Reader Comments

Post a new comment on this article

Response: Not so fast

Posted by moracamilo on 05 Sep 2015 at 00:42 GMT

Here we respond to some criticisms raised about our paper. Although comments allow for additional clarification of our paper, such comments do not change the results of our paper.

Comment by Ed Hawkins

Comment #1: Mora et al. only quote and use the multi-model ensemble mean, stating that it is more accurate than most of the models individually. Yet the authors have no basis for assuming that past performance is any guide to future performance. There are good reasons why the multi-model mean may look most like the observations in the past, but the models produce a range of outcomes for the future, and this range cannot be discounted as Mora et al. have done. They have not quoted any uncertainty ranges in the text, yet their Fig. S3 highlights that the spread amongst the models is as large or larger than the signal in many, if not most, locations. Their use of CMIP5 simulations in this way is in sharp contrast to standard practice and what the IPCC recommends. Not quoting an uncertainty range is indefensible.

Response: Our paper provides detailed results of model accuracy (i.e. how models predict current suitable days) and precision (i.e., difference among models in predicting suitable days). These results were presented and described in extended supplements of the paper (S2-S3). Further, the underlying data of such supplements were released in Excel sheets for anyone to check.

Comment #2: The individual models have biases in their simulated present-day climate. For example, temperatures may be too warm or too cold in particular regions. Indeed, simulated global mean temperature is up to 3C different across the CMIP5 models. This has critical implications for using the raw model output to determine the number of days above or below certain absolute thresholds. If these biases are not accounted for (as seems to be the case) then the number of days calculated is incorrect. Similarly, differences in the daily variance between the observations and simulations is also critical for the number of days above or below thresholds and would also potentially need to be corrected. There are well studied ways of trying to account for these biases, which appear to have been ignored.

Response: There are indeed well known biases about how models predict current conditions, and some papers try to account for such bias by quantifying the bias from modern conditions and then removing it from future projections. The problem of controlling this bias is that it is based on the assumption that the magnitude of the bias will be the same for the same place in the future. It is well known that such a biases results from model uncertainties, and that they more likely result from poor modeling of climatic tele-connections. So the underlying assumption to control the bias may not be entirely true, and thus we will be controlling an error with potentially another error. So, it would have been nice to remove the bias just to avoid this critique, but it will be prone to errors. Instead, our paper opted for quantifying and showing the magnitude of the bias, which was again an analysis of its own, which is presented in the supplements on our paper.

Comment #3: Mora et al. use the daily mean temperature in their analysis, yet much research has highlighted that the daily maxima and minima are important for plant growth. Daily minimum and maximum temperatures are projected to change at different rates, with daily minima warming faster than daily maxima. This feature has also been seen in the observations, but does not appear to be accounted for in the present analysis.

Response: Indeed there are many other variables that would be nice to add to our analysis. In addition to your suggestion of maximum and minimum temperature it would be nice to add the extent to which thresholds in other variables such as nutrients, ozone, CO2, etc would be crossed, reducing or increasing the suitability of the planet for plant growth. We chose the three most basic variables (mean temperature, water and solar radiation) as to be able to assess 3-way interactions (three-dimensional thresholds are shown in the figures of our paper). More variables would be a nice extension to the approach we propose, especially looking at the interactions among climate factors; that of course, will come with their own challenges (e.g., assessing, visualizing and interpreting more than three dimensions) and limitations (e.g., cumulative uncertainties of used variables).

Comment by Colin Prentice
Comment #1: Implications of global environmental change for the growth of plants (whether as natural ecosystems, managed forests, or crops) are a hot topic. Key aspects, such as the control of leaf canopy temperatures, the ability of plants to acclimate to high temperatures, and the effectiveness of CO2 "fertilization" and water saving by plants at high CO2 under different environmental conditions, are incompletely understood. There are obvious concerns, for example, for regions that are undergoing increasing drought and where this trend is projected to continue, and about the major geographic changes in global agriculture that will be necessary if "high-end" climate change projections come about.
This paper attempts to cut through these complexities by means of an empirical analysis of net primary production (NPP) derived from satellite observations. The results are dramatically more pessimistic than previously published analyses obtained with Earth System models, which have many of their own uncertainties and problems. However, there are good reasons to suspect that these new results are strongly biased towards a "glass half empty" interpretation of the evidence. In summary:
The specific method adopted implicitly assumes that climate combinations that are rare today are unsuitable for plant growth. So for example, projected future increases in temperature and rainfall together might create novel environments that would be suitable for plant growth, but the method assumes they are not suitable. This approach produces some strange but presumably significant side-effects, such as an upper temperature threshold that is below the optimum for photosynthesis as observed in many species.

Response: We actually had a similar comment from one of the PLoS Biology reviewers that we addressed by re-running our thresholds weighted by the magnitude of their occurrence (i.e. weighted by area, Fig S1). This analysis basically calculates the NPP as a function of how common the climate variables, giving less weight to conditions that are common. If the best conditions for growth currently occur in these rare conditions we would have expected to see much different thresholds than we did. However, the differences among the thresholds (weighted by area vs. not) were minimal (Fig S1). Please also note that our threshold analysis also allowed for interactions among variables (Fig. 1C-F).

We were also intrigued by the seemingly low upper thermal threshold but there are several things to consider:
• Our analysis is based on mean temperature. Considering an average difference between day and night temperatures of 10oC, then any day with 28oC mean temperature could easily have extremes as high as 32oC.
• When we analyzed the interactions among variables (Fig 1D-G) it was clear that the upper temperature threshold is higher at different levels of soil moisture and solar radiation.
• The upper thermal threshold we used is actually similar to that found for corn in the U.S., in which their production declines sharply at temperatures of 29 oC, although it is less than for corn at 30 oC and for cotton at 32 oC (Schlenker and Roberts PNAs 2009). It should be noted that these and other crop species could potentially have higher resilience/thresholds than other unaided species through intensive breeding efforts and other adaptive measures meant to increase their production.

Comment #2. The quantity analysed, called MODIS NPP, is not a measurement of NPP. It is a model of NPP, that has previously been criticized for the way in which it treats plant respiration, which tends to exacerbate the modelled effects of warming on NPP.

Response: We fully acknowledge that MODIS NPP is a model and state this in the paper itself: “MODIS NPP data are modelled using remotely sensed satellite data…” While there have been criticisms on the use of MODIS NPP, an analysis by Zhao and Running in Science (their Fig. 2) indicates that MODIS NPP is quite similar to observations of NPP from the Global Primary Production Data Initiative.

Comment #3. The approach assumes that any positive effects of CO2 concentration on growth or water use by plants do not exist. This is one extreme position on a continuing controversy. The authors refer to an "over-emphasis" on CO2 fertilization in current Earth System models, but they do not present evidence for their view which is, as far as I know, impossible to reconcile with the continuing and strong uptake of anthropogenic CO2 by the land.

Response: Our paper also acknowledges the possibility of (and uncertainties around) elevated CO2 influencing our thresholds in the “Caveats and Considerations” section of our paper. Specifically, we say the following:
“Interactions among CO2 and climatic variables could also broaden or narrow modern thresholds. For instance, elevated CO2 is known to increase resistance to drought by plants closing their stoma [48,49]. However, under warming conditions the closing of the stoma may induce overheating (by preventing transpiration) and/or if sustained could decrease carbon fixation [50,51]. Likewise, the temperature ranges over which elevated CO2 enhances plant growth are strongly mediated by water availability [49].”

Comment by Richard Betts
Comment #1: One of the other very odd things about this paper is how the authors have decided that it is legitimate to "adjust" the net primary productivity (NPP) projections from the CMIP5 models on the basis of the so-called "unsuitable plant growth days" which have been diagnosed from the CMIP5 climate. Irrespective of Colin Prentice's and Trevor Keenan's concerns (which I share) about the validity of these "unsuitable days", this seems like double-counting of climate change impacts.
The CMIP5 NPP simulations *already* take account of meteorological factors, often on a much finer timescale than Mora et al, and certainly in a way which is process-based as opposed to correlations. The HadGEM2-ES model, for example, uses meteorological and hydrological quantities on time steps of less than one hour to drive the NPP calculations. This means that the model is already not allowing plants to grow when it's too hot, dry etc. This is already factored in to the CMIP5 NPP projections. For these projections to then be adjusted further on the basis of a second (and less sophisticated) interpretation of the meteorology is giving undue weight to the unfavourable conditions.

Response: Identifying the climatic condition affecting NPP (In your words, what are the levels that are too hot or too dry for a plant to grow?) from CMIP5 models is complicated for a couple reasons. First, NPP data from CMIP5 models are available mostly at monthly scales, which prevent to see what are the daily climates when plants stop growing. Second, the intricate nature of Global (process–based) Vegetation Models prevents separating the role of individual factors. In the words of a paper by Cramer and colleagues, GVM differ considerably in their prediction of carbon storage, for reasons that are not completely understood, and are often “obscured by model complexity”. So what are the thresholds under which plants stop growing from CMIP5 models is not fully clear from available data.

However, the purpose of that one figure (Fig.6), which was cited in one line at the end of our paper, and was done mostly in response to a reviewer comment, was to show that even under very extreme climate conditions, even beyond those conditions we know surpass plant survival today, NPP is still projected to increase according to CMIP5 models. A key purpose of that figure was to highlight this discrepancy, which is not new to our paper as it has been previously mentioned by papers like that by Anav and Colleagues, Reichstein and Running to name a few. We do not mean to use this to dismiss the value of modeling NPP, quite the opposite; we need more of such studies. A side note on that figure, was a personal curiosity to see if human population demands for NPP could surpass what the planet can produce under different climate and human population projections.

Comment #2. I also find it strange that the authors pick and choose which components of the CMIP5 models they will use, without realising that all the components interact. They are somewhat dismissive of CO2 effects on plants when it comes to their effects on growth, but overlook the fact that these effects also contribute to the climate change itself through the surface energy and moisture budget. Warming over land is projected to be larger because of CO2 effects on vegetation, and this also affects evaporation and precipitation. By discarding the CO2 effects in one part of the system but not the other, the authors are introducing an inconsistency.

Response: We presume you are talking about the fact that we use climatic variables (temperature, water, solar radiation) from CMIP5 models while being critical and omitting the use of NPP, which as you suggest is modeled to the full extent of its complexity. We agree with your comment and this is entirely based on the fact that the CMIP5 models are good at modeling some variables but not others. For the case at point, there is a very nice paper by Anav and colleagues, who found that while Earth System Models are relatively good at modeling actual climate variables (e.g., temperature, and to a lesser degree water), they are less accurate at modeling NPP.

Comment by Trevor Keenan
Comment #1: This paper has many issues, but the majority of their startling results are based on one (bad) assumption - that plants will not thrive in any future conditions that do not yet exist. Future climates (in many regions) will likely be unlike anything currently experienced anywhere on Earth, but assuming plants will not survive in those climates simply because those climates do not currently exist is just plain wrong.
The tropics are a good example of why the authors’ reasoning is fundamentally flawed. There is nowhere on Earth as warm and wet as the tropics (conditions in which plants thrive like nowhere else). So what happens when the tropics get warmer under climate change? The author’s statistical approach predicts that plants will not be able to survive, as currently no plants are found in conditions warmer and as wet as the tropics (given no such conditions exist). The problem is that there is simply no information in the data used by the authors to inform many of their predictions.
The truth is we still do not fully understand how plants will respond to future climate change, neither in the tropics nor anywhere else. We do know however, that models, theory and observations all suggest that the dire (and unfounded) predictions from this study are very very unlikely.

Response: In our analysis, we actually found that many plants are currently living in conditions that, at least some time during the year, are not suitable for growth and this can be seen in our thresholds figure (Fig 1G), where red points indicate conditions under which plants did not grow/respired. Furthermore, there are several papers, including a good overview by Craig Allen and several papers by William Anderegg that show that current drought and high temperatures are already causing massive tree-die offs around the world, including areas of the tropics.
In addition, our paper also includes a re-calculation of our thresholds weighted by land area to account for variations in productivity in areas of the world that currently experience conditions that are rare and, as shown in Fig S1, we found that differences were minimal. It is correct that some areas of the world will experience future conditions that are unlike anything we have seen today (as indicated in our caveats section and also in earlier paper: The projected timing of climate departure from recent variability), and it is thus difficult to make predictions about those conditions based on data on how plants grow today. However, we felt that it was not an unrealistic assumption that if plants are currently not growing in some conditions they experience today (and at times dying) then if conditions move further from suitability they will still not grow. One caveat (discussed also above) is the extent to which plants will adapt to changing climate conditions, another uncertainty disclaimed in our paper. To truly resolve this issue, though, we need better experimental data that tests the effects of changes in multiple climate variables beyond current conditions on growth in multiple plant species.

No competing interests declared.