Enhanced net CO2 exchange of a semi-deciduous forest in the southern Amazon due to diffuse radiation from biomass burning
Abstract. Atmospheric processes and climate are closely linked to the carbon cycle in the Amazon region as a consequence of the strong biosphere-atmosphere coupling. The radiative effects of aerosols and clouds are still unknown for a wide variety of species and types of vegetation present in Amazonian biomes. This study examines the effects of atmospheric aerosols on solar radiation and their effects on Net Ecosystem Exchange (NEE) in an area of semideciduous tropical forest in the North of Mato Grosso State. Our results show a reduction in the NEE with a decrease in incident solar radiation of ≈ 40 % and relative irradiance between 1.10–0.67. However, an average increase of 35–70 % in NEE was observed when pollution levels (Aerosol Optical Depth) were above ≈ 1.25. The increase NEE was attributed to the increase of up to 60 % in the diffuse fraction of Photosynthetically Active Radiation. These results were mainly attributable to the biomass burning organic aerosols from fires. Important influences on temperature and relative humidity of the air, induced by the interaction between solar radiation and high aerosol load in the observation area, were also noticed; an average cooling of ≈ 3.0 °C and 10 %, respectively. Given the long-distance transport of aerosols emitted by burning biomass, significant changes in CO2 flux can occur over large areas of the Amazon, with important effects on the potential for CO2 absorption on ecosystems of semideciduous forests distributed in the region.
Simone Rodrigues et al.
Status: open (until 08 Jun 2023)
- RC1: 'Comment on egusphere-2023-684', Anonymous Referee #1, 29 May 2023 reply
- RC2: 'Comment on egusphere-2023-684', Anonymous Referee #2, 30 May 2023 reply
Simone Rodrigues et al.
Simone Rodrigues et al.
Viewed (geographical distribution)
The carbon cycle in the Amazon has been directly and indirectly impacted by both climate change and land use/land cover change, but the effect of each driver on net ecosystem exchange (NEE) remains uncertain. In this study, Rodrigues, Cirino and colleagues seek to use a combination of observations from an eddy covariance tower, and aerosol optical depth estimates from in situ and satellites to assess the impact of biomass burning aerosols on the radiation partition (direct/diffuse) and how it cascades into the carbon fluxes in a semi-deciduous forest site in the southern part of Amazon.
The topic is well suited and relevant for Biogeosciences, and the authors have data and analytical tools to provide an important contribution. However, the current analyses have important assumptions that are not clearly evaluated, and potential confounding effects are not addressed or discussed. In addition, at many places in the Results and Discussion section, the authors make statements that are not clearly supported by the results. I list some of these points below. Most of these concerns are fixable, but it will likely require substantial revision of the methods, analyses and discussion.
Most of the analyses presented here assumes that the variability of NEE is driven by gross primary productivity (GPP), and most of the variability in light quality is driven by aerosols. Ecosystem respiration, which had been pointed out previously as the main driver of seasonal variation of NEE in the same area (Vourlitis et al. 2011) is not mentioned as a potential confounding effect. More generally, if the goal of the research is to assess the effect of smoke on diffuse light and NEE, I wonder why the authors did not restrict the analysis to the dry seasons of the study period. That would likely reduce confounding effects due to seasonality (e.g., water stress, deciduousness, ecosystem respiration) and potentially provide better support for most of the assumptions in the derived quantities described in the methods.
Leaf canopy temperature (Section 2.3.7). The model used to estimate this quantity seems to come from Tribuzy et al. (2005), and applied to the focus site in Mato Grosso. Interestingly, Equation 9 does not depend on air temperature. Perhaps this is less of an issue for Manaus, the equatorial site where this equation was originally developed. However, as the authors indicate, Mato Grosso does experience temperature variations from weather systems. Moreover, in Figure 9, the air temperature range is broader than canopy temperature, which suggests that the modelled canopy temperature fails to capture the actual variability. I understand the authors do not have any validation data, but they should consider this limitation when using the canopy temperature estimates.
Clear-sky NEE (Section 2.3.8). I wonder about what this equation is actually capturing. The same SZA may mean different times of day (and temperature and VPD) at different times of the year, so water and heat stress and deciduousness may be adding time-dependent noise on the fitted model (Eq. 10). Perhaps the authors should include month as a factor in their models? Or eliminate seasonal confounding effects by focussing on the dry season only?
Results and discussion section. I found this section difficult to follow, and often found myself unsure on whether the authors were describing their results or previous research. If the authors prefer to keep the results and discussion together (as opposed to separate sections), I suggest reorganising the paragraphs so the distinction between results from this research and the discussion within the broader literature is very clear. For example, in Section 3.1, it was difficult to separate which results were from this study and which results came from Vourlitis et al. (2011). Perhaps focus more on describing Figure 4 and only briefly compare/contrast with the previous results by Vourlitis et al. (2011).
In addition, the authors used NEE and GPP interchangeably throughout most of the manuscript (including in the definition of light use efficiency). Within the same season and during daytime, I can see that this is less likely a problem, but unless I missed this, the authors used the time series across all seasons, which could confounding effects. I understand that GPP estimates from eddy covariance flux towers can be uncertain too, but I think the authors could explain why they opted for analysing NEE instead of GPP. Also, the authors highlight that the forest is semi-deciduous, which made me wonder about the mechanisms that could lead to an increase in NEE through increase in GPP during the dry and smoky season. The assumption of GPP-driven variability in NEE appears multiple times (e.g., L439-L457, L.494-502), so it seems central to the discussion, yet it is not fully supported by the presented data and analyses.
Likewise, the authors discuss the effects of potential confounding effects on the observed relationship between diffuse light and NEE in Section 3.6 and in Figure 9 (e.g., temperature and vapour pressure deficit), but they do not account for these other important drivers in their analyses. They mention in line 508-509 that they could not quantify the effect of these variables in this study, but they do not explain why, and I do not see any reason for not exploring it with statistical models that account for these other variables (similar to the models that they implemented, but with additional predictors). The authors did attempt to mitigate these effects by exploring the response of NEE to diffuse radiation fraction by binning data by solar zenith angle (Section 3.5, Figure 8), but that may have caused the bins to have different times of day grouped together across seasons, so it is difficult to interpret the results.
I found the text to have several typographic mistakes and sentences that appear out of place. I am not listing every one, but I suggest the authors to thoroughly revise and streamline the revised text for clarity.
L1. The opening sentence is a bit circular (atmospheric processes and climate are closely linked to carbon cycle as a consequence of biosphere-atmosphere coupling).
L2. The radiative effects of aerosols and clouds ON XYZ are still unknown…
L5. Relative irradiance: briefly explain “relative” (i.e., relative to which conditions).
L10. 10% increase or 10% decrease?
L17-20. The nature of the debate is unclear, consider briefly explaining it.
L24-25. Mention other sources of CO2 too? Deforestation and degradation (including fires).
L36-39. I found this discussion somewhat misleading due to the significant difference in scale across the studies (Gatti et al. is a regional study, whereas the other references are for specific sites)
L55. The sentence is vague. What are the current limitations are why do these limitations matter?
L57-59. Doesn’t Rap et al. (2015), which the authors already cite, discuss the effects of aerosols on productivity across the Amazon (including Mato Grosso) using numerical modelling?
L76. Remove “105”? It seems out of place
L84–85. The areas presented in this sentence (49.95 km2 and 20.50km2) seem very small for Mato Grosso.
L94. Which systems operate in northern Amazon?
L97. Flush new leaves? “Recover” strikes me as the incorrect word, as deciduousness is an evolutionary adaptation to droughts.
Section 2.2.1. What is the time span of the AERONET data, 1993-2018 (L105) or 1993-2021 (L121).
L143. Drop “in Amazonia” as eddy covariance has been used globally.
L148-149. Drop sentence? This does not seem to add much content.
L155-157. Sentence is confusing.
L170. Temperature should be in K, not °C, for equation 2.
L189. What is Meteoexploration (SolarCalculator)? Provide reference/citation/context.
Equation 4. The notation is somewhat confusing. Perhaps replace the numerator with SWia(t), so it is universal (as opposed to only when AODa > 0.10 and accounting for cloud cover)?
L244-245. The definition of LUE reads as a bit too circular to me.
L281. List the bins used?
L290. Elaborate and briefly describe/provide examples of what were the acceptable levels?
L305. What is the typical pattern of tropical forests?
L326. Statistical difference of what, exactly?
Figures 4–9. The authors often refer to top panel/bottom panel of these figures, but they are mostly side by side. I suggest labelling them with (a) and (b) and edit text accordingly. Also, in many captions, the authors could describe the figures in a bit more detail, and avoid using “correlation” as a synonym of “scatter plots”.
Figure 5. The authors present the binned averages as points, but presumably each bin has a significant variability that should be acknowledged/quantified.
Table 4. Last header column should be statistic.
L346-352. The discussion attributes the variation and increase in PARd to aerosol dispersion, but couldn’t that be partially attributed to clouds too? Presumably the solar zenith angle colours could be telling something on the seasonality, but this is not discussed in the text. Likewise, Table 4 is not really discussed, and I wonder if this is needed in the main text.
L376. Most of this paragraph discusses LUE but no direct link with the results of this study is provided.
L414. Alta Floresta (2 words)?
L415. Where do we see the standard deviation?
L415-417. Either discuss what the readers should get from Table 5 or move it to Supplemental Materials.
L445. Consider replacing “jumped” with “declined”
L485-487. This sentence seems to contradict the text in L474-481, and the authors did not present a clear separation between cloud and aerosol effects on NEE. I suggest dropping the sentence.
Data availability. The authors should consider depositing their code to a permanent archive too. I also found the Ameriflux remark unnecessary, considering that the authors provide a DOI link with the data (although I had to remove the .2 at the end to access it).