the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Getting the leaves right matters for estimating temperature extremes
Gregory Duveiller
Mark Pickering
Joaquin Muñoz-Sabater
Luca Caporaso
Souhail Boussetta
Gianpaolo Balsamo
Alessandro Cescatti
Abstract. Atmospheric reanalyses combine observations and models through data assimilation techniques to provide spatio-temporally continuous fields of key surface variables. They can do so for extended historical periods whilst ensuring a coherent representation of the main Earth system cycles. ERA5, and its enhanced land surface component ERA5-Land, are widely used in Earth System science and form the flagship products of the Copernicus Climate Change Service (C3S) of the European Commission. Such land surface modelling frameworks generally rely on a state variable called leaf area index (LAI), representing the amount of leaves in a grid cell at a given time, to quantify the fluxes of carbon, water and energy between the vegetation and the atmosphere. However, the LAI within the modelling framework behind ERA5 and ERA5-Land is prescribed as a climatological seasonal cycle, neglecting any inter-annual variability and the potential consequences that this uncoupling between vegetation and atmosphere may have on the surface energy balance and the climate. To evaluate the impact of this mismatch in LAI, we analyse the corresponding effect it has on land surface temperature (LST) by comparing what is simulated to satellite observations. We characterise a hysteretic behaviour between LST biases and LAI biases that evolves differently along the year depending on the background climate. We further analyse their repercussion on the reconstructed climate during the more extreme conditions in terms of LAI deviations, with a specific focus on the 2003, 2010 and 2018 heatwaves in Europe where LST mismatches are exacerbated. We anticipate that our results will assist users of ERA5 and ERA5-Land data to understand where and when the larger discrepancies can be expected, but also guide developers towards improving the modelling framework. Finally, this study could provide a blueprint for a wider benchmarking framework for land surface model evaluation that exploits the capacity of LST to integrate the effects of both radiative and non-radiative processes affecting the surface energy.
- Preprint
(7753 KB) - Metadata XML
- BibTeX
- EndNote
Gregory Duveiller et al.
Status: final response (author comments only)
-
RC1: 'Comment on gmd-2022-216', Anonymous Referee #1, 27 Feb 2023
Review of
Getting the leaves right matters for estimating temperature extremes
by Duveiller et al.
General comments:
This paper shows that the ERA5-Land product should be used with caution and that the ERA5-Land production chain should use a more modern approach for representing vegetation or include satellite-derived LAI into their LSM. I am not sure this paper has much practical value from a modelling point of view. "Getting the leaves right matters" is perfectly right but far from being a new requirement. The added-value of this work should be better explained. Interesting recommendations are given in the Discussion section. For example, the recommendation to use hourly LST data from geostationary satellites. It is well known that solutions to integrate LAI into LSMs do exist. They are not mentioned and not used in ERA5-land. Could ERA-Land incorporate interactive LAI at some stage? Assimilation of LAI observations? Why not using another more advanced LSM forced by ERA5 atmospheric variables? The joint use of LAI and LST is interesting and offers a good benchmarking framework for assessing model performance. However, this paper may give the wrong impression that LST biases are completely explained by LAI. Other factors include the absence of representation of irrigation, snow misrepresentation, altitude solar radiation bias in mountainous areas, and slope effects in complex terrain. Overall, the paper is well written and a few changes could be sufficient to address my remarks.
Recommendation: minor revisions.
Particular comments:
- L. 40: I don’t understand well the LAI definition used by the authors: “LAI is defined as half of the total green leaf area per unit horizontal ground surface area”. Why half? Because only one side of the leaf is counted? For the sake of clarity, I would recommend using this more precise definition: “LAI is the one-sided green leaf area per unit horizontal ground surface area”.
- L. 152 (GEOV2/AVHRR): The THEIA LAI data portal web page should be given. Not only the LTDR web page.
- L. 202 (GLEAM): Which satellite date are used in this version of GLEAM? Is LST used for example?
- L. 241 (“darker than”): I am not sure this can be considered as a general rule. At wintertime, wet soils might be darker than senescent vegetation. Is this represented in the model? Adding a reference showing what is occurring in the real world would be useful.
- L. 266 (Fig. 4): Figure 4 interpretation is not straightforward. This Figure does not show much more than Fig. 3 and has too many tiny sub-figures, difficult to read. The complete Figure could be moved to a Supplement and a selection of meaningful sub-figures could be left in the paper. Why not plotting the two HI and BD indices in a SM - Temperature space instead? I.e. replace Fig. 4 by the two left sub-figures of Fig. 5?
- L. 311 (Fig. 6): Connecting maps to the colour scale is difficult. I suggest plotting only 3 colour classes: one for significant positive correlation, one for significant negative correlation (significant meaning p-value < 0.01), and white for non-significant correlation.
- L. 324 (Fig. 7): The top Cfb subfigure is not readable (dark green is difficult to distinguish from the dark gray color used for ocean surfaces).
Citation: https://doi.org/10.5194/gmd-2022-216-RC1 -
AC1: 'Reply on RC1', Gregory Duveiller, 07 Jul 2023
We thank the reviewer for the comments and recommendations. We will try to take them fully into account in the revised manuscript. Below we respond first to the general comment and then to the particular comments evoked at specific lines in the text.
We tend to disagree to the comment that this work has little practical value from a modelling perspective. We see it as a study providing a new diagnostic of how to assess model performance (in-line with suggestions of topics from GMD). As mentioned by the reviewer, we agree that “the joint use of LAI and LST is interesting and offers a good benchmarking framework for assessing model performance.” We believe the work should be judged in this light. We will try to clarify the added value of this work in the revised text, in-line with what is also suggested by the second reviewer.
Following the present reviewer’s suggestions, we can elaborate on why LAI is not assimilated in the ERA5 model in the revised text. One of the main strengths of the ERA5 reanalysis is its consistency and temporal depth, going back to the 1950s. If LAI would be assimilated, it would have to be done all along the archive, which is not possible because we do not have any consistent observations of LAI before the satellite era. There could be solutions to work around this constraint, such as synthetically generating an LAI archive that is consistent with ERA5 (which is perhaps what the reviewer is suggesting using another LSM). We will discuss this further in the text.
FInally, regarding the remark that this study may give the wrong impression that LST biases are completely explained by LAI, we do agree that this could be misinterpreted. We will thus clarify this in the discussion and mention the other factors mentioned by the reviewer (i.e. absence of representation of irrigation, snow misrepresentation, altitude solar radiation bias in mountainous areas, and slope effects in complex terrain).
L40: The reason for this more convoluted formulation of the LAI definition is that it aims at encompassing all types of leaves and not just the more typical flat leaves. As explained in Yao et al. (2019), the simpler definition employed by the reviewer is akin to that by Watson (1947): “leaf area (one side) per unit area of land”. While this definition is suitable for flat leaves, it is more problematic for curly, concave or needle-shaped leaves. For non-flat leaves, such as conifer needles which are generally tubular, there is a distinction between half of total leaf area (i.e. half the surface of the tube) and the intercepted area (the cross-section of the tube). This is not a problem for flat leaves because these two concepts are the same, but the can be very different for non-flat leaves. This is why the more awkward definition tries to compensate for this by referencing everything to the total leaf area, which is proportional to the number of stomata for instance. We can add a phrase in the text to clarify this point.
L152: Thank you for pointing out that we referred to the LTDR more explicitly than the THEIA. We fully agree that it should be the other way around (directions to LTDR are not even very useful in this case) and we will make sure to fix this in the revised version.
L202: Regarding the satellite inputs of GLEAM, the reference Martens et al. 2017 provide the list of inputs (in table 1 therein), which includes Soil Moisture, Radiation, VOD and air temperature, but it does not include LST. We can add a note on this in our text.
L241: We agree that wet soils could indeed be darker than vegetation, especially if this vegetation is senescent. Overall, we did not want to go into too much detail about this question of when and where the vegetation is darker than the background, which is why we wrote that “generally” it is darker. The comment we made refers more specifically to the fact that we see that on the one hand, in winter, the model always simulates more leaves, and on the other, we see a warm bias in the LST for Northern areas that are very likely covered with snow at some point of the winter. In such cases (i.e. too much simulated vegetation above snow), it is very likely that this would lead to a warmer surface. We can change the text to be more explicit that we refer more specifically to this case (vegetation over snow) to avoid confusion with the case of dark moist soils. It is useful to note that the albedo of the model in snow-free conditions is based on a climatology of MODIS albedo observations, in-sync with the climatological LAI. It will therefore “see” if a soil is overall darker because it is generally (climatologically) wet, but it will not be sensitive to dynamic changes in soil moisture. However, the changes in albedo due to changes in snow is “prognostic”, i.e. it changes along with snow cover as modelled in the re-analysis system (i.e. in-sync with dynamic weather). We can add a reference to the modelling documentation to clarify this.
L266: We agree that Figure 4 is a bit complex. We liked the idea of having it to illustrate the variability of situations that could be encountered, but we take well the recommendation of perhaps moving it to supplementary material. We could contemplate adding 1 or 2 more sub-figures to Figure 3 to illustrate more cases, and leave the full Fig 4 in the annex. Regarding the point of splitting Fig 5 and use half to replace Fig 4, we do not see that as necessary. We would prefer just to remove Fig 4 and simply replace it by the current Fig 5.
L311: We have the impression and here on Fig. 6 the connection between map and colour scale is not difficult in this particular case, as the colourscale is clearly divergent both in colour and intensity. Keeping the continuous colormap seems also more ideal so that a hard (and possibly subjective) threshold in p values is not used. We could propose a second panel on the side with such categories to complement, although it is unclear (to us) how much more information this will bring.
L324: We agree that the contrast is not ideal. We would want to keep the original colours of the Koppen-Geiger classification (which are widely used) and therefore we propose to adjust the background colours and thereby improve the contrast.
Citation: https://doi.org/10.5194/gmd-2022-216-AC1
-
AC1: 'Reply on RC1', Gregory Duveiller, 07 Jul 2023
-
RC2: 'Comment on gmd-2022-216', Emanuel Dutra, 02 Jun 2023
The manuscript aims at assessing the impact of LAI errors and missing inter-annual variability in the ERA5-Land reanalysis land surface temperature (LST). This is a very relevant study for users of the reanalysis, but more generally to model development as it clearly identifies several limitations of the reanalysis LST which can be associated with errors and missing variability in LAI. The methodology is very robust, in particular (i) the focus on the hysteresis behavior between LAI and LST biases and (ii) the focus on several recent European heatwaves. However, there are some points in the methodology and results discussion that could be clarified and are listed in the following points.
Main comment:
In this study LAI is presented as a key variable that could explain some of the errors and it is also suggested that the assimilation of LAI could reduce some of the biases. I think that this is a simplification of a very complex problem, and deserves a more detailed discussion. Land surface models are a set of parameterizations with different degrees of complexity and uncertainty. In the case of LST, it depends on the turbulent exchanges and ground heat flux. There we can identify 3 key components: (i) coupling with the underlying soil temperature; (ii) aerodynamic resistance and (iii) canopy resistance. It is expected that LAI modulates both the aerodynamic resistance (e.g. increasing drag/turbulence when LAI is higher) and canopy resistance (increasing this resistance when LAI is higher). In the case of ERA5-land only the canopy resistance is modulated by LAI linearly and normalized by the minimum canopy resistance. In other models this relation is different, and other models also account for the LAI in the aerodynamic resistance (e.g. roughness lengths in SURFEX). The message that I think it would be important to be clear is that LAI is not directly used by the model, but used as a predictor of some parameters in some parameterizations. This study makes a significant advance to our knowledge by clearly showing the relation between vegetation status (via the LAI) and the LST biases, and in my opinion, suggests the need to revise both (i) the actual LAI data used (including assimilation) and (ii) the way that the model uses this data in the different parameterizations. My suggestion to the authors, if they agree, is to have some of this discussion in the paper (discussion and conclusions).
Results & discussion clarifications
Section 2.3: Selection of 5 warmest days: The selected days from MYD11A1 and ERA5-Land are the same, or where are the 5 warmest days selected independently from each dataset ? In the discussion “A possible improvement could be to work with daily values and select explicitly days in ERA5L that have clear-sky condition” suggests that the days are not the same. Please clarify. If the days are not the same, I see that this could be a limitation in the interpretation of some of the results, as they could be affected by the clear-sky bias.
Lines near 240: “... LAI during winter, and this corresponds to an overestimation of LST by ERA5L in the northern latitudes. This relationship between the bias in LST and the bias in LAI is consistent for such energy limited situation where biophysical effects of vegetation on climate are dominated by radiative effects. In fact, the modelling framework assumes there is an excess of leaves covering the background, the former being generally darker than the latter (especially when the background is covered with snow) resulting in more heat accumulation than what would be observed in a situation with fewer leaves.”
The authors suggest that the northern hemisphere winter warm bias could be related with excessive LAI. I have some concerns with this hypothesis :
- from a previous comment, are the 5 warmest days in each month the same in MYD11A1 and ERA5-Land ? In winter with snow conditions clear-sky conditions could be colder than cloudy-sky (stronger radiative cooling)
- The bulk snow scheme in ERA5 and ERA5-land is known to have a warm bias due to a large thermal inertia which is partially addressed with a multi-layer snow scheme (see https://agupubs.onlinelibrary.wiley.com/doi/full/10.1029/2019MS001725 )
- The results in Figure 3b also show a rather constant LAI bias in winter with a wide range of LST biases.
Therefore I suggest the authors revise these arguments regarding the warm bias of LST in winter in the northern hemisphere, also in lines 260:265 (discussion of Figure 3b).
Figure 6 : inter-annual correlations: From my understanding the correlations were computed for each point using 16 values (years). With this sampling a correlation of 0.5 will have a p-value near 0.05. Therefore only correlations above 0.5 (simple approximation here) should be considered to avoid miss-interpretation due to the very small samples. I recommend to show this in the figure, e.g. shading non-significant correlations, and treat it in the discussion with caution.
Line 307: “In other words, the years when the seasonally prescribed LAI of ERA5L is further from the reality, e.g. in years where the LAI peak is lower or shifted due to particular growing conditions of that year, the underestimation of LST can be expected to be more severe.” I recommend caution on the interpretation of these results. The correlations show that in years with negative anomalies of LAI ERA5-Land has a positive LST bias. The LST warm bias can be due to the missing LAI anomaly, but such LAI anomalies are also likely reflected on other aspects of the land surface, e.g. drier soil moisture, changes in albedo, changes in surface roughness. But these results are very relevant, as they show that ERA5L LST has larger errors during extremes, which are key in our understanding of the present and future climate.
Results and discussion in Figure 7: By mixing all seasons I find it hard to interpret the results. Considering the case studies focus in summer, would it be easier to interpret these results if only summer months were used ? I leave this to the author's consideration.
375: “warning for users of both ERA5 and ERA5-Land that the magnitude of heatwaves in particular may be underestimated in these datasets” This is the case if a user uses LST to define/evaluate the heatwave event. However the most common approach in the literature is to evaluate 2-meters temperature. Please clarify here that these results only apply to Land surface temperature.
376: “assess the sensitivity of LAI to temperature, not to…” To avoid misunderstanding, I suggest to change to “assess the sensitivity of LAI to land surface temperature, not to”
Line:408 “ This was currently not done because the values provided in ERA5L only represent snow-free albedo, which do not reflect the same reality as the MODIS albedo covering all conditions.” A clean way to compute ERA5L albedo and compare with MODIS would be to compute it from the surface net solar radiation and Surface solar radiation downwards, which are available in the CDS for ERA5-Land . I don’t think it would make a large difference and be necessary to be done for this study, but could be mentioned in the discussion.
Citation: https://doi.org/10.5194/gmd-2022-216-RC2 -
AC2: 'Reply on RC2', Gregory Duveiller, 07 Jul 2023
We thank the reviewer for the very insightful comments and we will strive to incorporate the suggestions as best we can in the revised manuscript.
We take good note that we may have given the impression that the solution lies “simply” in assimilating LAI, while this is indeed a much more complex problem. We fully agree with the reviewer that this would deserve more details in the discussion. As suggested, we will emphasize how LAI is not directly used by the model, but rather used as a predictor of some parameters in some parameterizations, and we will mention how LAI modulates both the aerodynamic and canopy resistance. We believe this discussion will overall enrich the general overview of the problem by the reader.
Below we further clarify the different specific points brought out by the reviewer.
Regarding the selection of LST values from MYD11A1 and ERA5-Land described in Section 2.3, we confirm the intuition of the Reviewer that they are not the same. We thought it was clear but we will ensure to make it more so. The choice of selecting the 5 warmest days is based on the assumption that there would generally be at least 5 clear-sky days in the month, and that these would likely be warmer at noon as more incoming radiation would heat up the land. We agree that this will not always be the case, and that this could, in some occasions (such as overcast winters where ERA5 might have higher surface temperatures simulated under the clouds), cause misinterpretations (which is why we wrote that phrase suggesting an improvement. The reason behind this choice is pragmatic, as it easily allows to get comparable values for most cases from both datasets in an easy way. The alternative would be to only select from ERA5 the clear sky days, but this would be more computationally complicated and would also not guarantee 100% the same days would be picked, as it implies that ERA5 currently estimates the clear-sky fraction. This is why we choose to stick to our simpler and more pragmatic approach, but we will keep the suggestion for improvement for further studies.
We take note of the remark related to the part around lines 240, as well as in lines 260-265, regarding the interpretation of the warm bias of LST in winter conditions. We will be more cautious in suggesting this is linked to excessive LAI and will nuance this with the uncertainties linked to our simple approach to deal with the clear-sky bias. We will also acknowledge the presence of the snow bias mentioned by the reviewer.
Regarding Figure 6, we agree that given the small sample size there is ground to be more prudent and shade out those where p-values are not appropriate. This is also in-line with the remark of Reviewer 1, and we will thus revise the figure.
L307: We take the Reviewer’s point that some caution is warranted in this interpretation. We will elaborate by stating how the anomalies in LST may be due to LAI underestimations, but also by the other effects mentioned by the reviewer (drier soil moisture, changes in albedo, changes in surface roughness). We will also stress, as suggested, that nothing this variability in LST bias is important as it indicates these are worst under extremes, which is a nice segway to the later part of the study on heatwaves.
For Figure 7, we agree that mixing all seasons together complicates a bit the interpretation. We will try to do some version with only summer months and judge if this relays better the message.
L375: We agree, we will clarify this remark is only for LST and we will further stress how heatwaves are generally defined by air temperature.
L376: We agree to the change
L408: We agree to the suggestion and will mention this in the discussion as a possible improvement
Citation: https://doi.org/10.5194/gmd-2022-216-AC2
Gregory Duveiller et al.
Data sets
Dataset in support of the study "Getting the leaves right matters for estimating temperature extremes" Mark Pickering & Gregory Duveiller https://doi.org/10.5281/zenodo.6976942
Model code and software
Code in support of the study "Getting the leaves right matters for estimating temperature extremes" Gregory Duveiller & Mark Pickering https://doi.org/10.5281/zenodo.7275088
Gregory Duveiller et al.
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 736 | 213 | 21 | 970 | 10 | 10 |
- HTML: 736
- PDF: 213
- XML: 21
- Total: 970
- BibTeX: 10
- EndNote: 10
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1