Review of: Simulating heat and CO2 fluxes in Beijing using SUEWS V2020b: Sensitivity to vegetation phenology and maximum conductance

Author(s): Yingqi Zheng et al.

General Comments:

=================

This manuscript evaluates the simulation results of the SUEWS on radiation flux, turbulent heat flux and CO2 flux at a densely built neighbourhood in Beijing. Using the site-specific gmax and optimized LAI parameters, the modelling of turbulent heat fluxes is improved.

The Fc module of SUEWS is applied in Beijing for the first time, and the simulation results of CO2 flux are satisfactory, which makes it possible to quantitatively evaluate the contribution of various CO2 sources and sinks and facilitate comparison with other observation sites.

However, there are some problems in the analysis and discussion results. For example, the radiation parameterization scheme NARP does not involve parameters gmax and LAI, so there is no way to say “gmax and LAI parameters has only a minor impact on the modelled radiation fluxes”.

Some of the analysis and discussion in the manuscript are relatively simple and one-sided. Please see the following specific comments.

Specific Comments

==================

Line 167-176, Is the data during precipitation included in the deleted data?

In Table 2, is the TL set at -10 ℃ applicable to IAP, and is there any possibility that the air temperature in Beijing will be lower than -10 ℃?

Line 290-291, Generally, air temperature (including daily minimum air temperature, daily range of air temperature, accumulated temperature, etc.) is the main factor affecting urban vegetation phenology in Beijing. Vegetation in cities is irrigated frequently, so there are few cases where vegetation growth is limited by soil moisture, which is different from the non-urban areas Omidvar et al. (2022) studied. The optimization of the LAI model is of course necessary and recommended, but it is not reasonable to explain the factors affecting vegetation growth and senescence rate here.

 Line 295-299, the radiation parameterization scheme (NARP) in the SUEWS does not involve gmax and LAI, so the simulation results of radiation flux among 4 cases should be identical, and the R2 is 1, but why are the RMSE and MBE are not equal to 0?

Line 307-308, In addition to the lower albedo of vegetation in summer, wet surface caused by frequent rainfall and radiation trapping caused by street canyons also lead to the decrease in surface albedo (Ao et al., 2016; Oke et al., 2017; Dou et al., 2019). Given the vegetated fraction is low, only adjusting the albedo for vegetation has a limited effect on improving the simulation results of radiation fluxes as stated by the author. Therefore, it is recommended to adjust albedos for all surface types in SUEWS, as Ward et al., (2016) did, in order to improve simulation results, especially Kup.

Line 313, the overestimated Lup might be induced by the lower emissivity of the building materials but does the Kdown of the reanalysis dataset WFDES also play a role? After all, it can be seen from Figure 5a-d that Kdown is obviously overestimated, especially in summer.

Line 349-351, the Parameterization scheme of Fc does not include the parameter gmax but is related to LAI. At IAP, compared with anthropogenic emissions, the amount of CO2 absorbed by plant photosynthesis is relatively small, so the Fc is not sensitive to the improvement of LAI model. However, this is different from the case where QE simulation results are highly dependent on gmax and LAI. It is inappropriate to simply say that Fc is less sensitive to the improvement of gmax and LAI than QE without further explanation.

Line 432, the NARP does not include the parameters gmax and LAI at all. They are not involved in the calculation of radiation fluxes, so it is not appropriate to say "hardly affected by gmax and LAI".

Line 433-435, For Beijing, plant phenology (leaf expansion time, leaf growth period, defoliation time, etc.) is generally more affected by temperature, not the transformation of dry-wet seasons.

Line 438, Case gmax_LAI improved the simulation effect of Q_E, but I am not sure whether R² increased by 0.02 can be called remarkable better.

Technical corrections/suggestions/language edits (not exhaustive!)

==================

Line 90-91, “while outgoing longwave radiation (Lup) is estimated by a surface emissivity, α, Kdown, Lup and Tair”. The second Lup should be Ldown.

In Figure 5, the shaded area is recommended to be represented by the IQR rather than the standard deviation to display more data information. The same cases are in Figures 6, 8, and 9.

Line 309, that the average seasonal and diurnal cycles of L_down are well captured by the model are shown in Fig.5 i-l rather than Fig.4 i-l.

Line 311, the full name of the NARP (Net All-wave Radiation Parameterization scheme) should be given when it is mentioned for the first time in the submitted manuscript.

Line 469, The cited reference is short of publication year.

In lines 33, Line 43, Line 360, Line 382-383, and Line 421, the arrangement of references is not consistent.

=================

Dou, J. X., Grimmond, C. S. B., Cheng, Z. G., Miao, S. G., Feng, D. Y., and Liao, M. S.: Summertime surface energy balance fluxes at two Beijing sites, Int. J. Climatol., 39, 2793–2810, doi:10.1002/joc.5989, 2019.

Oke, T. R., Mills, G., Christen, A., Voogt, J. A. (2017) Urban Climates. Cambridge: Cambridge University Press. 134-137pp.  https://doi.org/10.1017/9781139016476

This study applies the urban land surface model SUEWS on a neighborhood in Beijing to simulate the energy and CO2 exchange dynamics. A meteorological tower is located in the center of the study area, which provides observations of turbulent heat/CO2 fluxes and the four radiation components. The flux observations are used as reference to evaluate the model simulations and investigate the performance of the model under different parameterizations of the vegetated surfaces, focusing on vegetation phenology and conductance. The study concludes that is very important to adjust the vegetated surface parameterization according to site-specific vegetation information, especially for the accurate estimation of the turbulent heat fluxes.

This study contributes to the literature with insights on how urban vegetation affects the energy balance and the CO2 fluxes at local scale. Such information is still scarce in the literature, especially regarding the CO2 fluxes, and can help improve future model parameterizations and wider applications of urban land surface modeling for climate change mitigation and urban resilience planning.

There are however some problematic and unclear parts in the paper that deserve more attention (see general concerns below). Furthermore, even though the manuscript is in general well-structured and easy to follow, phrasing and grammar can be improved throughout the text. Some examples are given in the specific comments below.

General concerns:

1. I am very skeptical regarding the LAI model optimization method. It is very surprising to see that the Authors have used the MODIS LAI/FAPAR 500 m resolution product which is based on a sophisticated 3D radiative transfer approach to derive LAI. If I am not mistaken, such approach would only be applicable on specific biomes and not on urban areas. The resolution of MODIS is too low to discern the green areas within an urbanized landscape. I would expect that such product would have omitted or at least flagged the areas that are not within its biome specifications. I recommend the Authors to double check the product and its quality flags. It is very probable that the LAI estimations of this product over Beijing are very unreliable.

Moreover, even if the LAI product was reliable, it is anyway challenging to assume that the phenology patterns derived by such a big area (ca. 40 km x 40 km, 6^th ring area) would be representative of your case study (ca. 1 km2). The vegetation types and the management practices would be very diverse across such a huge area.

I suggest that the Authors would revisit their LAI optimization method by using high resolution satellite datasets, such as Sentinel-2 or Landsat, or field observations (e.g. phenocam imagery or field measurements).

2. The part of the paper that presents the model performance regarding the CO2 fluxes is not sufficiently developed. There is a lack of clarity on how the two model adjustments (LAI, gmax) affect the modelled photosynthesis and respiration. There are different ways that such parameters would affect the vegetation and soil processes. In the manuscript is seems that the two parameters have opposite effects on photosynthetic performance. LAI reduction means Fpho reduction, gmax reduction probably induces higher Fpho due to larger soil water content. However, higher soil water content would also induce higher soil respiration, but I understand that this last effect is not included in SUEWS.

More importantly, the modelled diurnal Fc patterns are not matching the observations during summer and to a lesser extent during spring and autumn months. The observed Fc patterns show clear seasonal changes. Morning Fc is decreasing during summer and increasing during winter, while the evening peak seems to be consistent during all seasons. This morning flux seasonal variability is not captured by the model. The Authors claim that the mismatch could be due to an underestimation of photosynthetic performance, but this is not supported by some evidence.

In order to gain a better understanding and interpretation of the results, I suggest that the Authors should do some further analyses: i. examine if the diurnal traffic patterns change seasonally in the study area, ii. examine if there are specific diurnal wind patterns for each season that would affect the observed land cover fractions per season and hour of day, iii. perform an analysis of the observed Fc according to wind sectors to investigate if there are “peculiar diurnal patterns” that would indicate the presence of point sources or wind sectors that are more affected by the green areas. iv. try to find if there are some unaccounted sources in the area (e.g. emissions from commercial/industrial buildings) from some emission inventory (if available).

3. A part of the paper focusses on the modelled radiation fluxes. However, it is not entirely clear how these are affected by the vegetation parameterization in SUEWS. Vegetation phenology and conductance would naturally affect radiation balance by modifying the surface albedo and emissivity over time but also by affecting the upwelling longwave radiation due to the cooling effect of evapotranspiration and the shading. To what extent are these processes directly or indirectly simulated by SUEWS? If they are not involved in the simulations, I wonder if the radiation fluxes evaluation is a relevant part of the manuscript. It is good to report the model performance, but if it is not connected to the study’s main objectives, then it could be moved to an appendix or a supplementary file. Also, a discussion on the model shortcomings in respect to the vegetation effects on radiation fluxes would be relevant.

Specific comments:

Line 2 and throughout the text: the term “sink” is used several times in the text to describe the negative CO2 flux. I believe this term cannot be used to describe the flux sign but to characterize the behavior of an ecosystem in the long term. The right term, as opposed to CO2 emissions, would be “CO2 uptake”.

Lines 6 – 7 and throughout the text: “For the simulation of ….”, “In the model evaluation, ….”. In several places across the text the use of the grammatical article “the” is omitted. I suggest the Authors to have the text revised again for English phrasing and grammar.

Line 24 - 25: sentences unclear, please rephrase.

Lines 31 - 32: recheck the grammar in this sentence.

Line 59: “… imply that the sub-models …”.

Line 64: In the main objectives the Authors state that they aim to evaluate the model under different vegetation parameterizations against radiation and turbulent fluxes. Is the radiation part of the model relevant? See main concern No. 3.

Lines 64 - 65: Using the term “partition Fc” in this sentence implies the use of a top-down approach. However, you do not apply any partitioning of the observed Fc in this study. The different Fc components are modelled separately by SUEWS (bottom-up). Overall, it is hard to assess the modeled contributions of each Fc component just by comparing to the measured net Fc.

Lines 126 – 127: More accurately: Fpho is the CO2 uptake by photosynthesis and Fres is the CO2 release by soil and vegetation respiration.

Line 127, Eq. 10: The negative sign of Fpho is not indicated.

Line 129: Repeated use of “based on”, consider replacing once with “with”.

Lines 133 -134: The descriptions of the terms Ha,h,d and CM are not very clear.

Lines 133 – 146: The units used in the parameters within the anthropogenic emission models are very confusing and do not match in some cases between the text and Table 1. Consider describing the units and the conversions to μmol m-2 s-1 more carefully.

Eq. 13: The term QF,cool is not explained.

Lines 142 – 144: The descriptions of frheat and frnonheat are not complete and this causes confusion. As described by Järvi et al. (2019), these are the fractions of fossil fuels used for heating and other non-heating uses within the study area (i.e. local emissions). I would assume that frheat considers also the fuels used for cooling in the study area.

Eq. 15: Just to be clear, the model does not take into account LAI variability in Fres estimation, right?

Line 156: I suggest you state here that the main model domain is the 1km radius circle around the tower.

Lines 170 - 171: A large fraction of the wind directions are omitted from the analysis. This could affect significantly the land cover fractions “seen” by the observations. Moreover, the wind seasonal and diurnal wind direction patterns are not presented in any way. This information is very crucial when interpreting the measured Fc. The LC fractions of the SUEWS domains can be very different to the actual flux source area defined by the wind patterns and the omitted wind sectors. It would be useful to add the omitted wind directions as shaded areas in Fig. 1c and also include a wind rose in Fig. 1 to give an overview of which directions are affecting more the observations (supplementary to the detailed analyses suggested in major concern No. 2).

Lines 177 – 181: In addition to the criticism on this method (major concern No. 1), this paragraph is not very clear for the reader.

Lines 195 – 197: Is here the right place to describe the method for heat storage flux?

Lines 199 – 200: Not clear. Rephrase.

Lines 202 – 203: I guess it is due to my lack of understanding of the units, but could you give more clear explanations on how these traffic rates are calculated?

Lines 216 – 218: This is an important part to interpret the measured Fc. You state that the emissions from the boiler plants are “very likely” to be observed by the EC, but you do not give any more information of why do you assume that. You also say that you have located at least 3 boiler plants in the surroundings. Could you give more information on how have you located them, where they are (add them in Fig. 1) and what is the height of their chimneys? This information can help to assess if these emissions are actually affecting the observations. Moreover, a wind sector analysis of the observed Fc (as suggested in major concern No. 2) can provide evidence on the effect of point source emissions on the data.

Line 231: The 20 % used as frnonheat sounds an arbitrary choice. Could you explain in more detail how you end up with this number?

Table 1: Should the unit for traffic EFs be kg km-1 veh-1?

Lines 245 – 252: I have the impression that the description of the case base can be much simpler by referring to Table 2 parameters.

Line 255: Table 4 is referred before Table 3.

Table 4: I am not familiar with the LAI model of Eq. 1, but it seems that all the parameters have some physiological meaning. The optimised parameters presented in Table 4 are very different to the original ones (the signs of both ω1 are even inverse) and I am wondering if they still make sense in terms of plant physiology.

Line 290: add comma after “LAI behaviour”

Section 5.1, Figure 3: The seasonal LAI patterns of MODIS and the optimised model are a bit strange. I find it hard to believe that the deciduous species and the grasses present so slow greening phase in spring and summer that they reach full greenness in late July. The original SUEWS patterns, even though they present very steep changes, are more realistic. Could you provide any ground truth to support the optimised model? See also main concern No. 1.

Section 5.2: I wonder if this section is relevant in this study since the vegetation parameterization seems to not have any effect on the model results.

Lines 313 – 314: I am confused by this statement, if the built surface emissivity is underestimated then the LWup would also be underestimated, but the opposite is true in your case. Since the overestimation of LWup occurs mostly during day, it might be that it is indirectly introduced in the model by the Kdown overestimation (Offerle et al., 2003).

Line 322: The performance of the case LAI seems very similar to the case base in QE estimation.

Lines 327 – 333: It could also be that that the opt LAI is underestimated during spring and autumn.

Lines 334 – 339: It would worthwhile to include some small discussion on the effects of the other energy balance parameters in QH performance. QF and ΔQs could as well affect the results of QH.

Line 336: Delete “(SON)”.

Section 5.4.2: The title “model uncertainties” is misleading. This section does not quantify or report model uncertainties but instead it discusses the results of this study compared to previous literature. I suggest to merge this section with the previous one.

Line 377: Plural: “Building emissions are …”.

Lines 400 – 415: This paragraph has a lot of repeating and unclear phrasing, it can be rewritten in a more concise and clear way.

Lines 409 – 410: What do you mean when stating that it is challenging to change the model domain when the soil processes are involved?

Lines 422 – 424: Make clear here that you talk about the local emissions within the study area, otherwise these values would not make any sense.

Line 432: Is this true because the radiation model does not take into account such changes?

Line 433: “… great sensitivity to gmax and the behaviour of LAI”. LAI seems not to be so important.

Line 440: Replace “In comparison of” with “Compared to”.

Lines 445 – 446: This sentence can be omitted from the conclusions.

Lines 453 - 454: Rephrase this sentence to be clearer.

Line 468: The year is missing from Hansen et al.

Line 473 and later: LAI at “tree level” is hardly defined as a concept. LAI is always relative to the ground surface area. When you define seasonal LAImax, LAImin for a vegetation species or type you usually assume a homogeneous area totally covered by this species/type.

Line 485: Singular: “The observed LAI is linearly …”.

Lines 527 – 528: Do you mean that you used the night-time flux partitioning method by Reichstein et al. (2005)? Be more specific.

Line 536: Add “applying a” before “bootstrapping”.

Line 542: Plural: “the species in the modelled area are known…”