Land-cover and management modulation of ecosystem resistance to drought stress
Abstract. Drought events are projected to become more severe and frequent across many regions in the future, but their impacts will likely differ among ecosystems depending on their ability to maintain functioning during droughts, i.e., ecosystem resistance. Plant species have diverse strategies to cope with drought. As a result, divergent responses of different vegetation types for similar levels of drought severity have been observed. It remains unclear whether such divergence can be explained by different drought duration, co-occurring compounding effects, e.g., of heat stress or memory effects, management practices, etc.
Here, we provide a global synthesis of vegetation resistance to drought and heat using different proxies for vegetation condition, namely the Vegetation Optical Depth (SMOS L-VOD) data from ESA’s Soil Moisture and Ocean Salinity (SMOS) passive L-band mission and EVI and kNDVI from NASA MODIS. L-VOD has the advantage over more commonly used vegetation indices (such as kNDVI, EVI) in that it provides more information on vegetation structure and biomass and suffers from less saturation over dense forests compared. We apply a linear autoregressive model accounting for drought, temperature and memory effects to characterize ecosystem resistance by their sensitivity to drought duration and temperature anomalies. We analyze how ecosystem resistance varies with land cover across the globe and investigate the modulation effect of forest management and crop irrigation. We compare estimates of ecosystem resistance to drought and heat between L-VOD, kNDVI and EVI.
We find that regions with higher forest fraction show stronger ecosystem resistance to extreme droughts than cropland for all three vegetation proxies. L-VOD indicates that primary forests tend to be more resistant to drought events than secondary forests, but this cannot be detected in EVI and kNDVI. The difference is possibly related to EVI and kNDVI saturation in dense forests. In tropical evergreen broadleaf forests, old-growth trees tend to be more resistant to drought than young trees from L-VOD and kNDVI. Irrigation increases the drought resistance of cropland substantially. Our results suggest that ecosystem resistance can be better monitored using L-VOD in dense forests and highlight the role of forest cover, forest management and irrigation in determining ecosystem resistance to droughts.
Chenwei Xiao et al.
Status: final response (author comments only)
- RC1: 'Comment on egusphere-2023-304', Anonymous Referee #1, 02 May 2023
- RC2: 'Comment on egusphere-2023-304', Anonymous Referee #2, 15 May 2023
Chenwei Xiao et al.
Chenwei Xiao et al.
Viewed (geographical distribution)
This study examines vegetation sensitivity to droughts and heat on a global scale using several satellite-derived proxies of vegetation conditions, including L-VOD, EVI, and kNDVI. To estimate vegetation sensitivity, the authors propose an autoregressive model that incorporates annual drought frequency, annual thermal condition, and the previous year's vegetation states. By comparing the sensitivities to drought and heat across different land cover types, including primary and secondary forests, the authors identify distinct differences in vegetation sensitivities, which they term ecosystem resistance. The study highlights the role of forest cover and land management in shaping the ecosystem resistance to droughts and suggests the advantages of using L-VOD in monitoring vegetation dynamics for dense forests. The research addresses relevant scientific questions within the scope of ESD and provides a novel perspective that considers land cover differences, particularly with respect to the effects of land management practices such as forest management and irrigation. The scientific methods and assumptions are clearly outlined, and the results are well-presented.
One main concern about this study is the ambitious claim of "management modulation" of ecosystem resistance, which may not be straightforward to conclude. The results are based on linear autoregression models, and the management effects are derived from comparing primary and secondary forests that span multiple climate zones. It is possible that the differences observed between primary and secondary forests (e.g., Fig. 5a) are due to climate differences rather than forest management. For example, NEU and EEU, which are both dominated by secondary forests (Fig. 1; Fig. 3a), have contrasting alpha values (Fig. 3b) and different Köppen climate classifications. To better isolate the primary-secondary forest differences and see the effects of forest management, it would be helpful to exclude variations related to other drivers or to group Fig. 1 based on climate zones. Otherwise, the current conclusion that "primary forests, typically associated with higher biodiversity, tend to show stronger resistance to droughts than secondary forests" could be misleading, as the differences may simply be related to ecosystem types shaped by climate rather than forest management.
The current title of the study seems ambitious when using the word "modulation." It may be more appropriate to use a different word, such as "differences," to accurately reflect the findings of the study. The word "modulation" suggests a strong causal relationship, but what we see here are simply differences between primary and secondary forests.
Comments on the analytical methods:
I have several questions regarding the linear autoregressive model used in this study. First, how is it ensured that the droughts used in the first term occur within or before the growing season and affect vegetation growth? Second, is there a justification for using yearly mean temperature instead of yearly maximum temperature in the second term? It would be helpful to either provide relevant references or explain the advantages of using annual mean temperature (e.g., smaller prediction errors compared to using yearly maximum). Third, the third term in the model is incorporated to consider vegetation memory effects, but the study does not present any results related to this coefficient. It would be helpful to briefly mention any relevant findings even if they are not significant. Additionally, it's worth questioning the inclusion of this term in the model if it does not contribute to reducing the overall prediction error. Fourth, for better readability, it would be helpful to mention the "c" term in section 2.4 of the study.
One concern I have is about the explanatory power of each regression over each grid point, given that only 10 values are available for the regression. This could touch a pragmatic lower bound for sample size, and it is therefore important to ensure that the derived coefficients are significant and that their spatial patterns are indicated. For example, in Fig. 2, it would be helpful to show the significance of the derived coefficients along with their spatial patterns. This would allow readers to better assess the reliability of the results and understand the degree of confidence that can be placed in the findings.
Ln 150: Are the land cover classification considering temporal changes? For example, land cover A for year 1 but becomes land cover B for year 2, what would be the eventual land cover for analysis?
Ln 188: It is not clear how the anomalies were standardized.
Ln 26: Up on the improved analysis of primary-secondary forest differences in alpha, this sentence “L-VOD indicates that primary forests tend to be more resistant to drought events than secondary forests” may need to be rephrased.
Ln 27: “EVI and kNDVI saturation in dense forests.”, do you mean for the biomass estimates? Note that EVI is designed to be less susceptible to saturation over dense forest areas (Huete et al., 2002: 10.1016/S0034-4257(02)00096-2).
Ln 39: any reference for the concept “ecosystem resistance”?
Ln 40: Studies related to vegetation recovery and legacy effects have been increasing recently, more latest references are needed for supporting the sentence “recovery trajectory following the disturbance”.
Ln 41: “The mitigation of climate extreme events and maintenance of land carbon sink are highly dependent on the resistance of ecosystems and their changes under land use and land cover change.”: Looks a bit abrupt to come to this sentence, some transition may be needed. Also, please provide references for this sentence.
Ln 55-56: “Taller tropical forests … because …. ”: note that the influence of tree height on the response of tropical forests to drought and subsequent non-drought growth remains controversial. The deep roots of the tropical forest may also play a critical role, check the studies Brando, 2018: https://doi.org/10.1038/s41561-018-0147-z and Giardina et al., 2018: https://doi.org/10.1038/s41561-018-0133-5.
Ln 58: “kNDVI is better correlated …”, compared to which indices?
Ln 63: DVGMs or DGVMs?
Ln: 64-66: “DVGMs and upscaled FLUXCOM GPP have suggested that GPP anomalies are less negative or even positive for pixels including …”: less negative than what? The entire sentence is a bit difficult to understand, good to rephrase it or divide it into several short sentences.
Ln 66: “However, ecosystem fluxes are not directly observable at the ecosystem scale.” the definition of an ecosystem is quite broad and may be good to indicate what the ecosystem scale is referring to here.
Ln 72-74: “For example, modifying forest density and structure by high-intensity overstory removal was tested in conifer-broadleaf mixed forests in Central Europe and considerably increased their growth resilience to droughts and decreased drought-induced mortality by two-thirds (Zamora-Pereira et al., 2021).”: It could be helpful to indicate the findings mentioned here are based on a stand-alone forest gap model.
Ln 101: “VOD has been used 100 as a proxy for biomass”, I guess you meant “aboveground” biomass.
Ln 162: Table 3 can not be found.
Ln 180: Figure 1, what does the white area represent?
Ln 228-229: “L-VOD is positive in Amazon, central Africa and Southeast Asia regions”, difficult to see central Africa and Southeast Asia are positive, are they significant?
Ln 236: clearer compared to what? I can see clearer patterns than those from alpha, but for beta, L-VOD is not as clear as EVI or kNDVI.
Ln 247: Figure 2, double check the unit of temperature coefficients, if it is needed for x axis for (i) and (p)?
Ln 264: what does the black text in Fig. 3a represent, no regression is applied, or zero coefficients?