 Thanks. So our next lecture will be Paul Dirmayer. Paul is a professor at the Department of Atmospheric and Oceanic and Earth Sciences at the George Mason University. Paul is also a senior research scientist at the Center for Ocean Land Atmosphere Studies. Paul works on studying the role of the land surface in the climate system. This includes the development and application of land surface models, studies of impact of land surface variability on the predictability of climate, interactions between terrestrial and atmospheric branches of the hydrologic cycle, and impacts of the land use change on regional and global climate. Paul is also a fellow of the American Geo Physical Union as well as the American Meteorological Society. Paul was also a senior Leonardo lecturer at the EGU European Geophysical Union. Thank you again Paul for accepting our invite. Look forward to your lecture. Thank you very much Anish and Judith for inviting me and giving me this opportunity. And thanks to Andy for a great presentation covering the hydrology side of the problem as well as the applications and operational hydrologic forecasting. I'm going to go more into you might say the theory and processes that go into our ideas, our current understanding, conventional wisdom about the role of the land surface in sub-seasonal to seasonal prediction. And so there is a sort of preamble, I would say that what we're really looking at or trying to discern is when and where is the land state important to the atmosphere, to weather and to climate? And it just so happens as I'll show that sort of the peak of this impact happens to be in sub-seasonal time scales. So the analogy is very much like with ENSO. We pay a lot of attention to the equatorial Pacific sea surface temperatures because we know that's a strong driver for aspects of weather and climate around the globe. And if you live in the US you're particularly paying attention during the winter because that's the season when there's the strongest impact. So there's similar geographical and temporal variations in how the land can affect the atmosphere as well and that's what we're really trying to get at. So for that let's see let's move along a little bit of history to start off with. Back in the 19th century during the westward expansion of European settlers into North America manifest destiny there was a phrase that was bandied about a lot saying that rain follows the plow. People would look at the great plains western US and say that's just a big desert. I mean people who grew up on the east coast or who come from Europe very different kind of climate. We can't farm there and you know the real estate speculators would say ah don't worry about it. If you plant your crops the rains will come and of course this was really just a marketing scam for real estate. In fact the same argument was used in the early 20th century to settle the interior of Australia and they sort of benefited from the fact that you could that they had a wet period a very unusually wet periods during this settlement when people moved out. Oh yeah look it actually is raining as I'm planting crops. Of course eventually comes a drought and then everyone starts to question well maybe they were lying to us. Maybe there's really isn't this land surface feedback on the atmosphere which mind you was not a scientific hypothesis it was an economic marketing hypothesis. So we might say well obviously the droughts proved that this was wrong or did they trying to build up a little little suspense here as we go forward make it more interesting. Okay into the modern era Jerome Namias we consider the father of long-range forecasting especially in an operational sense. He was not a modeler but he was really not much of a statistician either to some extent he was but really just a very clever and observant person working in what was called the Weather Bureau back then now the National Weather Service and he showed observational evidence and even proposed mechanisms that the land surface could be a source of memory from season to season that what happens in the previous season has a bearing on future weather that there's predictability here in terms of forecasting. He found that persistence of springtime conditions into summer existed over parts of the United States mostly in the central US not much on the east coast not much on the west coast but the central part he found that this was the case that heat and dryness were correlated a hot dry spring tended to be followed by a hot dry summer for instance and that these temperatures of the surrounding oceans which is kind of our conventional climate view of forcing really conditioned the atmospheric circulation over the continents. So the little contingency table here on the right is actually from one of his publications back in 1960 before even I was born and it showed that cold summers tended to follow cold springs over sort of the western Great Plains the high plains between sort of the front range of the Rockies on into say Kansas, Nebraska, Oklahoma and this area has the most continental climate in North America in other words least influenced by oceanic air masses and what you see here outline kind of the yellow bars show the temperature relationship spring temperatures on the the left column and then the summer temperature on sort of across the top that two and a half times more likely that a cool summer would follow a cool spring than a hot summer following a cold spring for instance and then furthermore warm summers tended to follow warm springs but not quite as high a probability but when you start to factor in precipitation which is sort of the other parts of the column it's another predictor hot summers followed hot dry springs and cool summers tend to follow cool wet springs in fact it's five times more likely that a hot summer will follow a warm dry spring than a cool summer would follow a hot dry spring you might say that's kind of intuitive but you know these were the first numbers to really put behind it that was observational on the modeling side evidence goes way back shukla and mince in 82 provided one of the first atmospheric climate modeling studies to demonstrate the impact of land moisture anomalies on precipitation it's kind of a hit it with a big stick experiment tremendous extremes in one case two runs one case all the continents are prescribed prescribed to be perpetually wet they're just swamps everywhere the other one perpetually dry you just pave the continents over turn them into parking lots and then you see how the boreal summer climate responds and they found positive feedbacks almost everywhere over land when the ground was wet you would get more rainfall they would keep the ground wet when the ground was dry you would get less rainfall the one exception curiously was over India which had negative feedbacks and that had to do with the thermal driving of the elevated heat source over Asia drawing in a stronger circulation actually changing the flow and bringing a lot more moisture from the Indian Ocean up over the continent kind of the classical monsoon driver but when you have a numerical model you can you can answer all kinds of questions you can do sensitivity experiments how large is the impact of a particular land anomaly on the atmosphere what are the relative roles of ocean versus land and the chaotic nature of atmospheric dynamics in determining climate over continents and weather predictability and predictions and models are great tools you can make changes and you can see the results in ways that you can't or maybe shouldn't do with the real world we're kind of doing that right now with climate change it's a real-world experiment but again what we really want to know is how do these land atmosphere feedbacks work in nature it's one thing to play with models are the models right and it gets to to David's question to Andy earlier about do we have the data to validate and to understand these models so the processes that link land and atmosphere it's the surface fluxes okay the physical link is there through the energy balance you have the radiative fluxes shortwave and longwave downward and upward what's class classically called the four components of radiation the thermal and the visible solar part and the sum of those is the net radiation that's available the net energy available at the land surface they can then be partitioned into sensible heat fluxes basically warming of the air by contact against a hot surface or the arrow going the other way if the ground is colder than the air then the essential heat flux will be from atmosphere to land latent heat flux which is the energy that it takes to evaporate water and that can latently be realized elsewhere when that water vapor somewhere else in the atmosphere condenses on the cloud that heat is then released and then ground heat which I list here is a residual and all the little terms there show the the variables the state variables and the parameters that affect these particular fluxes and then the ground heat fluxes the residual it's kind of like the the the budget the sum in your bank account of all the withdrawals and deposits ends up changing your balance in the end and that change in balance is what goes into additional heat storage or taking heat out of the soil likewise there's fluxes in the water cycle as well there's a water balance precipitation being the big input evapotranspiration is the same as the latent heat they're the same the same process except in one we're quantifying it in terms of a massive water flux and the other we're talking about an energy what's the energy it takes to evaporate that water and then runoff drainage which is basically you know much of andy's work andy wood's work is in that realm and then changes in storage soil moisture is a change in storage so is the ground water so is snowpack so is water and reservoirs and lakes these are all parts of storage change even water stored in the bodies of the plants themselves a vegetation is a variation in storage so each of these arrows that we show they're either sparsely measured poorly measured or basically unmeasured unfortunately again getting to to david's question about observations to validate these things and this has been a big issue and it's why progress in this sort of land atmosphere interaction realm has been slow things are finally improving significantly over the last 10 years or so both in terms of in situ measurements and being able to utilize satellites to do a lot of a lot of the observation on a large scale but you know that's kind of in progress in process right now and again these are all the things that are happening just right at the surface at that interface between land and atmosphere and there are process chains that link land and state anomalies back to the atmospheric components like clouds and radiation and I show this here all these little abbreviations basically we're going from the earth surface on the left on up to the free atmosphere on the right surface and subsurface soil moisture net radiation surface temperature are all things that can then feed into these surface fluxes the latent heat flux and the sensible heat flux e f you'll see a lot in my talk evaporative fraction that's basically the fraction of the net radiation that's going into evaporation that's going into late heat flux and those fluxes affect the near surface humidity q and the temperature t and those then start to affect the lower atmosphere boundary layer processes and quantities characteristics like the depth of the boundary layer how much entrainment at the top of a rapidly growing boundary layer the lifted condensation level from your thermodynamics class parcel theory how far do you have to lift a parcel before it will cool to the dew point and form a cloud and the moist static energy or moist enthalpy in the boundary layer all of these things go towards determining if and how strongly you create clouds you generate precipitation okay now these are all arrows going in one direction of course there are arrows that go the other way obviously rainfall directly affects soil moisture when it rains you get mud right so we're kind of taking that as given an obvious here the key is to look at the more subtle feedbacks from land back to atmosphere that they create this linkage now these pathways involve the water cycle or the energy cycle or both the blue lines are basically purely water cycle pathways links in the chain the red lines are purely energy cycle and then there's sort of a rainbow pattern of degrees of water versus energy they were showing these different links and also when we model this they fall into the purview of different components of a of an earth system model the first two columns are really the land state and the surface fluxes are really calculated by land surface models like Andy mentioned earlier and then the near surface temperature humidity and all the atmospheric components are part of a GCM a general circulation model and so these components are coupled and they have to talk to each other every time step and pass this information these fluxes back and forth to operate correctly and then the arrows also make it seem like it's this sequential linkage things kind of moving along and in fact when we have computer models to simulate this climate models they necessarily represent these processes in a sequence of subroutine calls but in reality everything's happening all at the same time which makes it very challenging to untangle especially in observations in a model you can go in and pick one of these arrows and sort of change the code to to remove a link and then you rerun the model and you see how things change and say ah this link is important and it causes this response you can't really do that very easily or very wisely in the real world so understanding these links and process chains in the real world in nature and observations is tricky to suss out but we do have statistical techniques and process based techniques and metrics to do this but that's a whole other topic in a whole other 45 minute lecture which we're not going to get into here. Now if we sort of turn this the right way around on its side so that the lands at the bottom and the atmospheres at the top oh I was going to also mention sorry this is a very local couple view a one-dimensional view through the atmosphere neglecting circulation this is just saying what's happening between the land surface and the atmosphere directly above it and then we'll get into the how the circulation comes into play in a moment but if we turn it on its side and we look at this as kind of a pipe diagram of flows of water and energy we can actually quantify this and using data from flux towers and from in situ measurements and soundings and so forth we can quantify over time a kind of climatology and see how does that compare observations versus models here a particular location in Arizona we're looking at where we have flux measurements and meteorological measurements and rainfall rain gauges and so forth we can see these linkages and here the the arrows the colors indicate the correlation between the starting variable and its downstream variable so if a is the forcing or a is the driver the land surface driving the atmosphere b is a response variable the colors from blue to red tell you the correlation between a and b the green boxes tell you the variance the standard deviation of that particular variable over Arizona specific humidity varies a lot and the moist enthalpy varies a lot and there's a strong connection of humidity determining the moist enthalpy in the lower atmosphere it turns out say in the era 5 reanalysis they under represent this variability and the thickness of the arrow is the coupling strength it shows a very weak coupling oh we've got a problem in the in the ECMWF model it's not representing this process well if we look in the UFS the newly developed forecast model it's under development for the weather service we see it does that pretty well but it's like way overdoing some of the land surface to atmosphere coupling in this desert regime so this is becoming a very handy diagnostic tool so this is all thermodynamics now not the dynamics this is sort of just looking in the in the column now as soon as the sun comes up we have land surface impacts on the atmosphere here we're looking at location in Kansas and showing the difference in the daytime daily evolution of various quantities in the atmosphere and at the land surface between having wet soil versus dry soil so here's our evaporator fraction in the upper left higher values mean we have more latent heat less sensible heat when the soil is wet we have a very high evaporator fraction throughout the day when the soil is dry it dips down to very low values around one two o'clock in the afternoon and then rebounds again towards sunset net radiation is pretty similar between the two and the ground heat flux is a little stronger when we have dry soils and we have wet soils but we see a big difference in the depth of the boundary layer a daytime boundary layer grows much deeper when the air is dry than when it's wet because when the air is wet you don't have to lift it as far before it saturates and makes the cloud kind of makes sense right from dynamics the lower right is what's called a mixing diagram and this is actually an extremely useful tool and metric for understanding what's going on the x-axis shows humidity in this case it's specific humidity in grams per kilogram the y-axis is the potential temperature two meter potential temperature in kelvin and each dot is an hour throughout the day from sunrise to sunset so we see when the when the soil is wet we start off relatively cool and moist atmosphere and it just gets moisture warmer but especially wetter and wetter throughout the day until sometime around four o'clock in the afternoon we hit the peak specific humidity and then it kind of levels out drops back a bit and our temperature stays at a certain level in a dry soil case on the days with dry soil we start with a much lower atmosphere humidity the atmosphere only gets wetter from evaporation until about 10 o'clock and then the rest of the day it just dries out and gets really hot and we hit a much higher temperature a much drier situation you could take this cue at the bottom multiply it by the late heat of evaporation and take this theta and multiply it by ccp and then you would have energy units you would have joules per kilogram on each axis and then you could actually connect this to your sensible heat flux input your latent heat flux input entrainment of energy from the top of the atmosphere and really understand the components of what's going on and also when you do that then the slope of one of these lines actually starts to correspond to the bone ratio so it encapsulates a lot of information to tell you how the land in the atmosphere are connected and interact okay so to give it sort of the very descriptive empirical right side of your brain picture of land atmosphere feedback so you can think of it as being like a recipe that needs to have three ingredients there has to be this sensitivity by which I mean when and where is there an active coupling from land to atmosphere if soil moisture is changing if the vegetation is evolving if there's snow or no snow and the atmosphere doesn't respond to it doesn't change then who cares right the land is not important there are parts of the world where this occurs like places where you have very strong onshore flow most of the time the coast of Oregon land surface state doesn't affect things because it's the pacific ocean it's really determining the climate there variability so a coupling results in a significant impact only when the land surface anomalies are large enough when they change you can have a lot of sensitivity but if the land surface doesn't change it's never realized the Sahara Desert is a great example there's really strong sensitivity of evaporation to soil moisture you add water to the soil there's so much heat there's so much sunlight it evaporates readily but there's just never any soil moisture because it never rains so there's a lot of sensitivity but no variability every day is as dry as the last and the last one is memory so if the coupling and the anomalies are not persistent if they don't last very long then the effect on the atmosphere will be short lived and the impact will be minimal for instance when you have very sandy soil over over a karst terrain a lot of underlying chalk or limestone and yes geology matters and all of this too you could have pouring down rain for three days and then the water will just drain out of the soil very quickly and it will go back to a dry state within a couple of days there's not much memory in a situation like that okay so sensitivity this is kind of what it looks like so here's a plot soil wetness on the x-axis going from zero is completely dry to one is saturated and then the y-axis is the evaporation rate in this case it's shown in in watts per meter squared and so there'll be some very low range of soil wetness where there's basically no evaporation the water molecules are locked into the soil they can't really get out and then you hit the wilting point and then suddenly evaporation increases very rapidly as soil moisture increases this red area is an area of very strong sensitivity a range of soil wetness then when you get above a certain point a wilting point then the soil moisture is no longer increasing with i mean the evaporation is not increasing with increasing soil wetness there's a lot of spread so clearly there's other things that are affecting evaporation but it ain't soil wetness so this is where you would have a range of sensitivity and again in that high range what's actually happening here is that things like wind speed and how dry or warm the air is is affecting the evaporation rate if there's clouds or no clouds that's really what's controlling it the atmosphere is controlling it in this range you're energy limited in this range as opposed to here where moisture limitations are controlling the evaporation rate and so that slope is a measure of the sensitivity and how tightly those dots are grouped along that line the correlation is also a very strong measure of this coupling strength the sensitivity another way to portray it here this is the mean evaporation rate and then the y-axis is the sensitivity of rainfall to a change in the local ET and it shows there's sort of a medium evaporation rate probably somewhere in the middle of this red line this slope where the rainfall is most responsive to changes in evaporation so the the green plot is that land surface component land connecting to the surface fluxes and then down here is that surface flux is connecting to the atmosphere links in the chain going all the way up so variability is whether a location is moving around if from day to day you're moving back and forth within this range then you have the variability to for the land surface to be driving changes in the weather if you're up in this wet range moving to higher lower soil wetness in this rather wet range isn't resulting in anything at all this is what the variability looks like this is a map of the standard deviation of daily soil moisture the left column is the surface soil moisture top 10 centimeters uh subsurface is on the right and this is going from April May June in the middle July and August so you see global patterns dark colors are a lot of day to day variability what you see this this strong band that kind of migrates north from month to month is basically the snow melt front uh whether the snow melts earlier later in a particular year when is it's melting how much is it melting melting snow increases the soil moisture so you see that in that region the largest variability is there also in transition areas between arid and humid zones like the Sahel of Africa between the wet tropics and the dry deserts or the us great planes where it's humid in the eastern us dry in the western us these are places where you tend to have a lot of this variability and here's a map of soil moisture memory for instance persistence of anomalies again surface soil moisture on the left deep soil moisture on the right there's more the deeper you go in the soil the the the more memory you have the longer the time scales but again we see these spatial patterns very long memories where it's arid that's deserts that's also non-desert areas that have a wet and dry season when you're in the dry season a lot of the southern hemisphere subtropics is in the dry season during june july august and so those areas will have a lot of memory if you had a if the wet season was really wet that wet anomaly will carry through the entire dry season as it's drying down whereas if you had a a failed monsoon a dry anomaly then that you'll have a dry soil moisture but they will stay separated those two years will always stay different from month to month to month very long memory long memory where the ground is frozen the ground's frozen the water can't move so the soil moisture stays the same when it's covered by snow it's sort of isolated insulated from the atmosphere short memories where it's humid where it rains a lot short memory in forests forests are deep rooted they can kind of moderate the soil moisture and in fact there's even a process called hydraulic redistribution by which plants will take soil moisture from meters deep in the soil and bring it up and redeposit it in the shallow soil and it tends to moderate and prevent soil moisture anomalies from lasting very long okay so quick recap of the theory pathways by which the land can affect the atmosphere energy balance or energy cycle water cycle we sort of have these two legs these two main links in the chain the land states affecting surface fluxes and then the surface fluxes propagating up into the atmosphere a series of chains going up to clouds and precipitation and then our three ingredients sensitivity variability and memory all the same mechanisms yes sorry to interrupt with three or four minutes oh okay i got to get moving here so predictability and prediction s2s you've probably seen this diagram um two to four weeks sub seasonal range is a very hot topic and it turns out that's right where the land tends to have most of its impacts so you need those ingredients you need good models you need accurate analyses and by the way this diagram you've seen it a few times already it's meant to be representative of a mid latitude location sort of mid continental where an ocean anomaly takes a while from to propagate from the tropics into uh into the mid continental areas um so i mean this is kind of reiterating i'm going to skip that our positive feedbacks they occur when a change in soil moisture an increase in soil moisture results in an increase in evaporation the plus areas positive correlation that means we're in an energy limited regime and soil moisture is driving evaporation that happens when we're moisture limited in an energy limited regime it's the opposite evaporation goes up when there's more energy and that draws down soil moisture they become anti-correlated so the blue shaded areas there's not a land surface feedback on the atmosphere the red colored areas there are and so that's kind of a reversal we've put a diode the wrong way in the circuit and now the current is not flowing and snow cover tends to cut off the connection and we become uncorrelated we get zero correlations and that kind of the situation now these now we were talking about where these couplings happening it is mostly in these transitions between arid and humid regimes and the glossay experiment showed this with a dozen models simulating climate in a way that we could quantify by playing with the soil moisture states we could see the atmospheric response and it turns out that these hot spots tend to be this is for June July August in these transitional regimes the great planes of the US between humid and arid the Sahel in Africa between humid and arid northwest India Pakistan between the humid and the arid it doesn't show up so well in this original experiment but many will quantify a band that kind of goes across the Eurasian steppes from Mongolia across to Ukraine is another band if you look into December January February you see areas in the southern hemisphere that light up in Australia and South Africa and South America so why is it there I already show that that soil moisture to surface flux region shows up in arid to semi-arid regions where you have a moisture limitation but the atmosphere has to play along too and it's in the humid areas it's sort of the eastern and especially the southeast US is where you have high convective available potential energy and where the atmosphere is sort of primed to respond an increase in moisture to the atmosphere will make it just that much more unstable and that much more moisture that can be condensed out into clouds and turn into convective rainfall and where these two regions overlap happens to be right along the central US kind of in that great planes area and so it turns out in a prediction experiment glossay to which was a hind cast experiment kind of like the s2s or sub-expo we did this about a dozen years ago we find that if you look at the wettest half of all of the thousand we did I think a thousand different forecast cases across many years ensemble members and forecasts and the wettest half of the initial soil moisture anomalies had more contribution in the dry areas and the driest half of the initial soil moisture anomalies impacted just on the wet side of that hot spot which kind of makes sense because if you're in a dry location and the soil is anomalously wet oh you've moved into the sensitive range now we've got some predictability from the land surface if you're in a humid region and the soils are kind of dry now you've moved down into that sensitive range and ah now we have some predictability from soil moisture and in fact what we found was and this is showing the temperature skill across 12 models the top is with realistic land initialization at various lead times out to from two weeks out to two months the middle rose when we randomized the initial states of the land surface but the atmosphere and the ocean states are the same so we see very dark in the short range that's initializing the atmosphere that gives us a lot of predictability um we have these patterns that show up in both that have to do with sst this large scale forcing changes in circulation if you take a difference map what you see the red areas is where the land surface initialization is contributing some extra predictability some extra skill into the forecasts interestingly it's not really in the hot spot it's in the western us and kind of across the northern tier and that seems to have to do with the fact that the memory is actually weak in this great plains hot spot the memory is stronger in the western us and early on in the forecast it's the quality of the soil moisture initialization that gives you skill in your forecast but as time goes on it's areas that have a lot of memory that began to dominate after three four five six weeks because if you don't have memory how are you going to communicate skill into a forecast six weeks out and so this is a very interesting and useful tool as well to understand when and where is a land surface state important for forecasting i'm going to skip this slide about the contributions and focus on this land impact on prediction skill so this was actually trying to quantify this with the cfs model looking at april may and june initialize forecasts on the top for temperature humidity depth of the boundary layer and rainfall so there's three curves here there's a red curve which is well let me start yeah red curve is forecasts where the land surface is not initialized realistically the green curve is where it is realistically initialized uh from reanalysis initial conditions and blue is where we specified we cheat we specify the soil moisture states throughout the entire forecast and in this case these are all leads that are going out from zero days out to two months 60 days and so you see skill of course drops off with time the difference in the curves are the shaded areas the tan region is the difference between the red curve and the green curve the effective land surface initialization the cyan shaded area is the difference between the green and the blue curve what happens when you include when you specify perfect forecast as it were of the land surface states and what you see is there's an immediate impact on day one but the peak depending on the variable you look at is always between one and two weeks is when the land surface has the greatest impact but that impact carries out for many weeks into the future all the way out the exception being precipitation it does have a peak in about two weeks but in this model precipitation was very insensitive to land surface states unrealistically so this is something we're looking at correcting in ufs so in fact okay so i'll go ahead and wrap up here uh let me skip ahead then and oops to um to get just to the point of application so oh here we go um forecasting applications a lot of talk about uh about droughts uh by andy to talk about the sort of the other side of the coin and heat waves um oh yeah right here the 2020 heat wave in the western us we saw very warm anomalies very dry dew point depressions it was very much connected with very dry soil moisture and you asked the question did a hot dry atmosphere dry the soil or did a dry soil heat the atmosphere and the the answer is there's a feedback it's both dry soils can greatly amplify a heat wave you can have warm air invoked into a region desiccates the soil that can feedback from day to day creating a warmer and warmer boundary layer storing energy in the atmosphere leads to an extended heat wave and our current understanding is that this is what happens and it was very prevalent in the 2018 european heat wave where we were able to show then in fact the land surface states the very dry soils impact fed back on the atmosphere and created these very warm conditions and in fact there is a critical soil moisture that when you cross below that threshold value suddenly you get hypersensitivity of temperatures to soil moisture drying conditions getting that break point right and forecasting that correctly in a model is critical for getting heat wave forecast which you need to do on a one to two week time frame so to quick summarize predictability of weather and climate from land surface states comes when there's feedbacks there's not always feedbacks there's two legs in this path land surface component and atmosphere component you need to have variability you need to have sensitivity you need to have memory and the effects start as soon as the sun comes up but the peak is around one to three weeks and can last for several months the impacts can be local but there's also i didn't get a chance to show the remote effects but there can also be downstream effects with teleconnections and rosby ways much like you have with sst feedbacks so i'll stop there sorry i went a little bit long no it was great fault thank you again there's really great explanation of these complex feedback so thanks