Post on 22-Jun-2020
Testing the soil C and N routines for use in JULES in the
QUERCC projectMatt Aitkenhead, Bente Foereid,
Jo Smith & Pete SmithSchool of Biological Sciences, University of Aberdeen, Scotland, UKE-mail: m.aitkenhead@abdn.ac.uk
First JULES Science Meeting, University of Exeter, 28-29th June 2007
Soil C & N routines (Sundial / ECOSSE)
N routines
RothC
DPMk=10 y-1
RPMk=0.3 y -1
BIOk=0.66 y -1
HUMk =0.02 y -1 IOM
co 2RESIDUE
soil surface
Testing the soil C routines
RothC – widely tested for SOCWoodlands:
Calhoun Forest, NC, USA(loblolly pine plantation)
Geescroft Wilderness, UK(natural regeneration from cropland)
RothC – widely tested for SOCCroplands & Grasslands:
Bad Lauchstädt, Germany(crop rotation)
Park Grass, UK(permanent grassland)
Waite, S Australia(grass to crop rotation)
Comparison to other soil C models
RothCCANDY
DNDCCENTURY
DAISYSOMM
ITEVerberne
NCSOIL0
5
10
15
20
25
RMSE
Using the Root Mean Square Error (RMSE)to compare performance, the models fell into two groups with "group a" models having significantly (p<0.05) lower errors than "group b" models.
aa a a a a
b bb
Still to do on soil C…
• Test in a wider range of climatic zones / biomes (ongoing within QUERCC)
• Test how coupled system responds within JULES (ongoing within Hadley Centre Project / QUERCC)
• Improve DOC routines – see next slides
DOC
• Field measurements available only for organic soils
• Sixteen plots, identical starting conditions for model
• DOC routines need more work
New version of the model – DOC
Ostle & McNamara (unpublished data) Organic soil, Lancaster, UK, DOC
The simple implementation of DOC needs more work
0
10
20
30
40
50
60
70
0 50 100 150 200 250 300 350 400 450Days
DO
C (k
g ha
-1)
Testing the soil N routines
Previous testing of N routines
0
50
100
150
200
250
300
04/0
1/19
89
08/0
2/19
89
15/0
3/19
89
19/0
4/19
89
24/0
5/19
89
28/0
6/19
89
02/0
8/19
89
06/0
9/19
89
11/1
0/19
89
15/1
1/19
89
20/1
2/19
89
24/0
1/19
90
28/0
2/19
90
04/0
4/19
90
09/0
5/19
90
13/0
6/19
90
18/0
7/19
90
22/0
8/19
90
26/0
9/19
90
31/1
0/19
90
05/1
2/19
90
09/0
1/19
91
13/0
2/19
91
20/0
3/19
91
24/0
4/19
91
29/0
5/19
91
03/0
7/19
91
07/0
8/19
91
Date
Soil
Min
eral
N /
kg h
a-1
Simulated Soil Mineral N (0-90c m) Measured Soil Mineral N (0-90cm)
Krummbach, Germany, Normal fertilisation, Soil mineral N (0-50cm)
Soil mineral N reasonable at this site
Previous testing of N routines
Eisenbach, Germany, Normal fertilisation, Soil mineral N (0-50cm)
Eisenbach (N4 - normal fert)Potatoes - Sugar beet - S. Barley - S. Barley
Soil N
0
20
40
60
80
100
120
140
160
180
200
01/0
1/88
01/0
3/88
01/0
5/88
01/0
7/88
01/0
9/88
01/1
1/88
01/0
1/89
01/0
3/89
01/0
5/89
01/0
7/89
01/0
9/89
01/1
1/89
01/0
1/90
01/0
3/90
01/0
5/90
01/0
7/90
01/0
9/90
01/1
1/90
01/0
1/91
01/0
3/91
01/0
5/91
01/0
7/91
01/0
9/91
01/1
1/91
01/0
1/92
01/0
3/92
01/0
5/92
01/0
7/92
01/0
9/92
01/1
1/92
Date
kg N
ha-1
Soil mineral N reasonable at this site
Previous testing of the modelSUNDIAL simulated versus observed crop N uptake
0
50
100
150
200
250
300
350
Dat
e
28/1
0/19
88
02/1
2/19
88
06/0
1/19
89
10/0
2/19
89
17/0
3/19
89
21/0
4/19
89
26/0
5/19
89
30/0
6/19
89
04/0
8/19
89
08/0
9/19
89
13/1
0/19
89
17/1
1/19
89
22/1
2/19
89
26/0
1/19
90
02/0
3/19
90
06/0
4/19
90
11/0
5/19
90
15/0
6/19
90
20/0
7/19
90
24/0
8/19
90
28/0
9/19
90
02/1
1/19
90
07/1
2/19
90
11/0
1/19
91
15/0
2/19
91
22/0
3/19
91
26/0
4/19
91
31/0
5/19
91
05/0
7/19
91
09/0
8/19
91
Date
kg N
/ha
Simulated Crop N Observed Crop N
Krummbach, Germany, High fertilisation, Crop N uptake
N supply to plant seemsreasonable at this site
Previous testing of N routines
Krummbach, Germany, High fertilisation, Soil mineral N (0-50cm)
Krummbach (NI - FYM)W. Wheat - Sugar beet - W. Wheat - W. Wheat
Soil N
0
50
100
150
200
250
300
350
01/0
1/88
01/0
3/88
01/0
5/88
01/0
7/88
01/0
9/88
01/1
1/88
01/0
1/89
01/0
3/89
01/0
5/89
01/0
7/89
01/0
9/89
01/1
1/89
01/0
1/90
01/0
3/90
01/0
5/90
01/0
7/90
01/0
9/90
01/1
1/90
01/0
1/91
01/0
3/91
01/0
5/91
01/0
7/91
01/0
9/91
01/1
1/91
01/0
1/92
01/0
3/92
01/0
5/92
01/0
7/92
01/0
9/92
01/1
1/92
Date
kg N
ha-1
Soil mineral N poor at this site
Recent model developments
• Initialisation using RothC• Layering (5cm layers whole profile)• Improved denitrification routines• Incorporation of partial nitrification to
produce N2O and NO• Simulation of saturated soils (above field
capacity)• Change to critical minimum N assumptions
Previous testing of N routines
0
50
100
150
200
250
300
04/0
1/19
89
08/0
2/19
89
15/0
3/19
89
19/0
4/19
89
24/0
5/19
89
28/0
6/19
89
02/0
8/19
89
06/0
9/19
89
11/1
0/19
89
15/1
1/19
89
20/1
2/19
89
24/0
1/19
90
28/0
2/19
90
04/0
4/19
90
09/0
5/19
90
13/0
6/19
90
18/0
7/19
90
22/0
8/19
90
26/0
9/19
90
31/1
0/19
90
05/1
2/19
90
09/0
1/19
91
13/0
2/19
91
20/0
3/19
91
24/0
4/19
91
29/0
5/19
91
03/0
7/19
91
07/0
8/19
91
Date
Soil
Min
eral
N /
kg h
a-1
Simulated Soil Mineral N (0-90c m) Measured Soil Mineral N (0-90cm)
Krummbach, Germany, Normal fertilisation, Soil mineral N (0-50cm)
Critical minimum N wrong
Critical mineral N
• Obvious disparity between modelled baseline mineral N and measurements
• Old model: Ncrit = 5+(0.15 x clay)– OK for silty loam, no good for heavy clay or
organic soils• Related to ‘surface area density’ of soil• Close correlation with clay, bulk density• Ncrit α (clay x BD)?
Critical mineral NCorrelation between critical N in top 25 cm and soil physical properties
y = 0.9564xR2 = 0.8905
0
10
20
30
40
50
60
70
80
0 10 20 30 40 50 60 70
Clay (%) x BD (g cm-3)
N c
rit (k
g ha
-1)
Critical mineral N
• Baseline values now work for– Heavy clay– Sandy loam– Silt Loam– Peaty loam– Organic soils
• R2 = 0.89– R2 = 0.76 for clay alone, 0.57 for BD
0
20
40
60
80
100
120
140
160
180
0 200 400 600 800 1000 1200 1400 1600 1800Days
Tota
l N (k
g ha
-1)
Macdonald et al. (1997) Butts, UK, Oil seed rape - wheat rotation, Soil mineral N (0-50cm)
New version of the model – soil mineral N
New version of the model – soil mineral N
0
50
100
150
200
250
300
35004
/01/
1989
04/0
2/19
89
04/0
3/19
89
04/0
4/19
89
04/0
5/19
89
04/0
6/19
89
04/0
7/19
89
04/0
8/19
89
04/0
9/19
89
04/1
0/19
89
04/1
1/19
89
04/1
2/19
89
04/0
1/19
90
04/0
2/19
90
04/0
3/19
90
04/0
4/19
90
04/0
5/19
90
04/0
6/19
90
04/0
7/19
90
04/0
8/19
90
04/0
9/19
90
04/1
0/19
90
04/1
1/19
90
04/1
2/19
90
04/0
1/19
91
04/0
2/19
91
04/0
3/19
91
04/0
4/19
91
04/0
5/19
91
04/0
6/19
91
04/0
7/19
91
04/0
8/19
91
Measured Soil Mineral N (0-90cm)
McVoy et al. (1995) Krummbach, Germany, Normal fertilisation, Soil mineral N (0-50cm)
Some aspects of model performance are better
Others are worse!
McVoy et al. (1995) Krummbach, Germany, High fertilisation, Soil mineral N (0-50cm)
0
50
100
150
200
250
300
350
0 500 1000 1500 2000 2500Days
Tota
l N (k
g ha
-1)
And some sites are still a complete disaster!!
New version of the model – soil mineral N
New version of the model – soil ammonium
Regina et al. (2004) Cultivated organic soil (potatoes), S. Finland, Soil NH4 (0-50cm)
0
10
20
30
40
50
60
0 200 400 600 800 1000 1200 1400 1600 1800Days
NH
4 (k
g ha
-1)
Soil ammonium appears to be simulated well
New version of the model – soil ammonium
Ute Skiba, unpublished data. Tulloch - organic soil - Scotland Soil NH4
Soil ammonium appears to be simulated well
0
2
4
6
8
10
12
0 200 400 600 800 1000 1200 1400Days
NH
4 (k
g ha
-1)
New version of the model – soil nitrate
Regina et al. (2004) Cultivated organic soil (S. barley), S. Finland, Soil nitrate (0-50cm)
time (weeks)
0 20 40 60 80 100 120 140 160
nitro
gen
(kg/
ha)
0
20
40
60
80
100
120
simulated NO3- simulated NH4+ measured NO3- measured NH4+ fertiliser application
Modelled by B. Foereid
N2O & CH4
• Soil C and N routines are getting there• CH4 and N2O emissions need further
parameterisation – this is ongoing work in CarboEurope, ECOSSE and NitroEurope
Ongoing work
• N2O• CH4• Characterisation of other vegetation types• Data acquisition (Great Plains)• Test for other biomes and climate zones• Refine and improve the model for different
soil types and conditions
In Sundial / ECOSSE, emissions of N are simulated due to denitrification and partial nitrification, the emissions are then partitioned into emissions of N2 and N2O.
Sundial / ECOSSE will be modified by cloning some of the subroutines from DayCent model to simulate NO emissions. Influce of Snow melting on NO emissions will be added.
Sundial / ECOSSE will be calibrated in some of the core sites ofNitroEurope IP using Bayesian Calibration method to reduce the parameter uncertainty.
Sundial / ECOSSE will be calibrated, evaluated and reformulated for simulation of NO and N2O emission estimates from site scale to regions scale in ongoing Nitroeurope IP project.
An example – improved N2O routines
Work by Jagadeesh Yeluripati