search

weecology / NEON_crown_maps

Generating tree crown maps for NEON sites
MIT License
1 stars 0 forks source link

BONA Tree Fall Rate #24

Closed bw4sz closed 4 years ago

bw4sz commented 4 years ago

@marconiS and I had some doubt as to the validity of BONA results in this figure. Documenting here.

The largest tree fall counts were 

>>> fall_mean.sort_values("mean")[-5:]
     Site       geo_index  mean
172  TEAK  317000_4108000  3048
583  BONA  472000_7231000  3065
58   BONA  473000_7221000  3265
419  BONA  477000_7231000  4814
403  BONA  470000_7224000  6256
bw4sz commented 4 years ago

TreeFalls

bw4sz commented 4 years ago

As we expected this is an artifact. BONA has very short and thin trees. the 1m canopy height model is too coarse to capture changes, leading to small artifacts where the entire tree fits inside 2-4 pixels. The smoothing here makes alot of difference. For example a drop from a 4m in 2018 to 2m in 2019 is a drop of 50%. Plus the 2019 height is below the 3m threshold we use to identify trees!

The original algorithm said if there is a drop of more than 30% height call it valid. Should we add in an absolute measure? More than 30% height and the final height is > 5m? @marconiS @ethanwhite what do you think? Do we need the % measure at all? My thought is if it is a 100m tree and it drops 10m its not a tree fall. We should ask stephanie what kind of bias this might introduce.