We developed a fast and inexpensive method for estimating the composition of boreal ground fuels using smartphones and downward facing (nadir) photos. We demonstrate our proof-of-concept in this methods paper: Cameron, Panda, Barczyk and Beverly 2022

Descriptions of forest ground cover are important for predicting potential fire behaviour. Surface fuels on the forest floor can include different types of biomass such as dead needles and leaves, moss, lichen, grass, forbs and fallen dead woody debris (i.e., branches). The types and amounts of these combustible materials on the forest floor will determine how easily the fuels ignite; and will directly affect the intensity of any wildfires that spread through the area.

​Measuring forest ground cover manually with field campaigns is labour intensive and time consuming. Due to the costs involved, field measurements are not widely used. Visual assessments offer a comparatively inexpensive alternative, and photo guides have long been used to help fire managers rapidly assess surface fuel conditions.

In this study, we analyzed the viability of using nadir or downward photos from smartphones (iPhone 7) to provide quantitative ground cover and biomass loading estimates. First, we collected sets of downward photos at the locations of fuel inventory plots. Then we had a technician partition (i.e., segment) the photos into different cover types on a desktop computer.

Good correlations were found between field measured values and pixel counts of each cover type identified by the technician using the photos. Although promising, segmenting photos manually was labor intensive and therefore costly. So we explored the viability of using a trained deep convolutional neural network (DCNN) to segment the photos automatically.

The DCNN was able to segment our downward photos with 95% accuracy when compared with the manually delineated photos. Our image segmentation algorithm also exhibited promising results when we used it to assess ground cover in a different location that had similar surface vegetation characteristics.