Snapshot Safari: A standardized

Transcription

1 Snapshot Safari: A standardized program for assessing population and behavioral dynamics of large mammals Craig Packer, Meredith Palmer & Sarah Huebner Department EEB, University of Minnesota & School of Life Sciences, University of KwaZulu Natal

2 Serengeti National Park : 225-camera grid covering 1100 km 2 Snapshot Serengeti NSF National Science Foundation NATIONAL GEOGRAPHIC

3 Margaret Kosmala Ali Swanson Meredith Palmer Annika Moe Online volunteers view photos, classify & count each animal

4 Since launch in December 2012: 180,000 volunteers 12 million classifications Each photo viewed ~10X Consensus 97.1% accurate Swanson, et al., Nature Scientific Data 2015

5 Spotted hyena WTF??

6 Species classifications are least accurate in rarest species Swanson, et al Conservation Biology False negatives: species present but not reported False positives: species reported but not present

7 Swanson, et al Conservation Biology Accuracy vs Effort

8 The animals have been classified and counted now what?

9 Migratory species vs. Residents Wildebeest Zebra T. gazelle Buffalo Hartebeest

10 lion hyenaspotted cheetah lion hyenaspotted cheetah Carnivore coexistence count tsince (Hours) Swanson, et al Ecology & Evolution

11 Temporal Variation in Predation Risk: Lunar cycles

12 Distribution Herding Vigilance

13 TIME SPACE % Lunar Surface Lunar Position Resources Encounter Risk Kill Risk

14 (Palmer et al. 2017, Ecol. Lett.)

15 Responses to lunar cycle interacted with spatially-structured risk Some anti-predator strategies based on predictable moon phase Others reflect current available light levels (Palmer et al. 2017, Ecol. Lett.)

16 Anti-predator strategies are species-specific Buffalo (largest, least vulnerable) responded least to temporal risk cues Gazelle and zebra made anti-predator decisions based on current light levels and lunar phase (Palmer et al. 2017, Ecol. Lett.)

17 Anti-predator strategies are species-specific Buffalo (largest, least vulnerable) responded least to temporal risk cues Gazelle and zebra made anti-predator decisions based on current light levels and lunar phase Wildebeest responded to lunar phase alone (Palmer et al. 2017, Ecol. Lett.)

18 Prey respond to specific risk cues by altering suites of behavior Respond strongly to areas of high lion encounter risk Distribution Avoided areas of high spatial risk Increased use of risky areas during safer times (Palmer et al. 2017, Ecol. Lett.)

19 Prey respond to specific risk cues by altering suites of behavior Use risky spaces at risky times with compensatory behaviors Vigilance Herding Form herds as nights become darker and in high risk areas More vigilant pre-moonrise and on darker nights (Palmer et al. 2017, Ecol. Lett.)

20 Daily cycles? Next logical step diel cycles Temporal Variation in Risk: Diel cycles

21 Lion activity across daily cycles Less active during day More active at night

22 (Palmer et al. in preparation.) For some species, responses to diel cycle interacted with spatially-structured risk Most sensitive to changes in likelihood of encountering lions Temporal refuge in the mid-afternoon

23 Medium-sized prey species: Attuned to spatiotemporal risk (Palmer et al. in preparation.)

24 (Palmer et al. in preparation.) Largest prey species: Low response to spatiotemporal risk Smallest prey species: Mixed responses

25

26 Proactive Risk Mitigation: Take-home points Dynamic changes in suites of anti-predator decision-making across spatiotemporal risk during lunar cycle Changes in spatial distribution across spatiotemporal risk during diel cycle. Behavioral changes still being analyzed.

27 Snapshot Safari Use camera-traps to monitor wildlife trends in different sites in Africa What works to conserve wildlife? Assemble Toolkit of Best Practices

28 Relative Abundance Indices (RAI s) reflect recent population estimates in the Serengeti Palmer et al African Journal of Ecology (in press) Wildebeest T. Gazelle Impala Zebra Eland Waterbuck Buffalo Giraffe Warthog Elephant

29 Random-Encounter Model correctly estimates Serengeti lion abundance but only at night Cusack, et al J. Wildife Management Day Night Day Night. Day Night Day Night Plains Woodlands Plains Woodlands Dry Season Wet Season

30 How many cameras are enough? Palmer et al African Journal of Ecology (in press)

31 Migratory Resident Migratory Resident 25 cameras 71 cameras 49 cameras 99 cameras

32 Larger camera grids provide greater precision for RAI s But for many species, maximum performance is achieved with ~50 cameras Migratory Herbivores Resident Herbivores Palmer et al African Journal of Ecology (in press)

33 Snapshot Safari Use camera-traps to monitor wildlife trends in different sites in Africa What works? Assemble Toolkit of Best Practices Launched: February 9 th Six sites in Mozambique, South Africa & Tanzania

34 Chalbi Akagera Grumeti Serengeti Ruaha Majete Niassa Hwange Mghadighadi Gorongosa Feb 2018 six sites Apr-Oct eighteen Selati Kruger Karingani Pilanesberg Mountain Zebra Karoo Gondwana Somkhanda

35 Snapshot Safari Use camera-traps to monitor wildlife trends in different sites in Africa What works? Assemble Toolkit of Best Practices Launched: February 9 th Six sites in Mozambique, South Africa & Tanzania 803,497 classifications in the first 25 days However, 1.2 million in first 10 days after initial launch of SnapshotSerengeti in camera-trap projects on Zooniverse in 2018 Next phase: machine learning with Jeff Clune, University of Wyoming

36

37

38

39 Animal count Species identification Accuracy of Machine-Learning Algorithm Top 1 = Correct species was the top choice Top 5 = Correct species in the top 5 choices Top 1 = Machine count was correct Top 5 = Machine count was within +/- 1 bin of correct count Norouzzadeh, et al. in press. PNAS

40 Snapshot Safari Use camera-traps to monitor wildlife trends in different sites in Africa What works? Assemble Toolkit of Best Practices Launched: February 9 th! South African component: FBIB National Biodiversity Project Jan Venter (NMMU), Rob Slotow (UKZN), Mike Somers (UP), Dan Parker (UMp), Mike Peel (ARC)

41

42

43 UMN Lion Lab team: 2 PhD students 23 Undergraduates