Hybrid modelling of Hybrid modelling of plantation cropsplantation crops

Euan MasonEuan MasonUniversity of CanterburyUniversity of Canterbury

OutlineOutline

Hybrid modelsHybrid modelsCurrent hybrid modelling projects at the Current hybrid modelling projects at the

School of ForestrySchool of ForestryStand level hybrid growth and yieldStand level hybrid growth and yieldHybrid initial growth modelsHybrid initial growth models

Potential for hybrid modelsPotential for hybrid models Geographic Information SystemsGeographic Information Systems

More known about each site and standMore known about each site and stand Variation in growth pattern from site to siteVariation in growth pattern from site to site

Less need for regional modelsLess need for regional models Variation in weather from year to yearVariation in weather from year to year

Predicting the pastPredicting the past Variation in monthly climate offers monthly Variation in monthly climate offers monthly

predictionspredictions Climate change may affect growth patternsClimate change may affect growth patterns Kyoto protocolKyoto protocol

Carbon storage explicit in some modelsCarbon storage explicit in some models

An example “hybrid” modelAn example “hybrid” model

3-PG Model (Landsberg & Waring 1997)

Allocation varies with fertility

Application of 3-PG Application of 3-PG 25-year old experiment25-year old experiment

05

1015202530354045

0 10 20 30

Age (years)

G (S

q m

/ha) Rip, actual

No Rip, actual3-PG, 50 ml ASW3-PG, 120 ml ASW

Potential issues with 3-PGPotential issues with 3-PG

Allocation of C is derived from allometryAllocation of C is derived from allometry Recursiveness, compounded errorsRecursiveness, compounded errors Over parametarisationOver parametarisation Fertility is inadequately representedFertility is inadequately represented Stand and stem geometry are not modelledStand and stem geometry are not modelled CircularityCircularity

DBH->Carbon, Carbon->DBHDBH->Carbon, Carbon->DBHMeasurement of LAI may partially solve thisMeasurement of LAI may partially solve this

Current HM projects at SoFCurrent HM projects at SoF

Physiological hybrid Initial Growth ModelPhysiological hybrid Initial Growth ModelComparisons between hybrid and Comparisons between hybrid and

traditional G & Y modelstraditional G & Y modelsModelling effects of nutritionModelling effects of nutritionModelling effects of spacing and genotype Modelling effects of spacing and genotype

on corewood propertieson corewood propertiesNational hybrid model for National hybrid model for P. radiataP. radiataModelling clonal standsModelling clonal stands

Physiological hybrid IGMPhysiological hybrid IGM

Comparisons of hybrid modelsComparisons of hybrid models

Guy Pinjuv, PhD candidateGuy Pinjuv, PhD candidateCanterbury estate, highly variable sitesCanterbury estate, highly variable sitesComparison of a range of hybridsComparison of a range of hybrids

Traditional growth and yieldTraditional growth and yield3-PG3-PGSeveral alternatives in betweenSeveral alternatives in between

Evaluate fit and utilityEvaluate fit and utilityTwo types of validation requiredTwo types of validation required

Modelling effects of nutritionModelling effects of nutrition

Horacio Bown, PhD candidateHoracio Bown, PhD candidateDevelopment of fertility modifier for hybrid Development of fertility modifier for hybrid

modelsmodelsLong-term goalLong-term goal

Controlled environmentControlled environmentField experimentsField experiments

Corewood stiffness vs actual Corewood stiffness vs actual stocking & clone – J.P. Lasserrestocking & clone – J.P. Lasserre

0

1

2

3

4

5

6

7

8

0 500 1000 1500 2000 2500 3000

Initial stocking (stems/ha)

Stif

fnes

s (G

Pa)

Clone 9 Clone 6 Others

Modelling clonal standsModelling clonal stands Rajesh SharmaRajesh Sharma What changes are needed when modelling for clonal forest What changes are needed when modelling for clonal forest

management?management? Potential for hybrid modelling to provide genotype-specific Potential for hybrid modelling to provide genotype-specific

modelsmodels Large plot clonal experiment - SPBLLarge plot clonal experiment - SPBL

10 clones, pure and mixed clone blocks10 clones, pure and mixed clone blocks Planted 1993Planted 1993 Variation in growth, form, wood qualityVariation in growth, form, wood quality Interchange in dominanceInterchange in dominance Competition with self or other genotypesCompetition with self or other genotypes Local genotype-dependent individual tree modelLocal genotype-dependent individual tree model RiskRisk

National hybrid model for National hybrid model for radiata pineradiata pine

Postdoctoral FellowPostdoctoral Fellow Dr Helge DzierzonDr Helge Dzierzon

Can hybrid modelling remove need for regions?Can hybrid modelling remove need for regions? Are hybrids more effective that traditional Are hybrids more effective that traditional

methods within regions?methods within regions? How much local sensitivity do we lose by How much local sensitivity do we lose by

adopting a national hybrid model?adopting a national hybrid model? PSP data from 13 companiesPSP data from 13 companies Monthly weather data from climate stations, with Monthly weather data from climate stations, with

local offsetslocal offsets Elevation and soils from GISElevation and soils from GIS

Weather stations used Weather stations used

Reasonable data Reasonable data coverage from 1960 coverage from 1960 to 2005to 2005

Stations over all New Stations over all New ZealandZealand

Data availability for Data availability for each station is each station is heterogeneousheterogeneous

Landcare Research Soil LayerLandcare Research Soil Layer

Aim: Investigate Aim: Investigate growth patterns over growth patterns over New ZealandNew Zealand

Format:Format: ESRI polygon layerESRI polygon layer

8 Soil Classes 8 Soil Classes (NZSCORD)(NZSCORD)

67 Subclasses 67 Subclasses (NZSC) covered by (NZSC) covered by PSPsPSPs

Results for MTH using LCR Layer – Results for MTH using LCR Layer – the ultimate regional approachthe ultimate regional approach

Regression for SC:Regression for SC: MSE=2.03MSE=2.03 N: 3202N: 3202 N classes: 26N classes: 26

Residuals for SC Residuals for SC From 20955 to 13182From 20955 to 13182

Results for G using LCR LayerResults for G using LCR Layer

Regression of SC:Regression of SC: MSE=9.70MSE=9.70 N: 3202N: 3202 N classes: 26N classes: 26

Residuals for SC Residuals for SC From 360825 to 302916From 360825 to 302916

Residuals of Residuals of Simulation runSimulation run Mean: 356021Mean: 356021 Std: 9154Std: 9154

An idea among manyAn idea among many Time = accumulated lightTime = accumulated light Use 3-PG type quantum efficiency modifiers to Use 3-PG type quantum efficiency modifiers to

accumulate accumulate potentiallypotentially usedused light light Use sigmoidal difference equations as usual, Use sigmoidal difference equations as usual,

fitted to PSP datafitted to PSP data Avoids some of 3-PG’s problemsAvoids some of 3-PG’s problems

Compounded errorsCompounded errors Allocation of CAllocation of C OverparametarisationOverparametarisation Lack of stand geometryLack of stand geometry

1

2

exp1ln

exp1ln

12

RADk

RADk

a

MTHaMTH

An idea among manyAn idea among many

Estimate genetic components of seasonal Estimate genetic components of seasonal variations in primary and secondary variations in primary and secondary growthgrowth

Different radiation sums for primary and Different radiation sums for primary and secondary growthsecondary growth

Seasonal growth in dbh (red), height (yellow) and Seasonal growth in dbh (red), height (yellow) and 3-PG predictions (blue). (Tennent 1986) 3-PG predictions (blue). (Tennent 1986)

Otago coast

0

5

10

15

20

25

1 2 3 4 5 6 7 8 9 10 11 12

Mohaka

0

2

4

6

8

10

12

14

16

18

1 2 3 4 5 6 7 8 9 10 11 12

Aupouri

0

2

4

6

8

10

12

14

16

1 2 3 4 5 6 7 8 9 10 11 12

Whakarewarewa

0

2

4

6

8

10

12

14

16

18

1 2 3 4 5 6 7 8 9 10 11 12

An idea among manyAn idea among many

Climatic variables as well as stocking and Climatic variables as well as stocking and radiation sum estimates in mortality modelradiation sum estimates in mortality model

To what extent can To what extent can temporaltemporal variation in variation in climatic influences inform us about climatic influences inform us about influences on crop growth and mortality of influences on crop growth and mortality of spatialspatial variation in climate? variation in climate?

An idea among manyAn idea among many

Compatible stand, distribution & individual tree Compatible stand, distribution & individual tree projection systemsprojection systems

Models that represent height vs basal area Models that represent height vs basal area growth as functions of site variablesgrowth as functions of site variables

Models that respond to climatic and local Models that respond to climatic and local weather variationweather variation

Models specific to each siteModels specific to each site Models that naturally provide growth estimates Models that naturally provide growth estimates

within yearswithin years

Preliminary Example – Preliminary Example – P. radiataP. radiata in in Central North IslandCentral North Island

0

10000

20000

30000

40000

50000

60000

70000

80000

Basal area/ha Mean top height (*10)

Res

idu

al S

S

Time

Radiation

Schumacher, daytimetemperature modifier

Von-Bertalanffy-RichardsDaytime temp, VPD modifiers

MTH: Results using light sum MTH: Results using light sum models - Dzierzonmodels - Dzierzon

Regression example Regression example MSE: 1.40MSE: 1.40 N: 2146N: 2146

Residuals:Residuals: From 4427.0 to 4118From 4427.0 to 4118

G: Results using light sum G: Results using light sum models - Dzierzonmodels - Dzierzon

Regression ExampleRegression Example MSE: 5.05MSE: 5.05 N: 2146N: 2146

Residuals:Residuals: From From 57796 to 5482257796 to 54822

Light sum initial growth model: Light sum initial growth model: Competition for light in R1835/2Competition for light in R1835/2

0

1

2

3

4

5

6

7

0 1 2 3 4 5

Time (years)

Hei

gh

t (m

)

No weeds

Weeds

)(0 DRHHT

Future directionsFuture directions

Two fertile strands for forest modelling Two fertile strands for forest modelling researchresearch Internal qualities of stemsInternal qualities of stemsPhysiological hybridsPhysiological hybrids

The University of Canterbury contributes The University of Canterbury contributes to bothto both

Challenge is to bring the two strands Challenge is to bring the two strands togethertogether

Future directionsFuture directions

Hybrids may obviate a regional approach Hybrids may obviate a regional approach to growth and yieldto growth and yield

Hybrids may provide an alternative or be Hybrids may provide an alternative or be complementary to 300 (500) index/Site complementary to 300 (500) index/Site index approachindex approach

Climatic and silvicultural effects may be Climatic and silvicultural effects may be more easily integrated in hybrid models more easily integrated in hybrid models than in traditional G & Y modelsthan in traditional G & Y modelsSite conditions mediate use of lightSite conditions mediate use of light

SummarySummary

Several key studies in hybrid modellingSeveral key studies in hybrid modelling Hybrid modellingHybrid modelling

More sensitive to local site conditionsMore sensitive to local site conditions Seasonal (monthly) predictionsSeasonal (monthly) predictions Reduced need for regions?Reduced need for regions? Better estimates of growth since last inventory?Better estimates of growth since last inventory? Carbon sequestration explicit?Carbon sequestration explicit? Linkages between mensuration, silviculture and Linkages between mensuration, silviculture and

physiologyphysiology