the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 24 Oct 2025
Gross primary productivity of forest ecosystems in a subtropical city and its decadal climatic and environmental drivers
Abstract. Vegetation plays a vital role in modulating climate and the carbon cycle on land through processes like photosynthesis, also known as gross primary production (GPP). The significant presence of vegetation in Hong Kong, covering over 70 % of the land area, highlights the potential for terrestrial carbon sink to contribute to achieving carbon neutrality in such a metropolitan city. Meanwhile, the terrestrial ecosystem is also influenced by climatic and environmental factors. This study investigates the historical spatiotemporal dynamics of GPP in the subtropical forests of Hong Kong and the key drivers behind its trend and interannual variability between 2002 and 2018. We used the Terrestrial Ecosystem Model in R-Hong Kong (TEMIR-HK), a localized process-based ecophysiological model, to evaluate the changes in GPP induced by changing CO2 concentration, temperature, ozone (O3) concentration, and changing leaf area index (LAI) shaped by these factors as well as land use. Simulation results indicate an increasing trend of GPP, with an average annual GPP of 1.75 TgC yr−1 , which is around 15 % of the annual total anthropogenic carbon emission from Hong Kong, suggesting a limited but indispensable potential of forestry to achieve city-level carbon neutrality. Model simulations of GPP show satisfactory results when spatially comparing with satellite-based GPP dataset (R = 0.89), with slight difference of +8.7 % on average. Factorial simulations reveal LAI changes dominate both trend (+0.0134 TgC yr−2 ) and interannual variability (standard deviation: 2.77×10−2 TgC m−2 yr−1) of GPP in Hong Kong. This result highlights that local-scale reforestation could influence GPP trend over the whole city and emphasizes the importance on the accuracy of LAI input in ecosystem-scale photosynthesis modelling. This work contributes to improving the scientific understanding on subtropical forest ecosystems, and highlights the potential, though limited, of Hong Kong forests to play their parts in working toward carbon neutrality targets.
- Preprint
(10483 KB) - Metadata XML
-
Supplement
(906 KB) - BibTeX
- EndNote
Status: final response (author comments only)
- RC1: 'Comment on egusphere-2025-4647', Anonymous Referee #1, 14 Feb 2026
-
RC2: 'Comment on egusphere-2025-4647', Anonymous Referee #2, 14 Feb 2026
This manuscript applies the TEMIR-HK model to estimate GPP in Hong Kong and attributes its variability to several drivers. While the high-resolution modelling for a complex urban landscape is interesting, the current version has several flaws regarding evaluation independence and the use of MODIS LAI. These issues must be addressed to support the primary conclusions.
Major Comments
- The manuscript primarily evaluates modeled GPP against MODIS MOD17 GPP, which is itself a modeled product strongly constrained by remotely sensed canopy structure (FPAR/LAI). Because MODIS LAI is also a key input in your simulations, the reported spatial agreement may partly reflect shared inputs rather than an independent validation of GPP simulations. Why do use independent data (e.g., FLUXCOM-X, SIF products, or local eddy-covariance data) to validate the model’s performance?
- It is not clear whether any parameter calibration or local constraint was performed for Hong Kong vegetation (e.g., using eddy-covariance site observations, trait databases, or parameter tuning).
- The authors partitioned grid-level LAI into PFT-specific LAI using a “base-year PFT LAI” and a single scaling factor. The appears to assume that all PFTs within a pixel vary synchronously, which is problematic. The authors need to clarify how the "base year" was selected and justify why a single scalar is sufficient to capture the phenological differences between competing PFTs in a mixed pixel.
Minor Comments
L13: Units for trend and interannual variability are not consistent
L30: Typo “Michaelis-Menton”
L166: “base year” is not defined. Specify what year it refers to and how LAIbase,PFT is derived and whether it varies by space/season.
L243: Check the unit for IAV
L296: Typo “Yangtsz River Delta”
Citation: https://doi.org/10.5194/egusphere-2025-4647-RC2
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 272 | 227 | 30 | 529 | 64 | 36 | 34 |
- HTML: 272
- PDF: 227
- XML: 30
- Total: 529
- Supplement: 64
- BibTeX: 36
- EndNote: 34
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Two points on the methodology caused me concern: The first relates to the timescales under consideration: your model (TEMIR-HK) produces hourly simulations of photosynthesis, but your study is addressing decadal trends. Could not the model be driven with longer-temporal inputs? At several points in the text, you discuss photosynthetic responses to temperature but it is not made clear if these are instantaneous or acclimated responses. There is no mention in your submission of acclimation, but this must be an important theme when considering long-term trends.
The second point relates to your finding of a preeminent role for LAI in driving GPP trends. I am not an expert on satellite products, but I think it highly likely that MODIS LAI serves as an input to MODIS GPP through estimation of that fraction of light absorbed by the vegetation (fAPAR) which is a central term of most light use efficiency models. If that is indeed the case, then it can not be surprising that you find such a strong correlation with the ‘observations’. You do concede a weakness in the model approach (L358), but I worry that the problem is much more fundamental.
Minor points:
L19: remove ‘vastly’
L33-34: what is meant by feedback here?
L36-38: somewhere in here you need to distinguish between instantaneous and acclimated responses.
L47-48: try rewording - you are talking about conductance of CO2? How does uptake cause damage? What does ozone do inside the leaf?
L60: but stomatal closure is a plant response to water shortage?
L63-64: unsure what this sentence means.
L135-136: more explanation required here; how are both sunlit and shaded leaves incorporated into a single big leaf?
L136-138: this assumes too much prior knowledge on the part of the reader. And the Farquhar model makes no mention of PFTs.
Table 1: Is there not a case for testing the effect of some metric of water availability e.g. VPD? Or is Hong Kong never short of water?
L179-180: it’s not clear here how those affect your chosen model (Fig 1). Useful to have some explanatory text here.
L186: try rewording. MODIS GPP is adopted as your ‘observations’?
Figure 3: label the plots a, b, c for clarity. It is very difficult to make out the lines in (c). Tell us what each point represents.
L256: The slope of the line in 4c is 0.69 indicating that simulations underestimate??
Figure 5: it is difficult to distinguish the case lines in (a). Even so, I think it would be useful to include the trend of the ‘observations’ (MODIS).
L268-269: I don't follow why the Full and Additive do not agree.
L285-287 and Figure 6: I don't understand the correlation metrics here: the CO2 effect on GPP is flat and yet we have R=0.94?? And how does that flat line reconcile with Fig5b indicating that CO2 is second to LAI in influence?
L326: soil fertility?
L333-334: try rewording, this reads as if the radiation becomes saturated. It is the (GPP?) response that levels off.
L371-373: again the time period under consideration is important.
L379-380: see optimality approaches (e.g. Stocker et al. New Phytologist 2025 Vol. 245 Issue 1) where Vcmax is assumed to be independent of soil N availability.
L404-406: but vegetations differ in their sequestration potential e.g. long-lived trees versus grasses. All green (as seen from space) is not equal in this context.