Model selection for trends of basal area

#1
I'm currently working analizing basal area trends contemplating three different temporal points using national forest inventories from two different countries in Europe (Spain and Sweden). Time between censuses varies from countries and I think it's something I should have into consideration.

The main idea is analize how extreme climatic events have affected these basal area trends over time. So I'm using the periods between inventories to calculate the frequency and intensity of these extreme climatic events.

My problem is that I'm not sure what model should be used to analize the effects of these events in basal area trends. Something I don't know how to manage is that basal area trends between de first and the second period are no independent each other. Also to analize the effect of extreme climatic episodes I was calculating annual mean climatic variables, and they are highly correlated with annual mean extreme climatic variables (droughts, heatwaves...). I'd like to have some suggestion about how to start dealing with this problema.

If anybody could suggest me ideas or factible ways to deal with this problem, I would be highly thankfull.

Thanks in advance for your time.
 

katxt

Well-Known Member
#2
I would start with a simple model. It isn't clear just what is going on but this is what I imagine.
You have a collection of sites in two countries. For each site you have basal area estimates at various times, and climate data of some sort over the periods between the times.
A simple model would be to make the response the average % change per year from earliest to latest time points.
The first predictor would be country. A second (and maybe a third or fourth) predictor would be the average climate event per year over the same period. These predictors would be continuous.
This structure is that of a general linear model (glm) with both categorical and continuous predictors. It can also be done with multiple regression if you're not happy with the glm.
A more complicated thought. If you wanted to keep the two periods as a before/after scenario, just try the same trick for each of the two periods. You now have another categorical variable - Time1 and Time2. The basal levels may not be independent, but the % differences will be.
 
Last edited: