MOSES calibration for coniferous forest

1
MOSES calibration for coniferous forest

• Chris Huntingford, Peter Cox, Richard Harding..

2
The datasets used

• A key biome in the represented in the Hadley
  Centre land surface model (MOSES/TRIFFID) are
  needleleaf trees. Two such FLUXNET sites are
• Loobos (Holland)
• Tharandt (Germany)

3
MOSES/TRIFFID structure
4
TRIFFID output
5
Parameters to be optimised.

• Four parameters associated with photosynthesis
  f0, D, NL0, a. These affect overall stomatal
  opening, VPD response, leaf nitrogen level
  (influencing temperature response) and light
  response.
• Vcrit Soil moisture content at which vegetation
  stress starts to occur.
• Soil carbon content, Cs (in the GCM this is
  initial diagnosed from pre-industrial control
  simulation and balance between turnover and
  respiration).
• A new light response curve (beyond tuning the a
  parameter)?

6
Optimisation.

• A full optimisation using a NAg search
  algorithm revealed multiple minima with
  associated risk of finding the wrong solution. No
  feel for the effect of changing parameters.
• Decided instead to simply assess the effect of
  manually changing a) the photosynthesis
  parameters, then b) critical soil moisture
  concentration and finally c) the soil carbon
  content.
• Once manual tuning has occurred, then return to
  NAg routines for final phase of optimisation, and
  including uncertainty bounds (E04YCF).
• Once finished, see whether difference between
  optimised parameters (Loobos vs Tharandt) is
  smaller than the uncertainty bounds associated
  with data error.

7
Loobos optimisation

• Left column original light dependence
• Right column new light dependence
• Different rows are sensitivity to soil carbon
  content.

8
Tharandt optimisation

• Left column original light dependence
• Right column new light dependence
• Different rows are sensitivity to soil carbon
  content.

9
Light dependence.

• Current model uses the Sellers (1994?) assumption
  that everything in the canopy scales with the top
  leaf level response (and according to Beers
  law).
• However, this may be failing to get a smooth
  transition between light limited response, and
  other environmental constraints that occur around
  midday.

10
Light dependence.
NEP
Measurements
Temperature limited
Light limited
Model
Time of day
Midday