Development and application of growth and mortality in SORTIE-BC

Presented by Dave Coates.

Growth and mortality act on:

• Seedlings - problematic - varies by species and ecosystem - the seedling size matters
• Saplings
• Adults

Resource - light

Juvenile growth is based on light availability from fisheye photographs. Basing growth on light works quite well, even though there may be other factors influencing growth. The model estimates species-specific transmission of light (i.e. % openness of a crown) that can be combined with data on the distribution, species, and size of neighbouring trees to calculate the percent of incident PAR that reaches the crown of each individual seedling and sapling (GLI). Neighboring tree crowns are modeled as cylinders.

Benefits of this approach:

• The model accurately predicts three-dimensional spatial variation in understory light levels in mixed-species forests
• Allows estimates of light levels experienced by individual seedlings and saplings
• Explicitly incorporates crucial spatial variation in light levels in and around canopy gaps

Sub-canopy growth

The function used to calculate juvenile growth is Michaelis-Menton - but other forms might be needed by other groups, so flexibility would be nice.

We are also interested in the effect of history on growth as a function of light - history doesn't matter for shade tolerants, but matters more and more the less shade-tolerant a species is. This makes predicting post-disturbance events interesting.

Sub-canopy mortality

Juvenile mortality starts high and drops off with increasing size. For mature trees, mortality is fairly constant, and then mortality increases again for older trees.

Separate mortality into 1) disturbance-induced mortality, and 2) mortality due to competition. Predict competition-induced mortality as a function of prior growth.

We use self-thinning for dense, young populations but it's very "feeble" - would like to discuss improvements.

Canopy tree growth

For management, we need to improve prediction of growth for canopy trees. There are non-spatial predictors, and we're actively working on spatially-explicit factors (competition).

General elements of competition:

• Neighbourhood Competition Index (NCI) - several forms are possible
• Shading
• Tree diameter

Analytical approach - likelihood estimation, hypothesis testing, comparison of alternate models.

Canopy tree mortality

• Disturbance-induced - insects, disease, wind, etc.
• Mortality due to competition (i.e. self-thinning) - this needs work - needs further discussion
• Senescence mortality

Discussion of this talk

In the old model mortality was either stochastic, or senescence for adults. Parameterization of senescence was difficult - there were self-sustaining cycles that sometimes amplified even at low levels of mortality.

It has been shown that what we think of as stochastic mortality is relatable to growth rate - and probability of windthrow death was also relatable to growth. Currently this is a manuscript.

To use inventory data for adult growth - you often don't know age of tree, just size - it's hard to separate size and age effects. But trees have two ages - chronological and physiological. We'll take this point to the general discussion.

In NZ - tried to see how much slower trees grew before they died - significant effect for young trees.

For using growth history - might want to include all options for equation forms - different options would work better in different systems. The current system is for users to tell us what equation forms they want and we supply them - but still, might want to anticipate demand to some degree.

Question: Can you separate suppression from effects of herbivory for small trees? Let's move this to herbivory discussion.

The placement of the threshold between juvenile and adult trees is very important - it has huge management implications.

Sapling growth as a function of light, soil moisture, and foliar N for 4 species across a landscape fertility gradient in northern lower Michigan.

Presented by Rich Kobe

Hypotheses on sapling growth

1. Species associated with mesic, high-fertility sites show higher sensitivity of growth to soil moisture and foliar N.
2. There is a trade-off among species between growth at high soil resources versus low soil resources.

Methods

• Searched for saplings across a broad range of light availability and soil conditions at Manistee National Forest
• Included species with different primary site associations -sugar maple, red maple, red oak, and A. beech
• Measured
  • light availability (hemispherical canopy photography)
  • soil moisture (time domain reflectometer) to 30 cm depth
  • Foliar N (CHN elemental analyzer)
  • Stem cross sections for radial growth / radii (Windendro)
• Regressed radial growth against radius, light availability, soil moisture, and foliar N.
• Tested factors and compared model formulations with maximum likelihood techniques.

Growth model

Sapling growth used double Michaelis-Menton function. We tried using different dependencies on water and N to tease apart the effects of each.

Results

Sugar maple:

• Light level is significant
• When light is low, increase in light OR increase in water leads to increased growth

Red Oak:

• Light and N significant
• Water not significant
• Foliar N acts on asymptotic growth
• At low light, light is only limiter; N only limits when light levels are high. Region of co-limitation?

American Beech similar to Red Oak

Red maple - Water and light co-limit at low light levels, N limits at high levels.

Conclusions

First hypothesis above - not true.

Second hypothesis - also no evidence.

Summary

N affected growth in 3 of 4 species

• operated at high light
• unlikely to affect understory dynamics

Soil water affected growth in 2 maple species at all light levels (double MM function)

S. maple sensitivity to N and water < r. maple and r.oak

There is evidence of a growth / survivorship trade-off - growth rates on fertile soil are inversely proportional to survivorship on low fertility soils.

There is a recent paper (Clark) - it's the stochasticity around the mean growth response that's important at community level. However, individual response explains a lot. What do people think?

If light, water, N, and tree size are used together to predict growth - R² is 0.75

N is unlikely to affect successional dynamics.

Soil microbes affect mortality - sterile soils have lower mortality - need to tease out effects of individual microbes.

Discussion of this talk

Messier student regressed growth against light, size and regressed the leftover against competition - competition explained much of what was left (so omitting use of soil resources).

Question: sterilizing soil kills micorrhyzae - does this explain? Effects of microbes appear to swamp that effect.

Is sugar maple's endo-mycorrhyzal status the reason for its response to N? Unknown, needs discussion.

Neighborhood effects on tree growth and mortality

Presented by Maria Uriarte

Approach

We can try to predict growth and mortality based on what happens in tree's neighborhood.

A tree has potential maximum growth which is decreased by neighborhood effects. Multiple effects are multiplicative. Estimating maximum growth is very difficult - it is hard to get data.

A size effect is optional. For a small range of DBH, you could use a linear relationship with DBH (size effect only).

Neighborhood crowding index

There is an effective neighborhood radius within which neighbors have an effect. Effects of neighbors increase with neighbor size and decrease with neighbor distance.

You can add an angular distribution parameter to add effects of neighbor position

Application

This approach works only with a large dataset since there are a large number of parameters needed (7 + n number of neighbor groups). You can study the difference between con and heterospecific neighbor effects by varying the parameters. The R² tends to be quite low - but the study areas are tropical with 10² species.

Discussion of this talk

Coates comment - you must have small fast-growers in your dataset for accurate max growth, and large old trees to get the function tails - and 50-100 individuals of each species across range of conditions. Then the simulations are very accurate.

SORTIE applications using large-scale forest inventory data: Challenges for eastern deciduous forests.

Presented by Will McWilliams.

The current forest service model is a large scale approach - there's a need for individual-based, more complex modeling.

Potential applications:

• Parameterization for general forest associations
• SORTIE data warehouse
• Projections of future forest conditions
• Identify older stand character
• Understand species composition shifts

Eastern broadleaf deciduous forests dominate in the eastern United States. They are extremely complicated.

FIA sampling

FIA sample grid - all states surveyed on 5-7 year cycles - gives good estimate of resource conditions. The grid is a set of hexagons mapped on a rotating basis, like this:

New sample design:
Trees 5.0 inches and larger: Four 24-foot radius fixed subplots spaced 120 feet apart.
Trees from 1.0 to 4.9 inches: One 6.8-foot radius plot per subplot.

Mapping Forest Conditions:
Forest conditions are mapped and separate condition-level variables are computed, i.e. land use, forest type, stand origin, and stand size.

Remeasurement:
All remeasurement is confined to the central 24-foot radius subplot

Currently this new mapping strategy is being implemented - some trees from the old dataset are being remeasured and the proportion of remeasured trees will increase through time.

Variables captured

• Slope and Aspect
• Land Use
• Stand Origin
• Owner Class

• Species
• DBH
• Tree history
• Total Height
• Azimuth
• Distance

Forest health monitoring measurements

The samples are widely spaced.

• Down Woody Debris - Coarse and Fine woody material on the forest floor
• Crown Condition Classification
  • Crown Ratio
  • Crown Density
  • Crown Dieback
  • Foliage Transparency
  • Vigor Class (seedlings)
• Vegetation Structure - Plant Occupancy of a three dimensional cone by Life Form
• Ozone Bioindicator Plants
  • Species
  • Number
  • Amount and Severity of Injury
• Lichen Communities - Collection of Voucher Specimens and Estimate of Abundance
• Soils
  • Soil Litter
  • Erosion Measures
  • Soil Texture
  • Lab Analysis of pH
  • Organic Carbon
  • Nitrogen
  • Exchangeable Calcium
  • Exchangeable Magnesium
  • Exchangeable Potassium
  • Phosphorus

Advantages of dataset

• Consistent measurements across the US - large studies encouraged!
• Spatially-explicit tree data
• Temporal data is building over time

Disadvantages

• Only covers forestland
• Some areas have not been inventoried (but eastern coverage is pretty good)
• Can be challenging to work with
• Do not collect enough details (like available light)

Other challenges

• Overstory composition does not match understory composition
• Often there isn't any tree regeneration - future composition unknown
• Herbivory rampant - white-tailed deer are extremely destructive to seedlings - overpopulated by factor of 2

Discussion of this talk

Question: Deer population size - based on? Enclosure study determined population density threshold above which deer are destructive to seedling regeneration.

Potential for community-level/ecosystem studies is huge. Vision is for FIA, NED, and SORTIE to be linked.

Question: How good is the location of stem mapping? Good - down to an inch, so trees can be accurately be identified for remeasurement.

Parameterization and application of SORTIE for mixedwood boreal forests of Quebec

Presented by Julie Poulin.

Species fall into shade tolerant and intolerant categories. After disturbance - shade intolerants grow to canopy and tolerants grow up underneath. A shrub was recently added into datasets.

Light submodel

Canopies are currently modeled as cylinders. Power functions for canopy radius and crown height fits the data better.

The method for parameterize openness:

1. Take a fisheye photo.
2. Use software to trace outline of crown.
3. Pixelate to black and white.
4. Calculate percentage of black.

To quantify light interception by snags, we defined three decay classes and calculated openness using the same method as for live trees. Conifers still block around 50% of light in decay class 1.

Discussion of this talk

With a very deterministic growth function, there was a huge cohort effect - so a subtle senescence effect and a stochastic effect was needed.

Question: Can you describe the dataset? Table reported density each 5 years until all trees were dead. So it fits a stand-level thinning curve.

Question: Would it have been possible to use permanent plot data? It was used for random mortality. What about mortality per DBH class? That's something Julie will try. The permanent plot had very few old stands in it so it couldn't be used for old-age senescence mortality.

Question: The successional sequence in Quebec - what relationship to stands at Date Creek? Same effect is seen - first shade intolerants, then shade tolerants. So initial conditions are very important for Quebec simulations.

General growth/mortality discussion

Self-thinning - couldn't use in black spruce stands because mortality rates were too high. Self-thinning is straightforward - you take empirical data - then work out mortality rate as a linear equation (although non-linear form is desired). It works, but it's not a good idea - we don't have a mechanistic approach of the self-thinning function. Juvenile mortality is a function of resources but we predict it as a function of light - when we sample, we pick a well-growing, non-crowded tree. We need to sample crowded trees too - this may modify the growth functions to reproduce mortality as a function of growth. Also, maybe neighborhood effects will produce the desired effect - Maria's data supports this.

We kill trees when they grow slowly - but maybe they just don't produce rings and survive a while - are we overestimating mortality? Does Dave check for missing rings? Can't get the data. Charlie says - this effect is just noise, and the problem is underestimation of mortality rather than overestimation.

Question: Could we model different genotype growth responses? Sure - the tradeoff is increased memory allocation per tree. Heritability would be very interesting - as is microsite effects.

Question: What are Dan Kneeshaw's results on juvenile mortality as function of size? In small trees, the mortality is like Rich Kobe's model - but in larger size classes the differences in species is smaller - the shade tolerant are getting less tolerant as they get bigger, and the intolerants get more tolerant.