Dataset
This dataset aimed to provide wildlife observations with a complete set of spatiotemporal metadata, to support the development of a novel cascading ensemble classification method. Each section below details the relevant essential information regarding the components of the dataset, and the final dataset taxonomic count breakdown for observations and images. Note the distinction between observations and images as observations may contain erroneous or mislabelled images.
The dataset is published and publicly available at the following link: https://www.kaggle.com/datasets/travisdaws/spatiotemporal-wildlife-dataset
The below figure displays the geographic distribution of the published dataset:
Observations
The observations are sourced from iNaturalist, a citizen-science platform. The images originate from a combination of camera-traps and citizen-science. Due to the time and resource limitations of the study, a subset of the available data is utilized, to form a generic subset, representative of the issues encountered within wildlife classification. The Felidae and Elephantidae mammalian taxonomic families were selected to populate the subset. This is due to the global distribution of the families stretching to remote corners of the world and the unbalanced number of sightings.
The resulting dataset spans 2 taxonomic families, 16 taxonomic genera, 48 taxonomic species, and 67 taxonomic subspecies. Note, the common household cat Felis catus was required to be removed from the original dataset, in order to maintain a wildlife only dataset. Additionally, iNaturalist offers the download capability to include the taxonomic labels (domain, kingdom, phylum, class, order, family, genus, species, and subspecies). However, the subspecies classification capability became apparent late in the study, hence subspecies labels were extracted from observations specifying the subspecies in the place of the common name. Hence, a greater quantity of subspecies labels may be available from iNaturalist than within the dataset.
Image Processing
The raw observation images contained erroneous and mislabelled images, such as images capturing footprints (spoor). Additionally, images may have captured multiple wildlife individuals within the same image, with varying quality. In order to extract images used to train a wildlife classifier, the raw images are pre-processed using Mega-detector. Mega-detector is a trained YOLO object detection model, capable of detecting objects from 3 classes (human, vehicle, and wildlife). The resulting objects, were cropped based on the object detection bounding box, enhanced to maintain resolution, and used as observation images.
For in-depth explanation, and visual aids please visit the Animal-detector repository documentation available here. The repository contains a working Mega-detector model, and Python scripts performing the image cropping and enhancement from the Mega-detector processing.
Spatiotemporal
The spatiotemporal data is sourced from Open-Meteo Weather API. The table below details the set of collected spatiotemporal metadata per observation in order to generate the spatiotemporal snapshot.
For more information on the Open-Meteo Historic API please visit Open-Meteo Historic Weather API. The table below uses information from the API documentation to describe some collected features. Additionally, the table details additional information extracted from the metadata such as day/ night, light/ dark, terrestrial, etc.
| Feature | Description | Unit/ Format | Timeframe |
|---|---|---|---|
| Observed on | Time of observation | ISO8601 | Instant |
| Coordinates | World Geodetic System (WGS84) | (latitude, longitude) | Instant |
| Positional accuracy | Publicly available positional accuracy | \(R\) | Instant |
| Elevation | Meters above sea level | Meters (m) | Instant |
| Ground temperature (2m) | Air temperature 2 meters above ground | Celsius (\(^{\circ}\) C) | Hourly |
| Relative humidity (2m) | Humidity 2 meters above ground | Percentage (%) | Hourly |
| Dew point (2m) | Dew point 2 meters above ground | Celsius (\(^{\circ}\) C) | Hourly |
| Apparent temperature | Real feel temperature considering additional factors | Celsius (\(^{\circ}\) C) | Hourly |
| Surface pressure | Atmospheric air pressure at the surface. | Hectopascal (hPa) | Hourly |
| Cloudcover | Cloudcover within the immediate area | % of area covered | Hourly |
| Low cloudcover | Cloudcover and fog up to an altitude of \(3\) kilometers | % of area covered | Hourly |
| Mid cloudcover | Cloudcover from \(3-8\) kilometers altitude | % of area covered | Hourly |
| High cloudcover | Cloudcover from \(8\) kilometers altitude | % of area covered | Hourly |
| Wind speed (10m) | Wind speed at 10 meters above ground | kilometers per hour (km/h) | Hourly |
| Wind speed (100m) | Wind speed at 100 meters above ground | kilometers per hour (km/h) | Hourly |
| Wind direction (10m) | Wind direction at 10 meters above ground | Degrees (\({\circ}\)) | Hourly |
| Wind direction (100m) | Wind direction at 100 meters above ground | Degrees (\(\circ\)) | Hourly |
| Wind gusts (10m) | Maximum wind speed of the preceding hour | kilometers per hour (km/h) | Hourly |
| Shortwave radiation | Average shortwave solar radiation of the preceding hour | Watt per square meter (W/m\(^2\)) | Hourly |
| Direct radiation | Average direct solar radiation of the preceding hour | Watt per square meter (W/m\(^2\)) | Hourly |
| Diffuse radiation | Average diffuse solar radiation of the preceding hour | Watt per square meter (W/m\(^2\)) | Hourly |
| Vapor pressure dificit | A high VPD effects the transpiration of plants | Kilopascal (kPa) | Hourly |
| Evapotranspiration | Water evaporation into the atmosphere | Millimeters (mm) | Hourly |
| ET0 FAO Evapotranspiration | Metric estimating required irregation for plants | Millimeters (mm) | Hourly |
| Precipitation | Hourly precipitation sum (rain, showers, snow) | Millimeters (mm) | Hourly |
| Snowfall | Hourly snowfall sum | Centimeters (cm) | Hourly |
| Rain | Large scale weather systems resulting rain | Millimeters (mm) | Hourly |
| Hourly Weather code | WMO numeric weather code | WMO code | Hourly |
| Soil temperature (0cm-7cm) | Temperature in the soil at 0-7 centimeters | Celsius (\(^{\circ}\)) | Hourly |
| Soil temperature (7cm-28cm) | Temperature in the soil at 7-28 centimeters | Celsius (\(^{\circ}\)) | Hourly |
| Soil temperature (28cm-100cm) | Temperature in the soil at 28-100 centimeters | Celsius (\(^{\circ}\)) | Hourly |
| Soil moisture (0cm-7cm) | Average water content in the soil at 0-7 centimeters | Meter cubed per meter cubed (m\(^3\)/m\(^3\)) | Hourly |
| Soil moisture (7cm-28cm) | Average water content in the soil at 7-28 centimeters | Meter cubed per meter cubed (m\(^3\)/m\(^3\)) | Hourly |
| Soil moisture (28cm-100cm) | Average water content in the soil at 28-100 centimeters | Meter cubed per meter cubed (m\(^3\)/m\(^3\)) | Hourly |
| Daily Weather code | WMO numeric weather code | WMO code | Daily |
| Max temperature (2m) | Maximum daily temperature at 2 meters above ground | Degrees (\(\circ\)) | Daily |
| Min temperature (2m) | Minimum daily temperature at 2 meters above ground | Degrees (\(\circ\)) | Daily |
| Apparent temperature max | Maximum real-feel temperature at 2 meters above ground | Degrees (\(\circ\)) | Daily |
| Apparent temperature min | Minimum real-feel temperature at 2 meters above ground | Degrees (\(\circ\)) | Daily |
| Precipitation sum | The sum of daily precipitation (rain, showers, snowfall) | Millimeters (mm) | Daily |
| Rain sum | Sum of daily rain | Millimeters (mm) | Daily |
| Snowfall sum | Sum of daily snowfall | Centimeters (cm) | Daily |
| Precipitation hours | The number of hours with rain in a day | \(Z\) | Daily |
| Sunrise | Local sunrise time | ISO 8601 | Daily |
| Sunset | Local sunset time | ISO 8601 | Daily |
| Wind speed max (10m) | Maximum daily wind speed 10 meters above ground | Kilometers per hour (km/h) | Daily |
| Wind gusts (10m) | Maximum daily gust speed at 10 meters above ground | Kilometers per hour | Daily |
| Dominant wind direction | Dominant daily wind direction for winds at 10 meters | Kilometers per hour (km/h) | Daily |
| Shortwave radiation sum | The daily sum of short wave radiation | Megajoules per meter squared (MJ/m\(^2\)) | Daily |
| Daily evapotranspiration | Sum of daily evapotranspiration | Millimeters (mm) | Daily |
| Terrestrial | Terrestrial or aquatic observation | \(\{0, 1\}\) | Instant |
| Hemisphere | Location lies in the northern/ southern hemisphere | \(\{0, 1\}\) | Instant |
| Day | Sighting occurrence in light/ dark | \(\{0, 1\}\) | Instant |
| Season | Season of sighting, dependent on hemisphere | Season | Instant |
Dataset Taxonomic Breakdown
Taxonomic Family
| Taxonomic Name | Observation Count | Image Count |
|---|---|---|
| Felidae | 44710 | 26668 |
| Elephantidae | 11292 | 14831 |
Taxonomic Genus
| Taxonomic Name | Observation Count | Image Count | Family |
|---|---|---|---|
| Lynx | 20139 | 12121 | Felidae |
| Panthera | 12411 | 7485 | Felidae |
| Puma | 5239 | 2725 | Felidae |
| Leopardus | 2220 | 1391 | Felidae |
| Acinonyx | 1872 | 1245 | Felidae |
| Felis | 991 | 633 | Felidae |
| Caracal | 595 | 323 | Felidae |
| Herpailurus | 448 | 315 | Felidae |
| Leptailurus | 442 | 279 | Felidae |
| Prionailurus | 291 | 124 | Felidae |
| Neofelis | 24 | 11 | Felidae |
| Otocolobus | 20 | 6 | Felidae |
| Pardofelis | 11 | 4 | Felidae |
| Catopuma | 7 | 4 | Felidae |
| Loxodonta | 9407 | 12810 | Elephantidae |
| Elephas | 1885 | 2021 | Elephantidae |
Taxonomic Species
| Taxonomic Name | Observation Count | Image Count | Family |
|---|---|---|---|
| Lynx rufus | 19043 | 11641 | Felidae |
| Panthera leo | 6782 | 4935 | Felidae |
| Puma concolor | 5239 | 2725 | Felidae |
| Panthera pardus | 2964 | 1360 | Felidae |
| Acinonyx jubatus | 1872 | 1245 | Felidae |
| Panthera onca | 1831 | 745 | Felidae |
| Leopardus pardalis | 1324 | 866 | Felidae |
| Panthera tigris | 788 | 433 | Felidae |
| Lynx canadensis | 727 | 355 | Felidae |
| Caracal caracal | 573 | 299 | Felidae |
| Herpailurus yagouaroundi | 448 | 315 | Felidae |
| Leptailurus serval | 442 | 279 | Felidae |
| Felis lybica | 372 | 252 | Felidae |
| Leopardus weidii | 362 | 209 | Felidae |
| Felis silvestris | 345 | 228 | Felidae |
| Leopardus geoffroyi | 288 | 174 | Felidae |
| Lynx lynx | 260 | 75 | Felidae |
| Felis chaus | 228 | 138 | Felidae |
| Prionailurus bengalensis | 178 | 76 | Felidae |
| Lynx pardinus | 109 | 50 | Felidae |
| Prionailurus javanensis | 77 | 31 | Felidae |
| Leopardus tigrinus | 67 | 36 | Felidae |
| Leopardus guigna | 55 | 37 | Felidae |
| Leopardus guttulus | 47 | 33 | Felidae |
| Panthera uncia | 46 | 12 | Felidae |
| Caracal aurata | 22 | 24 | Felidae |
| Otocolombus manul | 20 | 6 | Felidae |
| Neofelis diardi | 19 | 8 | Felidae |
| Felis nigripes | 17 | 4 | Felidae |
| Prionailurus viverrinus | 17 | 9 | Felidae |
| Leopardus pajeros | 16 | 10 | Felidae |
| Leopardus colocola | 16 | 7 | Felidae |
| Leopardus garleppi | 12 | 6 | Felidae |
| Prionailurus rubiginosus | 12 | 5 | Felidae |
| Pardofelis marmorata | 11 | 3 | Felidae |
| Leopardus emiliae | 10 | 5 | Felidae |
| Felis margarita | 10 | 1 | Felidae |
| Leopardus braccatus | 9 | 5 | Felidae |
| Leopardus jacobita | 7 | 2 | Felidae |
| Prionailurus planiceps | 7 | 3 | Felidae |
| Catopuma temminckii | 6 | 2 | Felidae |
| Neofelis nebulosa | 5 | 2 | Felidae |
| Felis bieti | 4 | 0 | Felidae |
| Leopardus fasciatus | 2 | 0 | Felidae |
| Catopuma badia | 1 | 0 | Felidae |
| Loxodonta africana | 8939 | 12313 | Elephantidae |
| Elephas maximus | 1885 | 2021 | Elephantidae |
| Loxodonta cyclotis | 222 | 214 | Elephantidae |
Taxonomic Subspecies
| Taxonomic Name | Observation Count | Image Count | Family |
|---|---|---|---|
| Panthera leo melanochaita | 5587 | 4035 | Felidae |
| Panthera pardus pardus | 2449 | 1160 | Felidae |
| Acinonyx jubatus jubatus | 1642 | 1079 | Felidae |
| Panthera leo leo | 786 | 672 | Felidae |
| Panthera tigris tigris | 710 | 430 | Felidae |
| Lynx rufus fasciatus | 410 | 263 | Felidae |
| Puma concolor couguar | 357 | 172 | Felidae |
| Caracal caracal caracal | 349 | 163 | Felidae |
| Lynx rufus rufus | 262 | 163 | Felidae |
| Panthera pardus fusca | 222 | 107 | Felidae |
| Puma concolor concolor | 167 | 99 | Felidae |
| Felis lybica cafra | 167 | 112 | Felidae |
| Panthera pardus kotiya | 164 | 77 | Felidae |
| Leptailurus serval lipostictus | 152 | 96 | Felidae |
| Felis silvestris silvestris | 144 | 98 | Felidae |
| Felis chaus affinis | 105 | 69 | Felidae |
| Leptailurus serval serval | 87 | 54 | Felidae |
| Felis lybica lybica | 81 | 74 | Felidae |
| Leopardus pardalis mitis | 58 | 44 | Felidae |
| Leptailurus serval constantina | 57 | 46 | Felidae |
| Acinonyx jubatus hecki | 46 | 37 | Felidae |
| Prionailurus javanensis sumatranus | 38 | 16 | Felidae |
| Prionailurus bengalensis euptilurus | 32 | 16 | Felidae |
| Lynx rufus escuinapae | 27 | 16 | Felidae |
| Felis lybica ornata | 26 | 16 | Felidae |
| Prionailurus bengalensis bengalensis | 22 | 13 | Felidae |
| Neofelis diardi borneensis | 16 | 8 | Felidae |
| Panthera tigris sondaica | 15 | 0 | Felidae |
| Panthera pardus delacouri | 15 | 10 | Felidae |
| Lynx lynx lynx | 14 | 6 | Felidae |
| Leopardus guigna guigna | 13 | 9 | Felidae |
| Leopardus pardalis pardalis | 11 | 8 | Felidae |
| Panthera pardus tulliana | 11 | 5 | Felidae |
| Panthera tigris altaica | 11 | 0 | Felidae |
| Panthera pardus melas | 10 | 1 | Felidae |
| Caracal caracal nubicus | 10 | 9 | Felidae |
| Felis silvestris caucasica | 9 | 7 | Felidae |
| Prionailurus javanensis javanensis | 8 | 1 | Felidae |
| Prionailurus viverrinus viverrinus | 8 | 5 | Felidae |
| Lynx lynx dinniki | 8 | 7 | Felidae |
| Leopardus guigna tigrillo | 5 | 4 | Felidae |
| Lynx lynx isabellinus | 5 | 0 | Felidae |
| Lynx lynx carpathicus | 5 | 4 | Felidae |
| Leopardus tigrinus oncilla | 4 | 0 | Felidae |
| Pardofelis marmorata marmorata | 4 | 2 | Felidae |
| Elephas maximus indicus | 877 | 920 | Elephantidae |
| Elephas maximus maximus | 567 | 740 | Elephantidae |
| Elephas maximus borneensis | 180 | 146 | Elephantidae |
| Elephas maximus sumatranus | 18 | 12 | Elephantidae |