**4.3 Accounting for positional uncertainty and spatial bias**

Data elicited from local villagers impose two challenges, namely positional uncertainty and spatial bias. First, local villagers often recall *R. bieti* sightings in the form of "I saw the monkeys over this area." Clearly, "over this area" can be depicted using a polygon, but this does not mean that the monkeys showed up at every location in the polygon area and certainly not at an equal probability within the polygon. Thus, taking all locations in the polygon as sightings is not appropriate. The question then is how to obtain locations that are representative of the actual presence of wildlife in the area outlined by the local villagers. The second challenge is the spatial bias in the elicited *R. bieti* sightings due to local villagers' opportunistic observation effort. For example, multiple sightings of monkeys at one location by many villagers do not necessarily mean that the location is highly preferred by the monkeys; it might be that the location is easily visible from multiple activity routes. Thus, every time a monkey shows up at this location, it is spotted by some villager(s). On the other hand, a monkey that shows up at locations that are preferred by monkeys but less visible to the villagers will have a lesser chance of being spotted. This spatial bias must be compensated for when using the elicited *R. bieti* sightings for wildlife habitat assessment.

Geospatial analysis methods provisioned by GIS were adopted to address the two challenges. First, a frequency sampling strategy [23, 70] was applied to reduce the

#### **Figure 3.**

*Sightings of R. bieti and activity routes elicited from the local residents through geovisualization interviews: (a) 1973–1981 period; (b) 1987–2005 period; (c) 2006–2010 period (extracted from [40] with permission from John Wiley and Sons).*

position uncertainty in sighting polygons provided by local villagers and to identify the representative locations for *R. bieti* presence within each polygon. It is assumed that locations at which values of environmental conditions are most frequent over the polygon area would approximate the locations of actual presence best. Under this assumption, the frequency sampling strategy implemented in GIS was applied to locate the representative locations within a polygon. Here, only the general idea was outlined as above; full details on implementing the sampling strategy in GIS can be found in [23, 70].

Second, the spatial bias was compensated for by inversely weighting each representative presence location with cumulative visibility of the location from the routes taken by local villagers [23]. In this particular case study, spatial bias in the elicited *R. bieti* sightings was a result of the non-random and uneven distribution of local villagers' observation efforts constrained by activity routes and terrain (as discussed in depth in Section 2.3.3). Thus, cumulative visibility was treated as a proxy of the underlying observation effort and can be incorporated to compensate for spatial bias. The cumulative visibility of a location can be computed in GIS based on a DEM and the activity routes of local villagers [23]. The efficacy of the frequency sampling strategy to reduce positional uncertainty and the visibility-weighting scheme to compensate for spatial has been demonstrated in [23].

#### **4.4 Habitat assessment**

A kernel density estimation-based method [23, 71] was applied to derive the relationship between *R. bieti* habitat suitability and environmental conditions. This method estimates a probability density function representing the probability distribution of wildlife presence over the gradient of each environmental factor based on the values of the environmental factors over the presence locations. In estimating the probability density functions, presence locations are weighted by the in situ cumulative visibility from activity routes of the local villagers to compensate for spatial bias. The probability density functions are then normalized to the range of [0, 1] to represent the relationships between habitat suitability and individual environmental factors (**Figure 4**). The overall habitat suitability considering all

#### **Figure 4.**

*Suitability-environment relationships derived from elicited R. bieti sightings in each historical period. Aspect group 1: 0–45° (starting from north), 2: 45–90°, 3: 90–135°, 4: 135–180°, 5: 180–225°, 6: 225–270°, 7: 270–315°, 8: 315–360°. Vegetation type 1: evergreen coniferous, 2: pasture, 3: yunnan pine, 4: farmland, 5: deciduous broadleaf (extracted from [40] with permission from John Wiley and Sons).*

**31**

**5. Conclusions**

**Figure 5.**

*John Wiley and Sons).*

*Integrating Citizen Science and GIS for Wildlife Habitat Assessment*

environmental factors is determined by integrating the relationships based on a "limiting factor" principle (see [23] for full details of the method). Computing the overall habitat suitability at every location (pixel) in the study area resulted in

*Habitat suitability maps predicted for the study area using elicited R. bieti sightings in each historical period (a) 1973–1981 period; (b) 1987–2005 period; (c) 2006–2010 period (extracted from [40] with permission from* 

Across the three historical periods, high suitability habitats were in forests (**Figure 4d**) at mid-to-high elevation range (**Figure 4a**) on the northeast hill slopes (**Figure 4c**). Overall, high suitability habitats shrank in the 1987–2005 period compared to the previous period. As an example, the area outlined on **Figure 5b** in the 1987–2005 period is of much lower suitability compared to the 1973–1981 period. This might be a result of the introduction of new village settlements and roads in that area in the 1987–2005 period which induced significant human disturbance. *R. bieti* habitats were recovering in the 2005–2010 period. The outlined area recovered to higher suitability in that period; this might be attributed to the monkeys

The derived relationships between *R. bieti* habitat suitability and individual environmental factors (**Figure 4**) confirmed the recovering trend in the 2006–2010 period. The elevation range of high suitability habitats in the 2006–2010 period shifted back to higher ranges close to those in the 1973–1981 period (**Figure 4a**). The ranges of distance to rivers and distance to village or road corresponding to high suitability habitats also shifted back to similar ranges as in the 1973–1981 period (**Figure 4e, f**). These were potential evidences that conservation practices initiated by the Yunling Nature Reserve have restored *R. bieti* habitat in the area.

Wildlife data required for wildlife habitat assessment can be difficult and expensive to obtain with traditional data collection methods (e.g., biological survey, geographic sampling), especially for conservation programs with limited budget support in remote and poor areas. Citizen science offers a cost-effective way of collecting wildlife data to sustain such programs. Nevertheless, average citizens are non-

professionals and their wildlife observation efforts are un-coordinated. Thus, wildlife

*DOI: http://dx.doi.org/10.5772/intechopen.83681*

habitat suitability maps as shown in **Figure 5**.

getting used to proximity to villages and roads (**Figure 4f**).

*Integrating Citizen Science and GIS for Wildlife Habitat Assessment DOI: http://dx.doi.org/10.5772/intechopen.83681*

#### **Figure 5.**

*Wildlife Population Monitoring*

can be found in [23, 70].

**4.4 Habitat assessment**

position uncertainty in sighting polygons provided by local villagers and to identify the representative locations for *R. bieti* presence within each polygon. It is assumed that locations at which values of environmental conditions are most frequent over the polygon area would approximate the locations of actual presence best. Under this assumption, the frequency sampling strategy implemented in GIS was applied to locate the representative locations within a polygon. Here, only the general idea was outlined as above; full details on implementing the sampling strategy in GIS

Second, the spatial bias was compensated for by inversely weighting each representative presence location with cumulative visibility of the location from the routes taken by local villagers [23]. In this particular case study, spatial bias in the elicited *R. bieti* sightings was a result of the non-random and uneven distribution of local villagers' observation efforts constrained by activity routes and terrain (as discussed in depth in Section 2.3.3). Thus, cumulative visibility was treated as a proxy of the underlying observation effort and can be incorporated to compensate for spatial bias. The cumulative visibility of a location can be computed in GIS based on a DEM and the activity routes of local villagers [23]. The efficacy of the frequency sampling strategy to reduce positional uncertainty and the visibility-weighting

A kernel density estimation-based method [23, 71] was applied to derive the relationship between *R. bieti* habitat suitability and environmental conditions. This method estimates a probability density function representing the probability distribution of wildlife presence over the gradient of each environmental factor based on the values of the environmental factors over the presence locations. In estimating the probability density functions, presence locations are weighted by the in situ cumulative visibility from activity routes of the local villagers to compensate for spatial bias. The probability density functions are then normalized to the range of [0, 1] to represent the relationships between habitat suitability and individual environmental factors (**Figure 4**). The overall habitat suitability considering all

*Suitability-environment relationships derived from elicited R. bieti sightings in each historical period. Aspect group 1: 0–45° (starting from north), 2: 45–90°, 3: 90–135°, 4: 135–180°, 5: 180–225°, 6: 225–270°, 7: 270–315°, 8: 315–360°. Vegetation type 1: evergreen coniferous, 2: pasture, 3: yunnan pine, 4: farmland, 5: deciduous* 

*broadleaf (extracted from [40] with permission from John Wiley and Sons).*

scheme to compensate for spatial has been demonstrated in [23].

**30**

**Figure 4.**

*Habitat suitability maps predicted for the study area using elicited R. bieti sightings in each historical period (a) 1973–1981 period; (b) 1987–2005 period; (c) 2006–2010 period (extracted from [40] with permission from John Wiley and Sons).*

environmental factors is determined by integrating the relationships based on a "limiting factor" principle (see [23] for full details of the method). Computing the overall habitat suitability at every location (pixel) in the study area resulted in habitat suitability maps as shown in **Figure 5**.

Across the three historical periods, high suitability habitats were in forests (**Figure 4d**) at mid-to-high elevation range (**Figure 4a**) on the northeast hill slopes (**Figure 4c**). Overall, high suitability habitats shrank in the 1987–2005 period compared to the previous period. As an example, the area outlined on **Figure 5b** in the 1987–2005 period is of much lower suitability compared to the 1973–1981 period. This might be a result of the introduction of new village settlements and roads in that area in the 1987–2005 period which induced significant human disturbance. *R. bieti* habitats were recovering in the 2005–2010 period. The outlined area recovered to higher suitability in that period; this might be attributed to the monkeys getting used to proximity to villages and roads (**Figure 4f**).

The derived relationships between *R. bieti* habitat suitability and individual environmental factors (**Figure 4**) confirmed the recovering trend in the 2006–2010 period. The elevation range of high suitability habitats in the 2006–2010 period shifted back to higher ranges close to those in the 1973–1981 period (**Figure 4a**). The ranges of distance to rivers and distance to village or road corresponding to high suitability habitats also shifted back to similar ranges as in the 1973–1981 period (**Figure 4e, f**). These were potential evidences that conservation practices initiated by the Yunling Nature Reserve have restored *R. bieti* habitat in the area.

#### **5. Conclusions**

Wildlife data required for wildlife habitat assessment can be difficult and expensive to obtain with traditional data collection methods (e.g., biological survey, geographic sampling), especially for conservation programs with limited budget support in remote and poor areas. Citizen science offers a cost-effective way of collecting wildlife data to sustain such programs. Nevertheless, average citizens are nonprofessionals and their wildlife observation efforts are un-coordinated. Thus, wildlife data contributed by citizens may be subject to data quality issues such as positional uncertainty and spatial bias. This chapter provides an overview of citizen science as a means of collecting wildlife data, GIS-provisioned geovisualization, and geospatial analysis techniques for tackling the data quality issues of citizen-contributed wildlife data, and the integration of citizen science and GIS for wildlife habitat assessment. A case study of mapping *R. bieti* habitat suitability using *R. bieti* sightings elicited from local villagers in Yunnan, China, was presented as an example to demonstrate the usefulness of integrating citizen science and GIS for wildlife habitat assessment.
