site productivity

[gregjohnsonbiometrics]

In the context of using Forest Inventory and Analysis (FIA) data for CONUS-wide growth modeling, site index is flawed for a number of reasons:

1. It is not consistently obtained for each plot due to missing site trees for each subject species,
2. It is estimated using a number of different and not necessarily compatible equations,
3. The available equations do not all use the same base age, and
4. The site tree measurements are informing the prediction system of effects to be predicted (i.e., not available in practical use scenarios).

It seems desirable to derive a measure of productivity (location effects) that is independent of the growth measurements being predicted.

We need a measure that is:

• observable
• repeatable
• highly correlated with location-specific effects
• available or derivable for all of CONUS

[holoros]

I think this is always the biggest limitation and should be resolved first. This raster isn't perfect, but should be doable for western US? I could regenerate the climate site index using more consistent data.

[gregjohnsonbiometrics]

I think that's a good idea Aaron. Since we have the admixture of site index estimates and they work reasonably well to quantify the location effect, we can use model bias trends across site index to evaluate the effectiveness of the climate site index data.

[gregjohnsonbiometrics]

Here is a graph of mean diameter growth residuals from the Alternative Model without site index (localizing variable). The under-predictions (dark colors are clustered to the west and the over-predictions -- lighter color -- are to the east). I think you can see the Cascades in the transition zone between under and over predictions. I'd like to think through using something like this to create an empirical productivity map.

[holoros]

This is impressive and aligns with Vanclay's idea that standardized diameter increment can serve as a proxy for site quality. I bet they correlate strongly with various environmental factors. For HT-age relationship, I always liked the idea of using quantile regression to better characterize patterns and a 50% percentile could be used as the base? Good to see the continued progress on this!

[gregjohnsonbiometrics]

@holoros , do you have a reference for Vanclay's idea?

[holoros]

This idea is not well explained in the paper, but he references it in his 1994 book (see pgs 143-145).

vanclay 1994 - modelling forest growth and yield.pdf

0612015v1.pdf

[gregjohnsonbiometrics]

@holoros , what would it take to get a Climate SI for all of CONUS? Do you already have it?

[gregjohnsonbiometrics]

Some ramblings from David.

SiteIndexRamblings_20260119.docx

[gregjohnsonbiometrics]

@DMtree-DougFir Here are the loblolly pine residuals for the alternative no site index equation by EPA Level II Eco Region. A hint of an over-prediction for the SE USA Plains, but it is small.

[holoros]

This is my first attempt at a North American-wide climate site index, and the map is far from perfect, but I figure we could test and see how it performs. I think this could be refined with additional inputs, but I am not convinced this is the ideal approach?

The weird line bisecting the map is caused by different site index equations. For the eastern half, I used species-specific and unified equations by Carmean et al. (1989), but in the west I just used a Douglas-fir equation by Monserud (1984). We could probably mine FIA to use a consistent equation for the western species?

https://drive.google.com/file/d/1waDAJMYXCdd7o0VksxcAhtMjFMbJHLNm/view?usp=sharing

Suggestions?

Carmean, Willard H.; Hahn, Jerold T.; Jacobs, Rodney D. 1989. Site index curves for forest tree species in the eastern United States. General Technical Report NC-128. St. Paul, MN: U.S. Dept. of Agriculture, Forest Service, North Central Forest Experiment Station. 142 p.

Monserud, R.A., 1984. Height growth and site index curves for inland Douglas-fir based on stem analysis data and forest habitat type. Forest Science, 30(4), pp.943-965.

[gregjohnsonbiometrics]

@holoros , I think the lines are problematic for our purposes of a CONUS-wide, no variant type of model. Within the bands (I see more than two), the si variation looks "reasonable" (whatever that means). FIA site index is wild. There a multiple equations used for a given species and multiple base ages (making the si value itself incompatible).

This is similar to the issue with using USGS soil site index. Soil mapping boundaries show up in a similar way to the equation boundaries in your map.

[holoros]

Good feedback and I have recompiled the North American ht-age data I had and fit a species-specific Champan-Richards site index to the data, which generates more consistent estimates.

The full database is here and I'll work to make a wall to wall map now:

https://drive.google.com/file/d/1vrt1-auDFYsw6aJEgyoleSkmVooJCUdQ/view?usp=sharing

[gregjohnsonbiometrics]

@holoros , I'm wondering about the circular features in the map and the apparent high site values east of the cascades into Idaho and Wyoming.

[holoros]

This was observed values and just a function of how I made the map. For this map of predicted values, I used ClimateNA predictors, EPA North American ecoregions, and a consistent species-specific site index equation that was derived from the data.

https://drive.google.com/file/d/1YANsYj3yAgwk21p6K1_t3OqfcFAedByk/view?usp=sharing

[gregjohnsonbiometrics]

@holoros , I can't download the tif because it requires UM credentials (which have expired for me).