the world’s population. Therefore, it is important to conserve this region for the livelihoods of all those people. The elevation in the province ranges from 1664 m to 6619 m (Fig. 1). From extensive alpine grasslands and wetlands to forests and deserts, Qinghai is home to a wide variety of globally significant, but fragile ecosystems. As a traditionally sparsely inhabited region with a variety of different climatic zones and natural habitats, Qinghai Province provides important habitats for many endangered species including the Tibetan antelope, wild yak, argali, snow leopard, black necked crane, saker falcon and many other key endangered wild animals.

Figure 1.

The location of Qinghai Province in China and the elevation range.

Efficient expansion of CAs requires simultaneous planning for species and ecosystems (Polak et al. 2015). Qinghai Forestry Department put forward a list of 79 rare and endangered species in 2013 as indicator species of biodiversity conservation in Qinghai Province. We thus used 11 endemic ecosystem types (Table 1) and 72 of the 79 endangered species (Table 2) as the surrogate of biodiversity in this region. In this study, we integrated conservation features from three sources to achieve maximum representation of biodiversity and compensate for limitations in data availability: (1) China key rare and endangered species database collected by The Nature Conservancy’s China biodiversity blueprint project. This database has been successfully used to predict climate change induced range shifts of Galliformes in China (Li et al. 2010). It was once employed to identify conservation priority areas in “China national biodiversity conservation strategy and action plan (2011–2030)” (Ministry of Environmental Protection 2010); (2) Chinese Endangered Species Information System (CESIS) (Xie et al. 1997). This system collected the latest endangered species information including mammals, birds, reptiles, amphibians, fish species or subspecies in China. Both the theoretical and practical simulations show that when the number of species presence points is greater than 14, the species distribution model can produce a better simulation result of species habitat (Proosdij et al. 2015). Therefore, we excluded these species with less than 15 presence points from the two databases, and obtained species presence data for 59 key rare and endangered species (Table 2). We checked the independence of the records and used them to input species distribution models to simulate their geographic ranges; (3) We identified the other 13 species’ suitability range using expert range maps from the online IUCN website (Table 2).

We applied a maximum entropy modelling technique with the MAXENT software (Phillips et al. 2006) to predict the graphic distributions of the 59 endangered species. This approach has been extensively adopted to project species range shifts and change in species diversity patterns and to inform conservation planning (Hernandez et al. 2008; Costa et al. 2010; Ponce-Reyes et al. 2012; McPherson 2014). A set of 19 bioclimatic variables at 30s resolution were collected from the WorldClim dataset for current conditions (average for 1951–2000) (Hijmans et al. 2005). We performed a principal components analysis of 19 bioclimatic variables to select the first three principal components as input climatic variables for our SDMs. We also included vegetation types and two human disturbance factors (population density and gross domestic product) into model input layers.

MAXENT was run in default settings with a maximum of 500 iterations. We used cross-validation procedures to model calibration, which randomly assigned 75% of species records while keeping the other 25% records for AUC computations. We assessed model performance with AUC, which provides a single measure of model performance and ranges from 0.5 (randomness) to 1 (perfect discrimination), where a score higher than 0.7 is considered a good model performance (Rebelo et al. 2010). Outputs from MAXENT models were reclassified to presence/absence predictions using the ‘‘Maximum Training Sensitivity Plus Specificity’’ threshold, which has proven to generally produce more accurate results than other thresholds (Fajardo et al. 2014; Liu et al. 2005).

We defined conservation targets for each species according to the current conservation status, spatial distribution range and endemic status (Fajardo et al. 2014). The target for each species was calculated as the sum of the following three indices: conservation status index, distribution size index, and conservation endemic index.

Distribution size index: Species with smaller distribution area should have a higher conservation priority and target, whereas species with larger distribution area should have lower a conservation target (Rodrigues et al. 2004). We assigned a more demanding representation target to species with more restricted ranges, acknowledging the negative relationship between species distribution size and extinction risk (Gaston and Rodrigues 2003). The value given to each species was scaled between a minimum coverage of 5% for species with a distribution equal to or greater than 300,000 km² in Qinghai Province, and a maximum of 25% for species with ranges equal to or less than 1,000 km² (Rodrigues et al. 2004). The 300,000 km² upper threshold corresponds to the range size observed in one third of the studied species in Qinghai Province.

Conservation status index: Like in Fajardo et al. (2014), we assigned goals to species identified as threatened by the IUCN following a decreasing scale: Critically Endangered (CR), 25%; Endangered (EN), 17.5%; Vulnerable (VU), 10%; Near Threatened (NT), 5%; Least Concern (LC), Not Evaluated (NE), and Data Deficient (DD), 0%.

Conservation endemic index: An endemic species is one whose habitat is restricted to a particular area, and can be easily under threat. As such, endemic species are of great conservation interest to conservation planning. We assigned goals of 10% for species endemic to Qinghai-Tibet Plateau, 5% for endemic species in China, and 0 for other species.

In Qinghai Province, wetland, forest and endemic grassland ecosystems have high conservation importance. Existing NRs already protect 70% of the important plateau wetland ecosystem (Liu and Li 2007). Therefore, we exclusively focused on endemic grassland and unique forest ecosystems. We used vegetation map of China (1:1 000 000) to represent ecosystem features of this region, and selected 11 endemic vegetation types as key conservation ecosystem types according to their endemism in this region or China (Qu 2011). We identified their conservation target as 10% for ecosystems endemic to the Qinghai-Tibet Plateau, 5% for endemic ecosystems in China, and 0 for other ecosystems. Although the conservation targets were determined arbitrarily, the results from our scenarios indicated that the changed conservation targets for each conservation feature did not radically affect the spatial distribution of the proposed CAs.

We performed a gap analysis that compared the defined conservation targets to species’ current representation within existing NRs. The species distributions and expert range maps were first intersected with the NRs, and then the percentage of its distribution within NRs was calculated and compared with its defined conservation targets. Species are considered insufficiently protected by the current NRs when the percentage is below their conservation targets.

We used the systematic conservation planning software MARXAN 2.4.3 (Ball et al. 2009) to identify the most efficient set of conservation priority areas to meet the above targets for both ecosystems and endangered species. It is a decision-support tool, which solves an optimization problem of representing a set of conservation features (species, ecosystems, ecoregions or ecosystem services) at a minimal cost, and has been widely used for identifying CAs in China (Wu et al. 2014; Zhang et al. 2014) and across the world (Powers et al. 2013; Hermoso et al. 2013; Tulloch et al. 2016; Powers et al. 2016). The Qinghai Province was partitioned into 4 km × 4 km grids or 44,475 planning units (PU). We unlocked Kekexili National NR and the Soka River Protection zone of Sanjiangyuan National NR, the two largest NRs of current network, and set PUs in them as “available”, because we assumed that the large extent of these two NRs may not be required to effectively meet conservation targets. PUs coinciding with other current NRs were prioritized in the MARXAN solutions. We set the cost of each PU as the value of the human footprint index (Sanderson et al. 2002). This index assumed that PUs with less human disturbance have higher social acceptance (Powers et al. 2013) and a lower conservation cost, and is widely accepted as a universal conservation cost surrogate (Fajardo et al. 2014; Wu et al. 2014). We ran different scenarios using the Zonae Cogito Decision Support System to test the most suitable parameters for MARXAN whereby we varied the boundary length modifier (BLM) and the species penalty factor (SPF). BLM and SPF were optimized to 100 and 1 respectfully since it offered an efficient tradeoff in our scenario analysis between cost, reserve compactness and achieving conservation targets. We ran MARXAN to identify 100 solutions using the simulated annealing algorithm and the default values for number of iterations (1,000,000) and temperature Decreases (10,000).

The best solution from the MARXAN output is the network most optimized with respect to achieving the conservation targets at the lowest cost. We thus proposed priority areas from MARXAN’s best solution. Given the financial challenges associated with the immediate implementation of these areas proposed in the best solution, we prioritized the areas outside of existing NRs according to three important decision making criteria: species richness, selection frequency, and vulnerability. Species richness was generated by calculating the number of studied species present in each 4 km×4 km grid cell across the entire study region based on the binary distribution maps from species distribution models and the range maps. It has long been recognized as a key characteristic determining biodiversity patterns and conservation selection. The grid cells with higher richness were assumed to have higher conservation value and were preferentially prioritized. MARXAN produced 100 solutions and a summed solution made up of the selection frequency across the 100 runs. This score of selection frequency represents the total section frequency of each grid. The vulnerability criteria is used to prioritize highly impacted areas that are in greater need of protection. We should give priority to protecting areas where human disturbance is more serious and ecologically more sensitive. To calculate the score, we used the human footprint index as a measure of the human influence on each PU.

The three criteria scores were normalized to values between 0 and 100, and summed to give each proposed CA an overall priority score. Priority areas were classified as high, medium, and low priority according to the overall priority score. The area of high, medium, and low priority was determined using natural break method (Fajardo et al. 2014).

The species distribution models were able to accurately predict the geographic distributions of the species. Specially, the models had AUC values between 0.843 and 0.999, which indicates that the generated geographic distributions can be used to estimate regional species richness patterns and conservation planning (Fig. 4a). Species richness was spatially heterogeneous and follows the well-known latitudinal pattern in Qinghai Province. Its spatial pattern shows a general reduction from the eastern to western areas. The maximum value of 57 species per km² is located in the Haidong and Xining regions. Regions with a relatively low number of species are situated in the western high altitude areas, including Haixi and Yusu (Fig. 4a).

The 11 NRs account for 30.2% of the total ​Qinghai Province area. The percentage of area with 10–20 and 30–40 species/km² protected by the current NRs was 37% and 35%. The two regions with the highest species richness encompassed an area of 110000 km² and 3000 km² respectively, and had a low protection level of 19% and 11% (Fig. 1). These two regions are mainly located in the farming-pastoral ecotone within the eastern and southern parts of Qinghai Province.

We found that 41 species, 53% of the total, are insufficiently protected in the current reserve system according to our defined conservation target for each species. We also found that targets for those species most at risk species are not well met under current NRs: 3 out of 4 critically endangered, on third of the endangered, and 8 out of 16 vulnerable species did not achieve their defined conservation goals (Fig. 2). There were 22 and 11 species whose protection under existing NRs exceeded conservation targets by 10% and 20% respectively (Fig. 2).

Figure 2.

Summary of the conservation gap for key rare and endangered species in Qinghai Province: (a) species number of conservation goals met and not met; (b) the area protected and unprotected in nature reserves (CR – Critically Endangered, EN – Endangered, VU – Vulnerable, NT – Near Threatened, LC - Least Concern).

A set of priority areas based on the best solution were selected in Qinghai Province (Fig. 3). We identified 57 optimal CAs for biodiversity conservation in Qinghai Province. The total area assigned as CAs in order to achieve the conservation targets for all conservation features is about 273,872 km², about 39.3 % of the total land area in Qinghai Province. Among these selected priority areas, 28 areas are located within the existing NRs. The total area of selected CAs inside current reserves is about 134,656km², 19.3% of Qinghai Province. In order to better guide conservation investment and management, we judiciously reduced the coverage of the Sanjiangyuan National NR (conservation zone A, B, C and D in Fig. 3). This very large protected area was not optimized; therefore, its conservation effectiveness (e.g., reduced conservation cost, greater transparency and objectiveness, and higher level of protection for more species) can be improved by systematic conservation planning.

Figure 3.

Spatial distribution of proposed priority areas (including high, medium and low priorities) inside and outside the existing nature reserves for Qinghai Province.

Figure 4.

Spatial distribution maps for the three criteria used to evaluate the conservation priority of the proposed conservation areas in this study: a species richness and current nature reserves across Qinghai Province b vulnerability, derived from the Human Footprint index c selection frequency of the planning units, including additional solutions with varying conservation goals; and d overall priority score.

To fully meet our criteria for our conservation features, 29 new or not previously conserved areas, approximately 139,216 km² (20% of Qinghai Province), were added to the current NR system (Fig. 3). The majority of these new conservation priority areas were located in the Qinghai Province’s east and, to a lesser degree, in parts of the central and southern regions. Some conservation priorities were selected to improve the connectivity among other conservation areas, which were located between Qinghai Qaidam Haloxylon ammodendron forest national NR and Qinghai Nuomuhong Provincial NR (13, 17, and 20 in Fig. 3), Protection Zones of Sanjiangyuan National NR (7 and14 in Fig. 3), and Protection Zones of Qilian Mountain National NR (12 and 16 in Fig. 3).

The prioritization of new selected areas outside the existing NRs was determined according to an overall priority score derived from three design criteria: species richness, selection frequency, and vulnerability. The vulnerability of the proposed priority areas, as measured by the vulnerability score, increases gradually from east to west. Of the 29 new areas, 10 were designated as high priority, 11 as medium priority, and 8 as low priority (Fig. 3). High priority areas are more abundant in the eastern and southeastern parts of this province. In general, five additional priority areas were larger than 1000 km², while 11 are larger than 5,000 km² in size. For the top five largest priority areas, two are in the northeastern region, while three are in the south of Qinghai Province. The largest one (24 in Fig. 3) was the attached Haixi in the southwest of Qinghai (11135 km²). The second largest (11 in Fig. 3) was in the central part of Guoluo (106906 km²), the third (19 in Fig. 3) was in the east of Yusu (9989 km²), while the fourth (3 in Fig. 3) and the fifth (16 in Fig. 3) are located in between Haibei and Xining (9802 km²) and between Haibei and Haixi (9437 km²), respectively.