Tree growth is an important aspect of forestry and forest ecology. Typically a growth rate is calculated as the difference between two stem diameter measurements over a given time interval: (dbh2 - dbh1) / dt
There are two sources of error which can occur to lead to erroneous (and oftentimes negative) growth estimates: 1) small measurement error, which is typically is proportional to the size of the stem. These errors can be due to slacked or crooked diameter tapes or not exactly measuring the diameter point on the tree stem, etc. and 2) recording error. Measurements may be by off by a decimal place, having transposing numbers, data collection or entry errors, etc.
Alternatively, trees may have the same dbh recorded at both time intervals, giving a growth rate of zero.
Unreasonably large tree growth estimates can be identified as outliers using a quantile approach, but many types of growth models (i.e., gamma likelihood models) will not allow for negative or zero growth rates.
How do you deal with those zero/ negative growth measures? If you have suggestions beyond simply not including them, please share.
Thank you
1. Adjust Model Parameters: Modify the model parameters such as initial conditions, growth rates, or mortality rates to account for zero or negative growth rates.
2. Implement Constraints: Introduce constraints in the model to prevent unrealistic scenarios like negative growth rates, ensuring that the model reflects real-world tree growth patterns accurately.
3. Consider Alternative Models: Explore alternative tree growth models that can better accommodate zero or negative growth rates, such as stochastic models or models with different underlying assumptions.
I would add another reason for negative growth rates: real change. On large trees with thick bark, sloughing of bark can result in a decrease in diameter measurement with a tape, even though the tree is still adding wood increment. Likewise, trees can be measured with unrealistic increases in diameter due to stems splitting or lumps forming at DBH. These situations could be identified and corrected in the field by taking an increment measurement from a tree core (if your species dependably produces annual rings). The paper by Woo et al. 2020 (Article Tree Height Increment Models for National Forest Inventory D...
) tried various ways of dealing with negative values (in this case height growth, but it's the same idea). I would never throw out data unless it's clearly a blunder and not standard measurement error.- Badges
- Science topic
- Similar topics
- Biological Science
- Bioecology
- Systems Ecology
- Landscape Ecology
- Connectivity