|This manuscript revision uses field data from the FIFE dataset|
to evaluate the JULES land model. It does not describe new
model development, but instead compares and evaluates three configurations
of JULES. The stated aim of the paper is to use the FIFE data
to show that none of the model versions can capture the response
of the vegetation to water stress (at least not for the "correct"
reasons). The authors propose to use these results as a
benchmark for future model development.
It is still not clear to me why the authors feel the need to
publish this study at this stage. An evaluation of a model is
clearly a necessary step in the model development process.
But after identifying a poorly represented process, why not
continue the development process and attempt to improve the
process representation? As a modeller, I would have liked
to see some resolution of some of the identified deficiencies,
in particular, the use of the Jacobs ci equation. In my
opinion, the paper would be more valuable to the land modelling
community if some of the identified biases were resolved.
Comment on calibration of Jacobs' ci equation
The authors make the following statement in regards to their
attempt to calibrate the Jacobs' equation for ci:
"We found that the slope of the ci -ca relationship changed as ca
increased... Therefore, we were unable to calibrate the JULES ci-ca
relationship to this data."
They note that the ci/ca relationship shown in figure S8 is
nonlinear, while the Jacobs' equation (eqn 1) is linear. Rather than
address this apparent model shortcoming, the authors instead
attempt to calibrate eqn 1 with the data in figure 6. Why is the
data in figure 6 any better than that shown in figure S8? If
anything it looks less useful; nothing about the data shown in
figure indicates that it could be fit with a linear function.
Later the authors state: "This implies that the Jacobs
parameterisation used in JULES, where the relationship between ci/ca
and specific humidity deficit does not vary over the course of the run,
does not have the flexibility needed to capture the behaviour of GPP
at this site." Why should one expect the slope to vary with time?
Could not the behavior also be explained with a nonlinear ci
Comment on stress function
It would be more intuitive to recast the stress function shown in
figure 1 in tems of theta_crit and theta_wilt only. Given that the
stress function is linear, p0 can be expressed in terms of theta_wilt
and theta_crit. Varying p0 can then be described in terms of
varying theta_crit, which is physically interpretable and can be
compared to other soil properties such as porosity and field capacity.
The authors allude to this fact later in the paper, but I suggest
making this observation up front.
Comment on assumed soil moisture stress
The authors state that Kim and Verna show decreasing leaf water
potential, but Polley et al. show no decrease in assimilation.
Might this indicate that the observed changes in leaf water potential
are not due to soil moisture stress, but simply increasing
vapor pressure deficit? Figure 8 shows that pre-dawn water potential
values are similar for all days, indicating limited water stress.
Given the previously noted inadequacy of eqn 1, is it possible that
the authors may be effectively compensating for the nonlinearity of
the ci(dq) response through their water stress function?
Comment on proposed implementation of leaf water potential
In section 3.3 the authors note that earlier studies used leaf
water potential to modify photosynthetic parameters. Is there
reason to believe that these parameterization are more physically
based than the temperature dependence used in Cox et al.?
The common aspect of temperature and leaf water potential
appears to be similar diurnal variations. Is leaf water potential
actually the key process, or does it simply have the "correct"
variability required to empirically model the diurnal cycle
of GPP etc.?
remove: "We proceed as follows."
dq looks like a differential, consider renaming Delta q or Dq ?