Title: Modeling Effects of Genetic Improvement in Loblolly Pine Plantations
1Modeling Effects of Genetic Improvement in
Loblolly Pine Plantations
- Barry D. Shiver
- Stephen Logan
2Modeling Genetic Effects
- Plantation Management Research Cooperative (PMRC)
established a study in 1986 in the Southeastern
USA to evaluate effects of improved genetics on
yields from block rather than row plantings - The level of genetic improvement at the time was
first generation improvement - Seedlings were planted in January 1987
3Study Design
- 16 Locations in Piedmont
- 15 Locations in Coastal Plain
- Six top ranked families in each region chosen to
represent single family genetic material - Unimproved seed obtained from region encompassed
by study - Bulk lot improved stock obtained by mixing equal
amounts of seed from the six selected families in
each region
4(No Transcript)
5Study Design
- Eight 0.4 ac. treatment plots were included at
each study installation - For this analysis only treatments one and three
are considered (no veg control and no single
family)
6Unimproved, No Veg Cntl-Age 13
7Bulk Lot, No Veg Cntl - Age 13
8Measurements
- Measurements made at ages 3, 6, 9, 12, 15, and 18
years - All typical measurements made to estimate yields
(dbh, total height, etc.) - Overall analysis results show that genetic
improvement and vegetation control significantly
improve yields and that the effects of the two
treatments are largely additive - Genetic improvement reduces fusiform rust
incidence by about half - Tree form and percentage of trees qualifying for
solid wood are significantly higher for
genetically improved plots
9Adjusting for Silvicultural Treatments
- A common method used by practicing foresters is
to adjust the exhibited SI value in an existing
growth and yield model - For most silvicultural practices (weed control,
fertilization, etc.) this method does not work
well because the response is not anamorphic (a
proportional (constant ) increase across ages)
10Silvicultural Treatment Response
- 4 type of silvicultural responses
- Type A growth gains on treated areas continue
to increase throughout the rotation. - Type B - growth gains achieved early in rotation
are maintained but do not continue to increase
after an initial response period. - Type C early growth gains are subsequently
lost. - Type D - growth gains on treated areas fall below
levels observed on nontreated areas.
11Silvicultural Treatment ResponseTypes of Response
12Silvicultural Treatment Response Pienaars
Modified Adjustment Type B
- Treatment Age, Rmax, and Yst until 90 of max
response occurs must be provided by users, so
that -
13Cultural Treatment Response
- Use Pienaar and Rheney (1995) adjustment
function to create C response
R growth response associated with the cultural
treatment of interest Yst years since cultural
treatment was applied c 1/(years to expected
maximum response) b (Maximum response)cexp(1)
14Type A Response
- Creates a response where the gain gets wider as
the stand gets older possibly even anamorphic
(stays the same amount larger proportionately as
age increases) - Same effect on height as increasing the site
index - Would get this with fertilization with P on a P
deficient site - Do we get this type response with genetic
improvement?
15Adjusting for Genetic Improvement
- Unlike the majority of silvicultural treatment
responses, there is some evidence through age 18
that the genetic treatment response is anamorphic - There is also some evidence that it is primarily
effected through a height response - A disturbing finding is the amount of variability
in the data in some cases genetic improvement
is negative rather than positive
16Dom Ht by Genetic Improvement
17Structure of our GY Models
- Models are actually a system of models
- H f (Age, Site)
- N f (Age, Site)
- BA f (Age, H, N)
- Y f (age, H, N, BA)
- With our intensive silviculture plots we have
found that if we have the basal area per acre and
the height correct we can accurately predict the
yield (weight/ac)
18Basal Area Prediction
- The actual observed heights and observed trees
per acre were used to estimate basal area per
acre using the PMRC basal area prediction
equation - Residuals were calculated and graphed
- Lack of much of a trend in residuals is an
indication that the only factor affecting basal
area is change in height
19Residual BA using Actual Height on Genetically
Improved Plots
20Height Residuals for Improved using 10 Adjusted
SI for Unimproved at Age 15
21(No Transcript)
22Genetically Improved Prediction
- Works reasonably well on average
- But, lots of variability on average the
improved are higher, but some are even lower - A real problem when trying to predict yields for
specific stands - If we have data, we can use projection from
existing inventory data in the existing stand
23Projection
- To evaluate projection, we took the dominant
height from the improved plot at age 12 - The projection was done by projecting the
dominant height from age 12 to age 18 using the
existing equation with no adjustment - The basal area at age 18 was projected from the
age 12 existing basal area using the projected
dominant height at age 18
24Height Projection Residuals
25Basal Area Projection Residuals
26Green Weight Residuals
27Conclusions
- Adjusting the site index is a reasonable way to
adjust yield models for genetic effects on
average - The response does not fit our other response
models well except for perhaps a response where
the maximum response does not occur until after
18 - There is much variability in such adjustments and
the variability increases with age, but appears
to be well behaved across the range of site
indices - Using actual height data from genetically
improved plots at some inventory age and then
projecting to an older age shows promise for
reducing variability by about half for green
weight
28Conclusions
- In this case each stand has its own adjustment
depending on what the dominant height is on the
plot at the inventory age there is no
adjustment in the model itself - The residuals found in this study point out just
how variable the stands are with very similar
inputs (age, site index, tpa, ba)