Developing an integrated terrestrial ecosystem model for

Developing an integrated terrestrial
ecosystem model
for global changing predictions
陸域統合モデルへの結合を念頭にした
植生動態モデルの構築(設計と進捗状況の報告)
Hisashi SATO (FRSGC) & Takashi KOHYAMA (Hokkaido Univ.)
Toward developing the land surface model
Land surface physical
process model
Land surface carbon
cycle model
Vegetation dynamics model
For simulating long time scales,
vegetation dynamics model must be
added to predict changes in vegetation
distribution
原図:伊藤昭彦
*
Feature of dynamics modules within previous DGVMs
HYBRID 3.0
TRIFFID
ロトカ・ヴォルテラ方程式を用いて、大胆なパラメタライズ
を断行。これを、陸上植生の様に本質的に混ざり合うこと
のないシステムに対して適用することは適切ではない。
唯一、個体ベースモデル。ただし、FORSKAと
いう極めて原始的なモデルを使用しており、ま
た植生内の水平方向の地理的ヘテロ性も扱っ
ていない。要するに1 Patchモデルをグリッド毎
に複数走らせ、その平均値をグリッドを代表す
る値として用いている。
IBIS/ LPJ/ SDGVM
水平方向のヘテロ性を無視したAreabased model。各PFTの優先度を葉群投
影被度で表現し、これが1.0を超えた段階
で、PFT間の光を巡る競争が生じる。
Limited computation power inhibit to directly incorporate spatial
hetero-structure of vegetations into the DGVMs.
* Dynamic Global Vegetation Models
However, spatial hetero-structure plays a central role in vegetation dynamics
Gap dynamics
Gap formation
Competition among saplings
例えばPacala et al.(1995)は、光環境を空間的に平均してしまうモデルでは樹種の交代
の様子が変化するだけでなく、総バイオマスも実際の森林の半分くらいになってしまうこ
とを示している。ギャップの下にはとても明るい環境があるはずなのに、これを暗いとこ
ろと併せて平均してしまうことで、ギャップ内での森林の再生が遅れてしまうためである。
この結果は、森林の水平方向の構造を無視してしまうことの危険性を示唆している(竹
中2002 より引用)
Feature of the DGVM (1)
Major advances from the previous DGVMs
*
(1) Individual Based Model (except for herbaceous PFTs)
(2) Explicitly simulate spatial structures of vegetations
A snap shot of the simulated forest stand (30m×30m).
Individual tree is composed of crown, stem, and root.
Shape of crown and stem are approximated by cylinder.
--- Individual characteristics for woody PFTs ---
Crown
Stem
Root
: biomass, diameter, depth
: biomass, height, sapwood & heartwood diameter
: biomass
+ reserve resource for sprouting
* Plant Functional Types
Feature of the DGVM (2)
By explicitly treating forest 3D structure, the model can reasonably
calculate individual light conditions
Estimate light intensity on the top of
the crown by using canopy location
within the forest stand (SORTIE like)
Estimate light distribution
within canopy using leaf
area concentration and
light attenuation index
Estimated light
intensity
Calculate NPP, and
adjust bole height by
perishing deficit
crown layer
NPP
0.0
To avoid ‘edge effect’, this scanning is performed among
replicated forest stands, which surround the examining area.
Feature of the DGVM (3)
Characteristics of herbaceous PFTs
Leaf
Stem
Root
: biomass in a forest stand
: biomass in a forest stand
: biomass in a forest stand
Competition between woody PFTs and herbaceous PFTs
木
本
PFTs
の
定
着
率
Grass layer
Grass layer can only use light on the
forest floor
草本PFTsのバイオマス
Luxuriant grass layer inhibit
establishment of woody PFTs.
Output example (1):
Dynamics of temperate summer-green forest
Current version uses
・tentative modules for daily processes, mortality, and phenology
・parameters, which have not adjusted yet
Output examples (2):
Dynamics of temperate summer-green forest 200 years
4000
1.5
Litter production ( Kg / ha year )
Leaf Area Index ( m2 / m2 )
3000
1
2000
0.5
1000
0
0
60000
250
Biomass ( Kg / ha )
50000
40000
200
Foliage
Stem
30000
Root
150
100
20000
10000
0
50
0
Sum of stem diameter at breast
height ( m / ha )
To see more details of the DGVM ….
Module that comprise the dynamic global vegetation
models, and its computation time steps.
Use modules of Sim-CYCLE
Scheme for connecting phenology module and photosynthesis module
Simulation procedure (1)
Simulation will be conducted on the T42 global grid (128×64), each
of which includes 10 replication forest stands.
Thus, assuming 1/3 of the earth surface is terrene, about 27000
independent forest stand will be independently simulated.
To date, this would be the most complex ecosystem model that have
ever made.
小サイズの林分を複数シミュレートさせる主な理由としては、攪乱の問題があげられる。
例えば寒帯林で頻発する森林火災は、一度生じると、シミュレートしている林分の大きさ
が30×30mだろうが1haだろうが、その殆ど全てが壊滅してしまう。このように機会的に大
きく変動する単一の林分をもって、グリッドの代表値とさせることは適当ではない。
Simulation procedure (2)
Simulation 1 (free seed dispersal)
assumes that all PFTs establish all grid, irrespective of previous
or current vegetation distribution
Simulation 2 (no seed dispersal)
assumes that PFTs that currently distribute for each grid only
establish in the grid permanently
The former simulation should provide maximum estimate of
vegetation change, while the latter should provide minimum estimate.
Procedure for parameter estimation and tuning
(1) Estimate parameters and algorithm of a tree growth
so that tree-form and leaf-density are reasonably simulated for each PFT
(2) Estimate dynamics parameters (Establishment, Mortality, Disturbance):
so that density and age distribution of tree are reasonably simulated when only one
PFT composes the forest
(3) Estimate metabolic parameters (Photosynthesis, Respiration, Allocation):
So that biomass, LAI, and distribution of DBH are reasonably simulated. This will be
conducted on forest that was composed of only one PFT.
(4) By repeating above (2) and (3), convergence parameters
(5) Conduct test run on global grid
then examine that distribution of vegetation and GPP at equilibrium are reasonably
simulated.
Schedule
・ within Oct
Complete to develop the the DGVM program except
daily processes
・ 2 to 4 month
Link it with daily process modules of Sim-CYCLE
・ 1 to 3 month
Parameter estimation and tuning
・?
Vectorize the program and conduct simulation at
global scale on the earth simulator