Finally we’ve come around to cleaning and publishing the data from the first version of virtualforest.io. This data was collected with acquisitions from 10min to a base setting of half an hour, for a total of roughly 13K images, across mostly 2017. The virtual forest project ran from late 2016 until early 2018 in Harvard Forest, Petersham, MA, USA. The project took spherical images of the forest canopy for an interactive online installation equivalent to the current project in Gontrode, Belgium.
Since these data might have research purposes outside the scope of the original outreach project we’ve now cleaned the data and published them online for re-use (with attribution) on Zenodo.
You can find the full dataset on zenodo at https://zenodo.org/record/4899637. A brief description of the data and procedure used to clean the data is provided below. Please cite the data as when using this data in any way:
Hufkens, Koen. (2021). Spherical forest phenology images - 2017 - Harvard Forest (Version v1.0) [Data set]. Zenodo. http://doi.org/10.5281/zenodo.4899637
The camera was located in Harvard Forest, Petersham, MA, USA. The exact location of the camera was 42.535685326, -72.189706396 decimal latitude and longitude and an elevation of 337 m (as determined by high precision GPS).
Hardware & Software
The setup consisted of a Ricoh Theta S camera, tethered via USB to a Raspberry pi. Images were taken on a varying time schedule (for visualization purposes), but guaranteed a half hourly acquisitions (during daytime). Images were collected at both a normal (exp0) and 2-stop underexposed value (exp-2). Only the full growing season of 2017 is included in this dataset. A detailed description, with updated hardware (Ricoh Theta V) is provided in the photosphere repository of BlueGreen Labs.
The naming convention of the files is compatible with the processing toolchain of the PhenoCam US network. Images were reprocessed from equirectangular to hemi-spherical images (top and bottom), to reduce their file size. The top of the image is approximately north in both up and downward looking orientations.
Processing of the images for colour indices is easily accomplished using the software of the PhenoCam project. These include the vegindex python package to convert image data to colour indices. Data were pre-processed for a lower, mid and high canopy region of interest (30 degree bands) and the understory of the forest. I report 3-day agregated values. Transition dates can be extracted using our phenocamr R package.
The data is delivered as a large compressed file. After extraction several folders with data are provided. The
virtualforest_canopy* folders include canopy data, looking overhead. The
virtualforest_canopy_underexposed folder was not processed for colour indices but includes images which are 2 stops underexposed as common when using hemi-spherical images to determine a leaf area index (LAI). These images can therefore be used within this context. The
virtualforest_understory folder includes all downward looking images (matching those in the
virtualforest_canopy folder). Colour indices for the understory were also calculated.
When colour indices are calculated the results can be found in the
ROI sub-folder, following the structure as described in the vegindex package.
Please contact us at email@example.com if you have any questions regarding this data. Equirectangular data files equivalent can be provided upon request, file size prohibits easy sharing however.