1. Else K. Bünemann1 and Christoph Müller2,3
A 33P tracing model for quantifying gross P transformation rates in soil
Else K. Bünemann1 and Christoph Müller2,3
1 Institute of Agricultural Sciences, ETH Zurich, Eschikon 33, CH-8315 Lindau, Switzerland
2 School of Biology and Environmental Science, Earth Institute, University College Dublin, Dublin, Ireland
3 Department of Plant Ecology (IFZ), Justus-Liebig University Giessen, Germany

2. The challenge Soil P dynamics: dominance of physicochemical processes
SOIL SOLUTION P
(Pi + Po)
MICROBIAL P
(Pi + Po)
INORGANIC P (Pi)
ORGANIC P (Po)
Pools
Sorption/desorption
Microbial processes
Simultaneous to each other and to the physicochemical processes
Soil P dynamics: dominance of physicochemical processes
Main microbial processes: mineralization and immobilization
=> How can we quantify them?

3. N mineralization and immobilization
Mineral N
IMM.
MICROBIAL
and
ORGANIC N t1
MIN. t2
For N:
incubation and extraction (t1,t2);
dNmin(t1,t2) = net N mineralization
15N isotopic dilution methods:
gross mineralization
- gross immobilization
= net N mineralization
For N:
Net N mineralization can be measured by incubation and extraction (adsorption considered negligible)
15N isotopic dilution methods to go further into the processes
Gross min – gross imm = net min
(very simplified; in reality many simultaneous processes, including gaseous losses)

4. P mineralization and immobilization
SOIL SOLUTION P
IMM.
MICROBIAL
and
ORGANIC P
MIN.
INORGANIC P (Pi)
For P:
incubation and extraction (t1,t2);
no accumulation of soil solution P due to buffering
33P isotopic dilution technique:
- assess gross rates (physicochemical and biological)
- derive net organic P mineralization
In most soils no change in soil solution P over time
Need 33P isotopic dilution for gross rates and for net

5. 33P isotopic dilution technique to measure gross and net organic P mineralization
soil + H2O (1:10)
33P
100 min exp, extrapolate specific activity (33P/31P) in the soil solution
=> extrapolated SA (physicochemical processes)
incubate 33P- labeled soils
=> measured SA (physicochemical+biological processes)
net
release of 31Pi to the soil solution
by mineralization of 31Po
gross
Oehl et al. SSSAJ 2001; Bünemann et al. SBB 2007

6. Case study on organic P mineralization
Else Bünemann-König: Summary of research presentation at MLU
Mon Dec 2, 2013
Cambisol, pHH2O 5.5, 44% sand, 22% clay
derived from:
-isotopic dilution method
-C mineralization
Mineral P input
Annual plant P uptake
Microbial P immobili-zation
Net P minerali-zation
Treat- ment
kg P ha-1 yr-1
mg P kg-1 d NK
6.2 b
5.5 a
2.7 a
0.36 ns
NPK 17
16.6 a
2.2 b
0.9 b
0.39 ns
Now to an interesting case study. Long-term grassland fertilization trial on a Cambisol (Braunerde bis Kalkbraunerde).
Similar N and K inputs, different P levels.
Plant P limitation clearly shown.
Thus: even on our soils P deficiency likely in the absence of mineral P fertilizer inputs.
Surprising result: very fast microbial P immobilization rate under P-limited conditions (high-affinity P transporters in action).
Estimate of net P mineralization: high uncertainty because of incomplete extraction.
Different approach: deduce from C mineralization?
Benefit from progress made in data evaluation of 15N isotopic dilution experiments
45 kg N ha-1 yr-1 83 kg K ha-1 yr-1
Uncertainty due to incomplete extraction of microbial P
Bünemann et al. SBB 2012
Bünemann et al. SBB 2012

7. A numerical 33P tracing model
Conceptual model with 5 P pools and 9 P transformations
Transformation rates: zero, first or second order (Michaelis Menten) kinetics
Initial pool sizes and tracer distribution based on measured values
In collaboration with Christoph Müller, Giessen:
Developed P cycle model to simulate experimental data (31P and 33P)
Explain P pools and transformations
Kinetics of transformation rates can be selected, including MM which is relevant for microbial transformations
Initial pools can be measured: e.g. Pif is the P isotopically exchangeable in 24 h, Pof is microbial P.
Müller & Bünemann SBB 2014

8. Model parameter optimization
Markov Chain Monte Carlo method:
«random walk in the parameter space»
Entrapment in local minima avoided, e.g. two parameter space:
=> Probability density function for each parameter (mean and stdev)
Explain random walk:
Improved parameters always accepted
Degraded fit sometimes accepted, sometimes rejected
Explain data output
Müller et al. SBB 2007

9. Observed vs. modelled values
Simulation of experimental data (31P and 33P in Pw and Pof): good agreement between measured and modeled values
Each pool defined by differential equations for change in 31P and in 33P/31P => transformation rates
Agreement between measured and modelled values
Different kinetics tested, best fit evaluated by AIC
Calculation of transformation rates
Müller & Bünemann SBB 2014

10. Conclusions from the modelling approach
=0.77 mg P kg-1 d-1
Net Po mineralization rate (NK) = 0.46 – 0.35 mg P kg-1 day-1
lower than estimate based on C mineralization
0.27
0.19
0.31
0.04
Dominance of microbial immobilization and remineralization over slow mineralization/immobilization
Ratio of microbial to physicochemical processes in this soil (NK): about 1:2.5
0.84
0.0004
0.44
0.29
Allows us to put numbers to the transformation rates
Here: Average over 32 days, but can also show rates for each time step
Net Po mineralization: lower than estimate based on C mineralization, posing questions of mineralised substrates
Microbial processes more important than mineralization of non-living SOM
Microbial processes not negligible! (consequences for E-value concept…)
0.31
=1.88 mg P kg-1 d-1
Müller & Bünemann SBB 2014

11. Outlook
Modelling approach allows application of isotopic dilution principles to non-steady-state conditions
(baseline of extrapolated E-values not needed)
Progress:
application to non-steady-state conditions! (baseline not needed)
e.g. Presence of growing plant, incorporation of plant residues