10.5061/DRYAD.MB7QK
Steger, Cara
Colorado State University
Butt, Bilal
University of Michigan-Ann Arbor
Hooten, Mevin B.
Colorado State University
Data from: Safari science: assessing the reliability of citizen science
data for wildlife surveys
Dryad
dataset
2018
hierarchical modeling
observer bias
Bayesian data reconciliation
validation
2018-04-04T00:00:00Z
2018-04-04T00:00:00Z
en
https://doi.org/10.1111/1365-2664.12921
406199 bytes
1
CC0 1.0 Universal (CC0 1.0) Public Domain Dedication
1. Protected areas are the cornerstone of global conservation, yet
financial support for basic monitoring infrastructure is lacking in 60% of
them. Citizen science holds potential to address these shortcomings in
wildlife monitoring, particularly for resource-limited conservation
initiatives in developing countries - if we can account for the
reliability of data produced by volunteer citizen scientists (VCS) . 2.
This study tests the reliability of VCS data vs. data produced by trained
ecologists, presenting a hierarchical framework for integrating diverse
datasets to assess extra variability from VCS data. 3. Our results show
that, while VCS data are likely to be overdispersed for our system, the
overdispersion varies widely by species. We contend that citizen science
methods, within the context of East African drylands, may be more
appropriate for species with large body sizes, which are relatively rare,
or those that form small herds. VCS perceptions of the charisma of a
species may also influence their enthusiasm for recording it. 4. Tailored
program design (such as incentives for VCS) may mitigate the biases in
citizen science data and improve overall participation. However, the cost
of designing and implementing high quality citizen science programs may be
prohibitive for the small protected areas that would most benefit from
these approaches. 5. Synthesis and applications. As citizen science
methods continue to gain momentum, it is critical that managers remain
cautious in their implementation of these programs while working to ensure
methods match data purpose. Context-specific tests of citizen science data
quality can improve program implementation, and separate data models
should be used when VCS variability differs from trained ecologists’ data.
Partnerships across protected areas and between protected areas and other
conservation institutions could defray the costs of citizen science
program design and implementation.03-Apr-2017
data_JAPPL201601169Species counts were collected using ecological
transects (Traditional Sampling, TS =1) and citizen scientists on game
drives (Safari Science, SS =1) from June - August 2013. Dist_Livestock was
created using points from livestock locations collected during the TS
fieldwork, and distance to those points was calculated in ArcGIS. TWI was
calculated in ArcGIS using slope and elevation from a 30m DEM. Dist_river
was calculated using stream data produced from the same DEM. The wildlife
location data were separated into groups ("Groups" column)
spanning 16 days, which were centered on each of the Landsat scenes, and
the NDVI values were extracted to those locations.
East Africa