vegetation traits from landscape photos

As a grassland ecologist, I use landscape photography to collect field data, which speeds up fieldwork and reduces disturbance to the study area. Photos can capture vegetation composition, dominant species or species groups, vegetation changes over time, the appearance of invasive species, disturbances, and more. The drawback of photo-based data collection is that extracting information from images is currently very time-consuming and still relies heavily on manual processing.
One idea for the workshop could be exploring ways to extract vegetation traits from landscape photos. I am wondering whether it would be possible to identify major vegetation traits such as trees, shrubs, perennial bunchgrasses, annual grasses, and forbs. Could we go beyond these broad groups to recognize, for example, junipers, pines, large and small sagebrush shrubs, or large and small bunchgrass? Could we distinguish plants with various growth forms and leaf types—for instance, mat-forming plants, grasses, broadleaf plants, divided-leaf species, and needle-leaf species? Could we even identify invasive species such as cheatgrass?
Another potentially useful application would be estimating plant phenology based on the dates when the photos were taken. For example, could we determine the timing of cheatgrass green-up and die-off across different years?
The amount of bare ground is also an important parameter for studying landscape dynamics. It would therefore be extremely useful to differentiate bare ground from areas covered by litter or biological soil crust. Could we also detect signs of soil disturbance, such as erosion or cattle tracks? Importantly, would it be possible to estimate the projective cover of vegetation versus bare ground, or even the projective cover of different vegetation traits?
Additional idea is calculating patch metrics for vegetation traits and bare ground (such as mean patch size, edge density, and contagion).

No files, tests, or entry points are mentioned. Start by narrowing the proposed vegetation traits, photo data, and analysis goals into a defined scope; the issue does not yet specify a testable completion criterion.

Written by the indexing model from the issue text.