Belinda Medlyn: Traits, from leaf to ecosystem: a perspective from an ecophysiological modeller

Belinda Medlyn: Traits, from leaf to ecosystem: a perspective from an ecophysiological modeller



thanks very much to the organizers for inviting me and particularly giving the opportunity to speak second and I have never spoken second at a conference before as a modeler I'm always on the last day look at any conference program model is always on the last day and it also has thrown me a little bit because I normally I get the chance to tweak my slides to suit the audience so I'm always there on the last day of the conference going also impose head there stones who said that I can just change stuff this time I don't get the chance to do that so I actually have to like start from first first principles and so being a modeler I felt that it was really necessary to nail down what a trait was before I started talking about traits because this is the traits term is very very broad and the standard definition as Peter says comes from this paper by V ol which says that a trait is any morphological physiological or phenological feature measurable at the individual level so it's basically anything you can measure on a plant anything at all it could be you know it could be an bark color that's a trait and it could be height that's a trait for me that that's too that's too general and I don't think it actually reflects what people really think about as traits so I did a bit of a sampling of some of the people that I work with to try and find out what they thought traits actually were and a lot of people came up with a definition which was something like what what mark chalk has suggested that it's a trait is actually something that characterizes a species so it's not it's not anything that you can measure it's something that's really quite specific to that species and but then mark and I went off on this long tangent about whether relative growth rate was a trait or not and so relative growth rate it does characterize a species there are some species that have fast some species that have slow but you can't really measure it in such a way as to show that very easily because it also depends on so many other things it depends on the plant size incredibly depends on the temperature depends on the environment so I don't think we agreed in the end but I I was moving away from marks definition and moving towards the definition that Ian actually gave me a couple of years ago when I tried to pick his brain about what a trait was and so in definition was just it's something that's more constant within a species than across species and I found that really quite helpful because we all know that everything you can measure on a plant varies but if it's something that kind of is more cut you can you can look at the variability in a huge data set and the variation is across the species rather than within the species so that's kind of gelling with my way of thinking about it I had to say I got a few more and more tongue-in-cheek definitions of traits as well so Brendan's version was anything that correlates with anything else and so someone else named this said anything you can measure anything that ecologists can measure so apparently like leaf mechanical strengths became a trait as soon as mark Westerby got the machine that does leech leaf mechanical strength anyway so my idea of a trait is something it's more comfortable in a species than across species and really that map's fairly nicely to a model parameter and so what I've been trying to get my head around for years actually ever since I moved to mark westerby's University was to think about the relationship between traits and model parameters so you know models have parameters which are constants which characterize the species their input to the model they also have drivers which are you know external things your temperature your tire your co2 and then they have variables which are things that the model actually simulates and predicts so things like internal variables realize photosynthetic rate or ecosystem GPP so there is there is quite a strong relationship between traits and model parameters it's not it's not as simple as that and pile the reason is because but but my parameter might be your variable so different models have different sets of parameters so some people take for example VC max as their parameter whereas other people take don't try to predict leaf nitrogen content and try to predict VC max from that so it's not it's not as simple as saying this is a trait because this is a parameter because parameters vary but I think that's the same thing with traits actually so my trait might be your variable so some people have traits which which other people would consider as variable so it's really it's a little bit context dependent and a bit model dependent and now they're really nice definition I got was from rowdy actually where I see but then if you remember this conversation rowdy but but I said I think traits are just parameters and you said no no traits are things that we can use instead of having to get parameters because a lot of model parameters are actually really complicated things that you can't measure and so this I mean just one example there's hundreds of them in the literature of things that that that we take as parameters for our model but you can't actually measure so the nice thing about traits is that it gives us a way of at least getting a handle on what the parameters are so what what Brad did here he's got some parameters a like leaf saturated water content leaf osmotic potential at full turgor not so easy to measure so what is done is to measure leaf mass per area and wood density and try and come up with a simple relationship to get the parameters from the traits and you definitely said it I thought it was very insightful and another thing to bear in mind is not not all model parameters are actually traits so some model parameters we don't think of as varying by species so things like Rubisco specificity and the q10 of respiration we just assume that they aren't very constant for all species so not all parameters are traits and and the other big thing about traits and parameters is that the least in the models that I work with you have to focus on the dominant species you can't really think about ecosystem average traits so here's a picture of a nice little woodlands where we go bike-riding some sometimes near my my house you've got three species in the overstory here and I cannot tell you how many species there are on the understory there's all sorts of things and then you know little tiny things as well so for me as an ecosystem modeler I really have to worry about what those big plants are doing I can less worry about these you know that dozens and dozens of plants in the understory so my parameters are going to be weighted towards the dominant species and I think that's a little bit different between the way that most people work with traits which is often to take an ecosystem average across the species so there's there's kind of thoughts always working through to try and understand differences between traits and parameters and so I've kind of ended up with a working definition of what a trait is which is that it's it's immeasurable something that that characterized that species that can be used as a model input parameter all at the same time we really need to be aware of the sources of variation in that in that trait so with that definition I can now move on and just talk about so talk about how we can use traits to get well how do we get from traits to ecosystem function and so my first example is thinking about the carbon side of things and this comes from a paper that some a couple of years old now and just gives an example of how we use traits in our modeling so this example what we were trying to do was to look at the co2 response of grassland species driven by Peters biocon experiment and and so what we had was a very simple model where we had inputs which are basically climate and the nitrogen availability in the soil and and then we were simulating photosynthesis npp that becomes root biomass leaf biomass nitrogen uptake photosynthesis leaf area index and nitrogen concentration and we're representing the different species in this model by their parameters which we also called traits and so the sorts of traits that we had were things like our photosynthesis traits which are VC max per unit nitrogen G 1 which I'm going to talk about in a lot more detail shortly and the respiration loss so this is actually a carbon use efficiency parameter so how much of your growth photosynthesis is lost to respiration allocation parameters so allocation between foliage and roots and turnover parameters so lifetime of foliage and roots nitrogen uptake traits FLA and then the ratio so another trait we had to have was the ratio of nitrogen content in the roots to that in the in the foliage we had a bunch of constants in this model as well so our constants were things like the quantum yield and the temperature dependences the phenology we didn't get to grips with that was just constant by species and and then we really just focused on growth grasses and forbs not legumes Nazi Falls nothing big nothing water-stressed so that's kind of our you know a limitation of our universe and in this universe the model actually comes to a nice equilibrium which makes it very simple to think about or relatively simple if you're a modeler simple to think about how those traits influence the ecosystem function and and the way that we represent this is to is to think about the relationship between net primary productivity of that species in monoculture as a function of the leaf area index and there are two relationships right so there's one relationship that says you've got an L AI that determines how much NPP you can have and that's just with that laa you absorb some light which depends on your light extinction coefficient you use that light to do photosynthesis which depends on your photosynthesis traits and you then turn that into productivity which depends on your carbon use efficiency and that's kind of a saturating relationship because because of the fact that as you get more and more lights you can't as you get more and more leaves you can't absorb more and more light the other half of the system is to say okay given your MPP that means that you have a certain amount of lii that you can have so as this thing comes into equilibrium the amount of MPP if you multiply that by the allocation of foliage the specific leaf area and the lifespan of the leaves so the Turner of the foliage you work out how much la I you have and eventually the system comes into an equilibrium where those two lines cross and so this this this visualization helps us to understand how I trait influence our productivity because we can easily see how if we change one of those traits we change where the system sits and so for example if I change my SLA all it does is to change this relationship between the LA I and the NPP so if my SLA becomes lower I can make less leaf area for the same MPP and hence I get a different relationship here in a lower la I if you then think about what happens with elevated co2 so this exercise was all thinking about elevated co2 responses so here what happens when you increase the co2 concentration is that you increase the light use efficiency of the foliage and so that's my green line and then there's a feedback because that increases the LA I that can be supported and so here we are at ambient co2 low SLA higher cell here we are elevated co2 low SLA high SLA and the the implication of this model is that the co2 response is going to be higher for the thing that had the lowest la I to begin with so you can see that this this response from here to here relatively speaking is quite a lot larger than this response from here to here and so what we're seeing is that this model with those assumptions predicts that the responsiveness to co2 is higher in low productivity systems so low SLA low foliage allocation species with those traits and so implications of this is really that the slope arms are predicted to be more responsive to elevated co2 and if you go further than that it means that elevated co2 should actually promote coexistence so you should get more of the slow plugs in our ecosystems and that is you know that's what the model tells us and people don't like this result I can see rich making faces over educated and I know I know right it doesn't gel and the reason it doesn't gel is because of the that you know the assumptions in there over the traits stay the same and so unless the unless is that perhaps the traits are going to change differentially among species with other rated co2 now some of those traits we know really well and we know how much they change with elevated co2 so SLA changes a tiny bit VC max per unit nitrogen changes hardly at all do you want I know very well does not change with elevated co2 and but there's a bunch of other traits in there which are much less well characterized like foliage allocation and like the carbon use efficiency and then there's a couple of traits which aren't in there at all and so one of the things that we're spending a lot of time thinking about at the moment is the trade-off the in growth and storage and so plants this this model really takes all that photosynthesis and sticks it straight into growth there's no there's no what's the word there's no possibility that the plants might actually be storing carbon for later on so one thing that we're doing with our models now is to try and add in a storage strategy so I would I would call it a hidden trait it's not in the models and yet you know when you talk to people about the implications of photosynthesis for growth they kind of most physiologists have a feeling that that trait should be somewhere in there okay so how am i doing all right moving right along that was carbon I need to talk about water I assume everyone was going to expect me to talk about water so thinking about water the main thing that we're main thing that characterizes the water use of the plant is the stomata conductance but this tomato conductance I would argue with you is clearly not a trait it's really not it's highly highly variable and so this plot comes from our 2011 paper where we were looking at stomata conductance and how it correlates with photosynthesis co2 and VPD and you can see I mean this tomato conductance it varies it varies over the course of the day it varies with all kinds of environmental conditions but it's predictable it's predictable from this relationship and the slope of this relationship varies consistently across species so the species trait is really the slope of that relationship which is basically rg1 so this this the stomata conductance we would argue the trait that we need to know about is g1 which is I know it's not a very catchy name and I often get pulled aside by people to say what is g1 really so g1 is a is a parameter in a model which we derived and from the theory of optimal stomata behavior I should say so colin has a similar but slightly different model which comes from a different optimization criterion which arrives at basically the same model and Roddy has another one now which he's got a poster about which you have to go and look at if you're interested in g1 that is and it's but basically yes okay so the same model can be obtained from different optimization criteria the way I think of this g1 parameter it's related to the marginal cost of water to the plant so it's kind of related to the plant's carbon to water trade-off strategy but we really haven't come up with a good name for it sometimes I call it the stomata and float parameter sometimes I call it the stomata operating point mostly I just say g1 and hope that everyone knows what I mean so g1 we know does vary systematically across pft and so this is from a paper in 2015 led by Angie Lin where she collected a whole bunch of stomata conductance data and put it together and found systematic differences with lower G ones in gymnosperms than in angiosperms shrubs grasses and crops and these data are all online by the way figshare if you want to play with stomatal conductance data and it also would appear at least in this data set to be related to our density which is a nice thing it's clearly a trait because it's correlated with something else REM and one of the one of the problems that we have at the moment is that scaling up to the ecosystem it doesn't work as we thought it was going to so Yin she found these really nice differences across species and we expected across pft sorry and we expected that they would translate to differences in a similar parameters g1 at canopy scale and so what what we did was to derive these G ones using three different sources of data the leaf gas exchange data leaf isotope data which we've got a very large database put together by will Cornwell and the flux net data and we had thought that we would see the same patterns across ecosystems but we don't and this really troubles me because it says that there's something that we don't know yet so if you look at this figure the green ones are our leaf gas exchange data and there's clear differences between the needle leaves the broad leaves those patterns are kind of there in the isotope data as well so again you can see the difference between the needle leaves and the broad leaves I don't know what's going on with the tropical forests and the isotope data that's you know there's there's a bunch of questions in this paper it has to be said but if you look at the flux net data there's absolutely no difference in this G one parameter across eco across pft and so gonca has a PhD student who's been looking into this in some detail but we still haven't really figured out what's going on so perhaps it's a problem with the way that we scale from the leads to the ecosystems could also be a problem with the fact that the the eddy covariance energy balance doesn't always close so perhaps the transpiration fluxes aren't as reliable as we might help so this paper is one of those papers that was kind of difficult to publish because it doesn't have an answer it has lots of questions in it you know we need to figure this out to be able to use this g1 parameter and okay let me see I have so many things I could possibly talk about and they're all in here I just got to decide which ones to put in okay and we'll see how far we get and very briefly G nought wasn't in the model originally we had to put a G naught in there to make the model stable and and people do often pull me aside and say what are we doing about G naught and so what we're thinking at the moment is that the problem with the G naught what it is is there is what you get for conductance when photosynthesis has gone to zero and photosynthesis goes to zero for lots of different reasons so sometimes it goes to zero because it's not and then you have stomata conductance but that's really different from what goes on during the day so we want to have night time separate it goes to zero at low par during the day and so we need to have something that happens at low power and rod is optimization model actually I think might have a way of dealing with this which I'm hopeful will work it also goes to zero at really high temperatures so as soon as photosynthesis it we know observance bubs forty degrees Fotis into this goes to zero John Drake has a really nice paper that's in review at the moment that shows that at really high temperatures canopy carbon uptake goes to zero but folks what the transpiration really doesn't and so of course it's because the believes want to cool if they possibly can so the stomata remain open during heat waves if they possibly can and that's a really different process from what happens at low par and so we're trying to differentiate between these things there's also what happens when you have really low soil moisture so photosynthesis can go to zero at low soil moisture and the plant is then trying to shut its tomato as much as it can but you still have a particular conductance or a leakiness which determines the plant desiccation rate which is actually really important in simulations of part mortality so we're also working on trying to figure out how to simulate that G naught at low soil moisture and but they're really different simulation lived in different situations and you can't use you can't substitute one for the other okay so I'm going to skip this bit sorry and the other thing that we know about G one is that it does vary with soil moisture and it varies systematically again so very systematically across species and so this figure comes from a paper by a PhD student of Collins Tracy Jo who looked at this g1 parameter how it responded to pre-dawn leave water potential across species from different environments and really what he found is that species from from wet environments had a much steeper response banded species from dry environments and we're now finding this across wider sets of species and it's also nice because it correlates with other features of the plant so these data come from an experiment that's just been run by another PhD student she Ming Lee and Chris Blackmon is a postdoc with us where he was looking across that gradient in rainfall that Peter was showing the gradient rainfall that crosses New South Wales and goes from over a thousand millimeters down to two hundred millimeters and what we find across that gradient is that the xylem silent volubility so this is what I'm showing here is p50 it's the point at which there's a 50% loss of conductivity in the xylem and it's it's the trait that Tim Broderick now calls the super trait and is in his most recent commentary in new fight because it's so well predictable from from environment it really characterizes the plant's vulnerability and is very strongly correlated with those climatic envelope that the species come from so a square to 0.72 n would be very happy with that and what we find is that the rate at which this demands a shut with River with respect to the afforded potential is also very strongly correlated with that trait and so what I'm showing here are the water potential for 90% tomato closure so stomata are pretty much shut versus the water potential for 50% loss of hydraulic conduct tivity and there's a really good relationship across species from different environments and this relationship can be predicted from not from optimization theory so another PhD student yogi Lu are attempted to implement the optimal stomata conductance theory to predict how plants should optimally respond to drying in a stochastic environment and found that it really didn't work very well so moved on to thinking about competition for water and once you come up with a a model which assumes that plants are competing for water but that there is a carbon cost of embolism to the plant then you end up with this really nice prediction that the stomata closure should be really quite strongly related to the p50 and so here what Lu has done is to make different predictions for different carbon costs and compare that with a relationship coming from a meta-analysis of these data paper and it's you know as you can see it's um it's pretty well spot-on so so to summarize where are we at with modeling water use we feel like we've come really quite a long way in recent years we know a lot about G 1 and how and this trait really characterizes the stomata conductance we know a lot about how it varies across species even though we're a bit worried about how it scales to the ecosystem we know a lot about stomata closure in response to soil moisture we know a lot about this p50 this loss of hydraulic conductivity thanks in part to the work of my colleague Brendan Choate and others there's lots and lots of data coming in looking at p50 but of course there are things that we don't know and so one of the things that I am really struggling with at the moment is that to put all of this into a model you have to like together to working tomorrow you have to tie up all the loose ends and one of the loose ends is really about what the roots are doing and so the ruching depth is the thing that determines the soil water potential that the plant sees and if we're going to predicts tomato behavior as a function of the water potential in the plants we need to understand how the water potential in the plant relates to the soil water potential and what you see in almost every ecosystem is that you get plants growing side by side same soil same environmental things flow moisture apparently but very different water potential and so this is just an example from Queensland comparing blood woods with iron box minus one two minus five pre-dawn water potential even though they're growing side by side and I've got lots of other examples that are like this so rooting depth and reading depth is one of those traits but that I'd love to have but I I know it's going to be a long time before we really get a bigger get a bigger data set there and so this has been a whirlwind tour of some of the traits affecting carbon and water cycling I feel like there's been really incredible progress over the last few years in understanding plant ecosystem function via traits and clearly you know there's a lot more to come but I guess a couple of notes of caution we do need to be quite careful I think what we mean by traits and not just say traits you know sometimes you think papers would say traits explains it and saying traits explain something is you know is not an explanation that's all you need to mean you need to say what you mean by traits and be clear about what sorts of the variations are in your universe and what times what sources around and the other thing is is that we also need to be aware at least of that the hidden traits the storage allocation the rooting depth that can that can confound our expectations and that's that's it's for me you

Posts created 34089

Leave a Reply

Your email address will not be published. Required fields are marked *

Related Posts

Begin typing your search term above and press enter to search. Press ESC to cancel.

Back To Top