CZAI
27,898 imgs · 450 TN

Biodiversity & soundscape

Standard ecology indices computed from camera detections plus BirdNET-style acoustic events. The composite health score is a transparent weighted blend (40% richness · 30% Shannon H' · 30% NDSI) intended as a single-glance number for donor and programme leadership; the underlying components and time-series are below.

UG-ZWR
Ziwa Wildlife Reserve (savanna)
savanna_grassland
71
stable
Species richness (S)
12
Shannon H'
2.37/ 2.89
Simpson 1−D
0.899
Pielou's J'
0.954
NDSI
0.32-1…1
Detections
8,137
health components
richness 26.7Shannon 24.6NDSI 19.8
soundscape · 387 events
biophony 232anthro 120geo 35
UG-BFE
Budongo Forest Edge
moist_semi_deciduous_forest
67.4
stable
Species richness (S)
11
Shannon H'
2.24/ 2.89
Simpson 1−D
0.879
Pielou's J'
0.936
NDSI
0.31-1…1
Detections
6,320
health components
richness 24.4Shannon 23.3NDSI 19.7
soundscape · 337 events
biophony 202anthro 106geo 29
UG-MFR
Murchison Falls Riparian Zone
riparian_wetland
71.8
stable
Species richness (S)
12
Shannon H'
2.42/ 2.89
Simpson 1−D
0.906
Pielou's J'
0.973
NDSI
0.33-1…1
Detections
6,872
health components
richness 26.7Shannon 25.1NDSI 20
soundscape · 331 events
biophony 196anthro 98geo 37

Shannon H' (daily)

NDSI (biophony − anthrophony)

Species richness (daily)

Method & caveats

  • Shannon H' = −Σ pᵢ ln(pᵢ). Higher = more even spread of detections across species.
  • Simpson 1−D = 1 − Σ pᵢ². Probability that two random captures are different species.
  • NDSI = (B − A) / (B + A) using BirdNET event counts as a proxy for biophony / anthrophony power. Real soundscape pipelines compute this from raw audio power spectra (Kasten et al. 2012).
  • ACI proxy: Σ |Bₜ − Bₜ₊₁| / Σ (Bₜ + Bₜ₊₁) over hourly biophony bins per day. Pieretti et al. 2011 use the same formulation on FFT bins of the spectrogram itself; in production this needs raw audio.
  • Health score is a programme-defined composite — swap the 40/30/30 weights for whatever your science team agrees on and rerun the aggregator.
  • Richness scales against the 18-class Ugandan-mammal pool used in this synthetic dataset; production should scale against the real species checklist for the bioregion.