The ecology of numerical models and other baked goods Andy ... · The ecology of numerical models...
Transcript of The ecology of numerical models and other baked goods Andy ... · The ecology of numerical models...
The ecology of numerical models and other baked goods
University of Bristol, University of California – Riverside
Andy Ridgwell
OutlineComputer models and
other baked goods
(1) Muffins.
(2) Cupcakes.
(3) Other baked goods.
OutlineComputer models and
other baked goods
(1) Some GENIE results.
(2) Current/emerging GENIE developments and potential for a GUI-based ‘teaching model’.
(3) Future directions in ecological modelling.
introductionComputer models and
other baked goods
GENIEGrid ENabled Integrated Earth system model
www.genie.ac.uk
introductionComputer models and
other baked goods
open oceansea ice coastal seas
terrestrialbiota
atmosphere
marine biota
sed
imen
ts
soils
land surface and hydrology
icesheet
2D energy-moisture balance(no clouds, dynamics)
fully 3D (‘reducedphysics’)
ocean
simplifiedthermo-dynamic
cGENIECarbon-GENIE
www.seao2.org
introductionComputer models and
other baked goods
open oceansea ice coastal seas
terrestrialbiota
atmosphere
marine biota
sed
imen
ts
soils
land surface and hydrology
icesheet
2D energy-moisture balance(no clouds, dynamics)
fully 3D (‘reducedphysics’)
ocean
0 30 60 90 120 150 180 210 240 270 300 330 360
-90
-60
-30
0
30
60
90
36x36x8
175
411
729
1158
1738
2520
3576
5000
Depth (m)
0 30 60 90 120 150 180 210 240 270 300 330 360
-90
-60
-30
0
30
60
90
64x32x8
simplifiedthermo-dynamic
introductionComputer models and
other baked goods
open oceansea ice coastal seas
terrestrialbiota
atmosphere
marine biota
sed
imen
ts
soils
land surface and hydrology
icesheet
2D energy-moisture balance(no clouds, dynamics)
fully 3D (‘reducedphysics’)
ocean
surf
ace layer
bioturbatedzone of 1 cm
sedimentstack layers
partially-filledupper-most
layerb
iotu
rbati
on
al
mix
ing
non-bioturbated
zone ofburied layers
simplifiedthermo-dynamic
major changes vs. ‘GENIE’
deletion of ‘legacy’ science modules, e.g. IGCM, GLIMMER, MOSES/TRIFFID, etc.(improving compiler compatibility)
attempt at parallelization(concurrent GOLDSTEIN/BIOGEM, domain de-compositoin of GOLDSTEIN)
reorganization of main GENIE.F loop and introduction
(completion) of ‘GEMlite’
simplification of runmuffin.sh script configuration, name-
list checking
complete reorganisation of redox (in progress)
continued tracer addition and functionality such as proxy ‘inversion’ methodology
added netCDF restarts for biogeochem modules
added netCDF sedcor data saving
MATLAB plotting and analysis function development
muffinComputer models and
other baked goods
0
Age (Ma)
-5.0
-4.0
-3.0
-2.0
-1.0
0.0
1.0
10 20 30 40 50 60
18
d0 (
‰)
Zachos et al. [2001, 2008]
muffin application: Paleocene-Eocene Thermal MaximumComputer models and
other baked goods
-0.5
Age relative to the PETM (Ma)0.5 1.0 1.5-1.0 0.0-1.5-2.0-2.5-3.0-3.5
4.0
3.0
2.0
1.0
0.0
13
dC
(‰)
PE
TM
(ET
M1)
Zachos et al. [2010]Lunt et al. [2011]
0
Age (Ma)
-5.0
-4.0
-3.0
-2.0
-1.0
0.0
1.0
10 20 30 40 50 60
18
d0 (
‰)
-0.5
Age relative to the PETM (Ma)0.5 1.0 1.5-1.0 0.0-1.5-2.0-2.5-3.0-3.5
4.0
3.0
2.0
1.0
0.0
13
dC
(‰)
PE
TM
(ET
M1)
Zachos et al. [2010]Lunt et al. [2011]
Computer models andother baked goods
Zachos et al. [2001, 2008]
muffin application: Paleocene-Eocene Thermal Maximum
13d C (‰)
To
tal carb
on
rele
ase (
Pg
C)
0
2,000
0 -10 -20 -40 -60-30 -50
4,000
6,000
8,000
10,000
12,000
14,000
16,000
18,000
20,000
mantle CO2 @ -6‰
marine organic matter
thermogenic methane
biogenic methane
13
+6‰
dC
ini
13
-2‰
dC
ini
terrestrial:C3 plants
terrestrial:C4 plants
Contours of carbon release vs. source isotopic signature for a global -4‰ carbon isotopic excursion. Contours differ according to the initial mean
13global d C.
Ridgwell and Arndt [2014]
Computer models andother baked goodsmuffin application: Paleocene-Eocene Thermal Maximum
+ - + 2-CO +H O « H +HCO « 2H +CO2(aq) 2 3 3
pCO2(g) me
tam
orp
his
m,
vo
lca
nis
m (
0.1
)
sil
ica
tew
ea
the
rin
g (
0.1
)
sh
all
ow
wa
ter
C b
uri
al
(0.1
)o
rg
carbonate (0.1), kerogen (0.1)
weathering
calcification (1.1)net CO fixation2
(10)
CaCO dissolution (0.4)3
C oxidation (9.9)org
C oxidation (0.1)org
deep sea CaCO burial (0.1)3
vo
lca
nis
m
low-temperaturebasaltic alteration
CO emissions2
ocean [38000]
terrestrial biosphere [2000]
CaCO dissolution (0.6)3
sh
all
ow
wa
ter
Ca
CO
bu
ria
l (0
.1)
3
atmosphere [560]
Computer models andother baked goodsmuffin application: Paleocene-Eocene Thermal Maximum
1. Calculate model-data errorat a weekly time-step:
too high Þ emit carbon‘OK’ Þ do nothing
(too low Þ remove carbon)
-1emissions (PgC yr )
cumulativeemissions (PgC)
Computer models andother baked goodsmuffin application: Paleocene-Eocene Thermal Maximum
13model d C trajectory
13observed d C
13d C error
time
Earth systemmodel (’GENIE’)
13model d C trajectory
13observed d C
13d C error
time
-1emissions (PgC yr )
cumulativeemissions (PgC)
Computer models andother baked goodsmuffin application: Paleocene-Eocene Thermal Maximum
2. If emissions required:Add CO to atmosphere2
in an Earth system model
CO 2
assume:13d C signature
of fossil fuelsfor emissions
Earth systemmodel (’GENIE’)
-1emissions (PgC yr )
cumulativeemissions (PgC)
Computer models andother baked goodsmuffin application: Paleocene-Eocene Thermal Maximum
assume:13d C signature
of fossil fuelsfor emissions
3. Calculate new atmospheric13
CO d C value in model2
<REPEAT>
13model d C trajectory
13observed d C
13d C error
time
-8.6
-8.4
-8.2
-8.0
-7.8
-7.6
-7.4
-7.2
-7.0
13
atm
osp
he
ric
dC
O (
‰)
2
-6.8
-6.6
200019401920 198019601900
year
310
320
330
340
350
360
370
380
atm
osp
he
ric
CO
co
nc
en
tra
tio
n (
pp
m)
2
300
290
280
-27‰
muffin application: Paleocene-Eocene Thermal MaximumComputer models and
other baked goods
-8.6
-8.4
-8.2
-8.0
-7.8
-7.6
-7.4
-7.2
-7.0
13
atm
osp
he
ric
dC
O (
‰)
2
-6.8
-6.6
200019401920 198019601900
year
310
320
330
340
350
360
370
380
atm
osp
he
ric
CO
co
nc
en
tra
tio
n (
pp
m)
2
300
290
280
-12‰
-60‰
muffin application: Paleocene-Eocene Thermal MaximumComputer models and
other baked goods
0 100 200 3007.2
7.6
7.4
7.8
time (kyr)
year 2100 seawater pH
peak PETM
sur. pH (pH )(sws)
muffin application: Paleocene-Eocene Thermal MaximumComputer models and
other baked goods
0 100 200 3007.2
7.6
7.4
7.8
time (kyr)
sur. pH (pH )(sws)
muffin application: Paleocene-Eocene Thermal MaximumComputer models and
other baked goods
Earth systemmodel (’GENIE’)
-1emissions (PgC yr )
cumulativeemissions (PgC)
Earth systemmodel (’GENIE’)
-1.0
0.0
1.0
2.0
3.0
0 100 200 300
time (kyr)
13sur. d C (‰)DIC
-1emissions (PgC yr )
cumulativeemissions (PgC)
muffin application: Paleocene-Eocene Thermal MaximumComputer models and
other baked goods
0-100
100 200 300
0
-50
Earth systemmodel (’GENIE’)
-1.0
0.0
1.0
2.0
3.0
0 100 200 300
time (kyr)
13sur. d C (‰)DIC
0 100 200 3007.2
7.6
7.4
7.8
time (kyr)
sur. pH (pH )(sws)
-1emissions (PgC yr )
cumulativeemissions (PgC)
muffin application: Paleocene-Eocene Thermal MaximumComputer models and
other baked goods
-1.0
0.0
1.0
0 100 200 300 400 500
Model f
orc
ing &
resp
onse 2.0
3.0
13
su
r. d
C (
‰)
DIC
7.2
7.6
7.4
7.8
su
r. p
H (
pH
)(s
ws
)A
tm.
tem
p.
(°C
)
022
24
26
100 200 300 400 500
28
30
32
muffin application: Paleocene-Eocene Thermal MaximumComputer models and
other baked goods
Time since start of model experiment (ka)
muffin application: Paleocene-Eocene Thermal MaximumComputer models and
other baked goods
Time since start of model experiment (ka)
å (
Eg
C)
15
0-100
100
100 200 300 400 500
Dia
gnose
d c
arb
on e
mis
sions
Em
iss
ion
s-1
(P
gC
yr
)
-0.5
0.0
0.5
1.0
5
0
13
Em
issio
ns d
C (
‰)
DIC
10
50
0
-50
muffinComputer models and
other baked goods
issues
continued use of FORTRAN77 code, e.g. preventing compile-time array dimensioning
netCDF installation issues (partly an issue of the use of C++ code in the netCDF comparison model ‘test’)
linux-only (a problem?) (actually, can also now be natively on a Mac)
non-intuative construction of experiments
limited applicability in teaching due to linux and command-line basis of the model
cupcakeComputer models and
other baked goods
major changes
code management under git (not svn), so has lost its
explicit historical link with ‘GENIE’
conversion to F90 throughout
progress towards all run-time dimensioning of arrays (and no need for re-compilation)
simpler directory structure and job creation/submission
xml removed (and hence no need for python xml libraries (and hence a simpler install))
cross platform support ... can be run under linux/MacOS/Windows
cupcakeComputer models and
other baked goods
to-do thorough performance profiling and optimization
on-line /off-line matrix transport option (eventually replacing GEMlite)
addition of the Darwin ecosystem model
addition of the PALEOGENiE project ‘PAM’ (paleo-assemblage model)
addition of ECBILT AGCM??
addition of JeDi terrestrial ecosystem model??
cupcakeComputer models and
other baked goods
cupcakeComputer models and
other baked goods
to-do GUI development ... ?
PALEOGENiE – motivationComputer models and
other baked goods
Terrestrial weathering inputs
Organic matter burial, scavenging
FeMo
pCO2
(climate)
O2
- +NO , NH3 4
3-PO4
The nature marine ecosystems and strength of biological productivity and remineralization affects:
Oceanic macros nutrient inventories,
esp. P and the form of fixed N.
Ocean oxygenation and hence micro
nutrient inventories, esp. Fe – scavanged in an oxic ocean, and Mo – scavenged in a sulphidic ocean.
Atmospheric pCO and climate.2
PALEOGENiE – motivationComputer models and
other baked goods
The nature marine ecosystems and strength of biological productivity and remineralization affects:
Oceanic macros nutrient inventories,
esp. P and the form of fixed N.
Ocean oxygenation and hence micro
nutrient inventories, esp. Fe – scavanged in an oxic ocean, and Mo – scavenged in a sulphidic ocean.
Atmospheric pCO and climate.2
/end speculation
In turn, changes in the physical and biogeochemial (nutrient) environment will affect ecosystem composition and drive selection.
The approximate coincidence between plankton evolutionary time-scales and the residence time of many of the key ocean and atmospheric tracers raises the possibility of interesting dynamical behaviours of the full system.
Terrestrial weathering inputs
Organic matter burial, scavenging
FeMo
pCO2
(climate)
O2
- +NO , NH3 4
3-PO4
PALEOGENiE – motivation
Gib
bs e
t al. [
2006] (S
cie
nce
)
Fragile Robust Robust Robust Fragile
Originations Extinctions
PETM onset
tim
e‘PETM’
time
PALEOGENiE – motivationComputer models and
other baked goods
time
recovery
‘gap’
‘K-Pg’
65.2 65.3 65.4 65.5 65.6
Time (Ma)
13
dC
(‰
)
1.0
1.5
2.0
2.5
3.0
1210
465
sea ice
heat loss
winds
mixing
eva
po
rati
on
pre
cip
ita
tio
n
Eq
ua
tor
gyr
e
Po
le
coastal ocean
open ocean
deepocean
atmosphere
surfaceocean
strategies for modelling complex (marine) systemsComputer models and
other baked goods
strategies for modelling complex (marine) systemsComputer models and
other baked goods
Creating models is effectively, the art of encapsulation of one’s understanding (or preconceptions) of a system, numerically.
BUT ...
strategies for modelling complex (marine) systemsComputer models and
other baked goods
What happens under climate change?
What did the system look like in the past (e.g. Cretaceous)??
What if the structure of the system is not correctly understood???
strategies for modelling complex (marine) systemsComputer models and
other baked goods
heat loss
winds
mixing
eva
po
rati
on
pre
cip
ita
tio
n
Eq
ua
tor
Po
le
strategies for modelling complex (marine) systemsComputer models and
other baked goods
heat loss
winds
mixing
eva
po
rati
on
pre
cip
ita
tio
n
Eq
ua
tor
Po
le
sea icegyr
e
Ocean circulation becomes an emergent rather than prescribed property of the system.
strategies for modelling complex (marine) systemsComputer models and
other baked goods
phytoplankton
zooplankton
strategies for modelling complex (marine) systemsComputer models and
other baked goods
zooplankton
#2#1
Creating models is effectively, the art of encapsulation of one’s understanding (or preconceptions) of a system, numerically.
BUT ...
strategies for modelling complex (marine) systemsComputer models and
other baked goods
zooplankton
#2phytoplankton #1
str
ain
#1 in v
itro
str
ain
#2 in v
itro
predominantly short-term laboratoryperturbation experiments
model with measured properties of ~2-5 cultured strains encoded
Again:
What happens under climate change?What did the system look like in the past (e.g. Cretaceous)?What if the structure of the system is not correctly understood?
But also:
What about adaptation (or even evolutionary responses) to global change?
strategies for modelling complex (marine) systemsComputer models and
other baked goods
(Ocean) General Ecological Models? (O-GEMs?)
?
strategies for modelling complex (marine) systemsComputer models and
other baked goods
‘in s
ilico’
#1 #2 #3 #4 #n#5 .......
Follo
ws
et al.
[2007] (S
cience
)
Marine ecosystems in silico:
The MIT ‘Darwin’ model typically considered ca. n = 76 randomly-generated trait vectors (’plankton’).
Plankton trait vectors set according to physiological ‘rules’, e.g. larger cells have a higher nutrient limitation threshold, the ability to fixed N comes at the expense of reduced 2
growth rate, etc.
Plankton compete and the ecosystem is an emergent rather than prescribed property.But ...
‘PALEOGENiE’Computer models and
other baked goods
‘in s
ilico’
#1 #2 #3 #4 #n#5 .......
terrestrialweathering
terrestrialbiosphere
se
dim
en
tb
uri
al3D
circulationocean
sea-ice
‘paleoassembledge
model’
Marine ecosystems in silico:
The MIT ‘Darwin’ model typically considered ca. n = 76 randomly-generated trait vectors (’plankton’).
Plankton trait vectors set according to physiological ‘rules’, e.g. larger cells have a higher nutrient limitation threshold, the ability to fixed N comes at the expense of reduced 2
growth rate, etc.
Plankton compete and the ecosystem is an emergent rather than prescribed property.But ...... the geochemical environment and climate co-evolves as global nutrient cycles are modified.
‘PALEOGENiE’Computer models and
other baked goods
‘in s
ilico’
#1 #2 #3 #4 #n#5 .......
terrestrialweathering
terrestrialbiosphere
se
dim
en
tb
uri
al3D
circulationocean
sea-ice
‘paleoassembledge
model’
Marine ecosystems in silico:
n = 1,000-10,000 randomly-generated trait vectors (’plankton’).
Plankton trait vectors set according to physiological ‘rules’, e.g. larger cells have a higher nutrient limitation threshold, the ability to fixed N comes at the expense of reduced 2
growth rate, etc.
Plankton compete and the ecosystem is an emergent rather than prescribed property.But ...... the geochemical environment and climate co-evolves as global nutrient cycles are modified.
At very high resolved diversity, we can explore questions of adaptation and rates of evolutionary change by spawning new plankton with perturbed characteristics.
‘PALEOGENiE’Computer models and
other baked goods
reproductive regime
most happy
would rather be elsewhere
environmental variable (e.g. temperature)
gro
wth
ra
te
fatallyunhappy
‘re
-dra
wn
’, w
ith
sin
ce
re a
po
log
ies,
fro
m S
ch
mid
t e
t a
l. [2
00
6]
‘PALEOGENiE’Computer models and
other baked goods
environmental variable (e.g. temperature)
gro
wth
ra
te
am
bie
nt S
ST
‘Epp
ly cur
ve’
climate warmingclimate coolingIn traditional ‘functional type’ ecosystem models, diversity is not resolved, but instead its effects highly parameterized (e.g. the ‘Epply curve’).
The response to a change in climate is then instantaneous and fully reversible.
‘PALEOGENiE’Computer models and
other baked goods
environmental variable (e.g. temperature)
gro
wth
ra
te
.......sp
ecie
s #
1
sp
ecie
s #
2
sp
ecie
s #
3
sp
ecie
s #
4
sp
ecie
s #
5
Instead, in a highly diverse model, the environmental response of individual ‘species’ can be resolved ...
‘PALEOGENiE’Computer models and
other baked goods
environmental variable (e.g. temperature)
gro
wth
ra
te
.......sp
ecie
s #
1
sp
ecie
s #
2
sp
ecie
s #
2
sp
ecie
s #
2
sp
ecie
s #
2
sp
ecie
s #
2
sp
ecie
s #
1
sp
ecie
s #
1
sp
ecie
s #
1
sp
ecie
s #
1
....... sp
ecie
s #
3
Instead, in a highly diverse model, the environmental response of individual ‘species’ can be resolved ...
... or instead, the capability for adaptation (environmental selection within existing genetic diversity) can be represented(?)
‘PALEOGENiE’Computer models and
other baked goods
environmental variable (e.g. temperature)
gro
wth
ra
te
climate warming
‘PALEOGENiE’Computer models and
other baked goods
environmental variable (e.g. temperature)
gro
wth
ra
te
climate cooling
If climate cools, the low SST optimized species/varients no longer exist. Ecosystem dynamics are presumably different.
Niches are unfilled, so ...
‘PALEOGENiE’Computer models and
other baked goods
environmental variable (e.g. temperature)
gro
wth
ra
te
climate cooling
mutation
Allow non-viable plankton to be replaced with ‘mutations’ of surviving species, using the trait based trade-offs.
Q. How ‘frequently’ to mutate, and as a function of what?
Q. What ‘step size’ to take for mutation?
‘PALEOGENiE’Computer models and
other baked goods
#1 #2 #3 #4 #n#5 .......
terrestrialweathering
terrestrialbiosphere
se
dim
en
tb
uri
al3D
circulationocean
sea-ice
‘paleoassembledge
model’
‘PALEOGENiE’: A radical paleo model-data
concept for theoretically exploring questions of marine plankton adaptation and evolution.
Specific questions:Cause(s) of the delayed recovery (100s of kyr) from end Cretaceous extinctionDetermining which factor(s) best explain ecological responses to PETM carbon release.
A tool for gaining understanding about future ecosystem stability (+ proof concepts for future models).
‘PALEOGENiE’Computer models and
other baked goods
‘in s
ilico’
#1 #2 #3 #4 #n#5 .......
terrestrialweathering
terrestrialbiosphere
se
dim
en
tb
uri
al3D
circulationocean
sea-ice
‘paleoassembledge
model’
Marine ecosystems in silico:
n = randomly-generated trait vectors (’plankton’).......
At very high resolved diversity, we can explore questions of adaptation and rates of evolutionary change by spawning new plankton with
1,000-10,000
There is clearly a very significant computational expense involved, even if using low resolution/efficient Earth system models such as ‘GENIE’.
‘PALEOGENiE’ – computational strategiesComputer models and
other baked goods
60-90
0
90
0120 180
-120 -600
‘Color’ tracer pattern to unambiguously diagnose surface ocean transport
‘PALEOGENiE’ – computational strategiesComputer models and
other baked goods
=> Diagnose full 3D circulation, and employ (sparse) parallelized matrix multiplication.
=> Calculate plankton transport separately from nutrients (and other dissolved tracers)?
60-90
0
90
0 120 180
0-120 -60
Dispersal of a single ‘color’ after 1 year
‘PALEOGENiE’ – computational strategiesComputer models and
other baked goods
=> Diagnose full 3D circulation, and employ (sparse) parallelized matrix multiplication.
=> Calculate plankton transport separately from nutrients (and other dissolved tracers)?
Terrestrial ecosystem modellingComputer models and
other baked goods
A B
Terrestrial ecosystem modellingComputer models and
other baked goods
Use a crop (plant) model based on carbon resource allocation.
Incorporate basic plant biomechanics (60ft long leaves == not a good idea ...).
Terrestrial ecosystem modellingComputer models and
other baked goods
Explore the coupled evolution of terrestrial plants and environment?
Mutate the plants across millions of generations and across millions of ‘fields’ on a massively parallel
computing basis.
Select for yield (but at the field scale, hence dealing with ‘competition’).
Can also select for e.g. water use efficiency, tolerance to gusty wind conditions, etc. etc.
Q. Would an ‘optimal’ crop plant have 6 triangular leaves that enables a hexagonal space-filling
tessellation across the soil surface??
Terrestrial ecosystem modelling ... and cropsComputer models and
other baked goods