 Mic'd up here, and then I'll explain what all those words mean in my title Okay, so this is uh, I'm gonna talk to you this morning. First of all, how's everybody doing? How's everybody doing? All right I'm gonna talk to you this morning about some work. I've been doing with Joe Goluseki on using atmospheric models climate models to give us to drive landscape evolutionals, drive, or provide input through landscape evolution model And so I'm gonna start off by introducing you to our amazing field setting and the natural experiment it provides And and then I'm gonna talk about how the model, the modeling kind of plays into our research strategy there And in that context, I hope to highlight some of the issues of scale space and time scales You guys can hear me, right? Space and time scales, but they come up when crossing these kind of modeling boundary All right, so here's Hawaii, 20 degrees south which puts it in the Northeast trade wind zone in the sub-tropics, and it's in the descending wind with the habit of circulation So So we have the descending habit of circulation air coming down in this area, and it's very dry. It's coming out from the top of the troposphere and So you can see this low cloud deck as I'm flying into Hawaii. This is the big island underneath me There's a valley out there You can see it poking up above this cloud deck, which is Which is constrained and height by a temperature inversion That forms as a result of the interaction of this is in heavily air Which is he getting warmer adiabatically as it comes down So this descending air intersects the maritime boundary layer And Thanks, I'm not usually this technically challenged. I don't know. It sounds like one Okay, so anyway you form these clouds that And you end up with Longwave radiation from the tops of the clouds shortwave absorption There's turbulent entrainment at this interface of dry air from that's That's Descending into the marine boundary layer, and then there's a vaporative cooling associated with that the net effect of all of this Is to create a temperature inversion That caps the cloud layer and appears Something like 85% of days. It's very consistent. It appears across the subtropics because of these effects This is a this is a radius on profile that There's one of the there's a balloon goes up from Hilo twice a day This is from the day that this picture was taken. I stuck into the archive and found the right afternoon And so height versus temperature and you can see a strong temperature inversion right here You know six or seven degrees C at about 1500 meters As I said this occurs across the subtropics as you move from the subtropical high over towards the The intertropical conversion zone or the equator the inversion goes from very strong to very too much weaker and You start to get in it it rises in elevation As you do that it's influenced by sea surface temperature as you might expect and it's also the height of that is also influenced by the strength of the trade wings But the statistics of the inversion height are fairly well characterized and Just a paper from Chow and others um the So at Hilo 17% of the time there's not an inversion, but the inversion is there the rest of the time And so from these radios on Datasets you pull up the statistics of this you see there's some seasonal variation the frequency of Inversion goes down in the mid-summer when temperatures are a little bit warmer in the boundary layer and The height of the inversion goes up there which is consistent with that height increasing towards the equator So this is these these these statistics can provide constraint on climate models that that that manifest the inversion in modern setting At the last glacial maximum sea surface temperatures are probably different wind speeds and patterns were probably different and so this was These statistics might have shifted, but it's not really well explored how in a particular setting such as Hawaii There's also topographic interactions So here's what it looks here's the how the inversion manifests itself on the island So here's here's Hilo Monacao or the observatories are and one alone or the other observatories are And you can see this very sharp dividing line between vegetated and not vegetated and that's because that's about the average elevation of the inversion where the inversion layer caps this those clouds and so you have steady drizzle all the time below that when these clouds form and basically desert conditions Above just why it's good for the observatories So here's a tropical rainfall measuring measurement mission data averaged from 1998 to 2008 or it's the probably the monthly average So You can see the scale bar here. That's not a typo. There's not a period missing there It's 2.7 meters of rain per year that trim is picking up and that's probably an underestimate gauges gauges in here show at least isolated places where it's more like four meters of rain a year and you know Couple tens and millimeters a year up in up in this upper elevation mostly delivered by frontal storms that pass through This is the average inversion height from that histogram and you can see that it draws a beautiful bull's-eye around these These areas of very low precipitation And the river networks as you might expect are responding to this this is the mapped hydrology just put the USGS map hydrology and This blank patch in here is mostly because it's very young lava flows on on Montoloa, but you have You have high drainage densities I'll just zoom in a little bit here and you can see as you come down from the top of the mountain There's a very few well-developed channels that are mapped by the USGS You get down below the inversion where you have perpetual drizzle and you start to have a lot of You know a lot of river channels a very dense drainage network Little bit about the geology Because that's relevant here The lava flows on Monacaia that we're interested in are in the tens to hundreds of thousands of years old so these The the landscape here has evolved over at least one and maybe a couple of glacial glacial interglacial cycles so it's important to understand how how the Properties of the the inversion and the rainfall distribution have changed In in those two different kinds of climate extremes There's the high just the hydrology again just plotted a little differently on top of the geology And I'll point out that there was an ice cap on Monacaia Not just at LGM, but Several stages prior it's very well described by Stephen Porter in 1979 and there's been some recent work doing cosmogenic dating to nail down the timing of deglaciation the glacial stages and and The last deglaciation was about 15,000 years ago. There's also good evidence that there were outburst floods occasionally from From this ice cap on the top of the mountain Which would have very quickly dumped a lot of water into the top of the drainage network Which is very different from the very bottom-heavy signal that you see because of the inversion I'll just take you on a very quick photographic tour This is up at 3,800 meters near the top of the mountain right about at the glacial limit Foot is a little dark, but you can see This is just kind of the boundaries of lava flows and a cinder cone here, but There's a channel head that starts kind of in this depression and wanders off down the hill Over the mountain, but you can imagine that that doesn't see a lot of flow the desert. It's very rocky Get down to about 3,000 meters. You're starting to pick up some scrubby vegetation And you can follow these channels down. There's very nice scouring features even up here Or you don't have much discharge and so at times there must be a fair amount of water moving through these things Carrying sediment to a braid the a braid the bed right now. The only water in this potholes in this plastic bubble Moving down to about 2,000 meters a little bit more continuous vegetation cover. You're kind of in the grazing zone There's been a fair amount of ranching that's gone on so that's impacting it the channels look vaguely the same You see the same kind of scouring potholes and things and still no perennial flow very ephemeral Getting down to about 1,100 meters now you're now you're kind of below the you're starting to get down into the clouds Below well below the inversion that caps them This doesn't really show up, but there's a trickle of water coming down here. You can see the ripples That's a spring. So you're starting to pick up groundwater And yet you start to have perennial flow at some low level in the streams As you get as you imagine down towards sea level you start to have much higher perennial discharges And you get these big nick points big waterfalls that are probably good for showering under and Again down low compared to the really rocky thin soils above you have these big thick Sapper lights and weathered deeply weathered red Which are going to impact the hydrology they're going to impact the hill slopes delivery of sediments channels and so forth Let's take a look at a couple hydrographs from gauges at the White Luca River, which has this big drainage area here And Honolii stream which kind of runs up into here and Peters out and doesn't have a lot of tributaries That's just kind of the throw it up there. These these records actually go back pretty continuously to the 50s and Back to the 1917 I think sporadically before that and so the White Luca drainage area is hundreds of square kilometers Honolii is About 30, but when I blow it up and just look at the past couple of years You can see that the the hydrographs are pretty They're pretty similar to each other the bigger drainage area, you know tails off more slowly in the falling stages of these these storm events or these these runoff events which are almost certainly frontal storms and occasional tropical cyclones So it's important to understand this because the the inversion might cause this pinning of the clouds and meters of rainfall a year Down low, but it's falling as a slope drizzle And it's not you know it's not making a really flashy hydrograph, but then you get these storms that come through and several orders of magnitude higher discharges when that happens and the So what's you know, what's doing the geomorphic work? So I'm gonna start with a few examples of the scientific questions that this setting Gives us It's like to what extent does discharge variability impact the fluvial erosion rates Both above and below the inversion layer is to go from ephemeral to perennial streams Was the LGM fluvial system or top-heavy because you had an ice cap sitting up there Then it is currently How does the delivery to the sediment of sediment to the channels? How's that affected by the deeply weathered soils down low? and How are the positions of are the positions of the large nick points better explained by stream discharge or just by when Whenever they were kicked off by landsliding at the coast or something else these are the kinds of questions We can ask I could I could go four slides of this I'm sure you guys so You can imagine a number of approaches to to looking at this, but one of the things we want to do is is First is use modeling atmospheric modeling to to try to estimate the precipitation changes that might happen over these glacial interglacial time scales and Then and come up with the hydrological impacts of that So we want to simulate the modern precipitation climatology at the relevant scales Which you know, we're going to use something like a regional climate model and you need data to drive that The we want to characterize the glacial interglacial changes to inversion height storm frequency temperature hydrology And so we need we need to be able to drive our regional climate model with last glacial maximum boundary conditions Which you might get from a GCM run that was set up for the last glacial maximum And we want to incorporate these changes into a landscape evolution model to narrow down hypotheses About you know about the scientific questions that we had just asked before we go out and spend a lot of money putting in sensors and We wanted what are we going to measure when we go out there if you do the modeling first you have a better idea of what's a target and So we need a hydrologic simulation that captures the daily to seasonal hydrograph the glacial interglacial precipitation changes and then Critically responds to channel up network evolution over longer time scales like you might get from a landscape evolution model Apologize for the big messy unbroken up slide, but so we're going to take So what we're going to do is we're going to take global scale climate model output At something like one degree resolution And we're going to use it to drive a regional scale climate model at kilometer kind of scale resolution and use the mapped hydrology and geology to to to simulate Hydrographs using a hydrologic model and use the use the statistics of discharge from that somehow to see to drive a landscape a Evolution model to test some of these geomorph hypotheses All right, so about the scales of these models global climate models here's a here's spatial spatial resolution up here and Time step So minimum resolution minimum time resolution GCM's tend to sit up here tends to hundreds of kilometers at a degree is a common And but time steps of a second to a minute or so Depending on the physics you're using regional climate models, you know, you could you come down to the hundred meter scale But again there you're integrating on very short time steps Hydro hydrologic models also you generally pretty short time steps if you want to resolve daily Hydrographs and things like that and you might do that at tens a hundred meter resolution for a particular watershed Big hole in the time scale here And then you have landscape evolution models again at the watershed scale but Time steps of a year a hundred ten years a hundred years depending on the physics So The the time step of a landscape evolution model Might be longer than the entire an entire Climatological length run of a GCM so incorporating the variability that you get from the fields derived from these GCMs into something that's integrating over these very long time scales is tricky For so so what we're going to do we're going to take output from global climate model simulations These have been run by People who do that I'm not one of them We're going to use those output fields as Boundary conditions for our regional climate model. So we're doing dynamical downscaling here And so that will allow us to get some of you know, we'll have storms moving through We'll have the inversion moving up and down different conditions. We'll have some of this variability that's important and then we can take those precipitation fields and Put them over topo flow or or integrate them with par flow or you know Put them into a hydrologic model. That's you know, that's That can be done. That's been done And then we then we have to bridge this gap here So we're going to use a wharf in cars weather research and forecasting model To to do the dynamical downscaling It's fully compressible non-hydro static atmospheric climate simulation that's doing the fluid dynamics of the atmosphere Along with physics for precipitation and so forth radiation Wharf allows you to use domain nesting so you can take a very low spatial or high spatial resolution You can go to a high spatial resolution from very low spatial resolution in several steps so if we're going to simulate at the kilometer scale and our input fields are from a GCM at the one degree scale We can we can kind of step down into that and I'll show you the effects of that and why it's kind of important Wharf can ingest climate model output We need to do that can take a reanalysis data station observations can take all kinds of input to use as boundary conditions and The CCSM folks have recently Rerun some of their parts of their 20th century and LGM Runs at with six hourly output, which is what you need for wharf to do a good job So those are just becoming available in the last Six months or so All right, so here's what we're working with the This would be like so the one degree box is the size of the island. That's your input field from a global climate model And just as an example with the topography here's the topography of Hawaii with At 48 kilometer resolution. This would be like an outer outer nest in wharf And you can see it doesn't do a particularly good job of representing the island in fact the island is Doesn't even poke up above the inversion in this At this resolution so you got to go a little farther in we're still missing one of the mountains here Basically, you're doing a little better on the overall elevation at 16 kilometers But to really resolve these turret the terrain interactions with the inversion We're gonna need to do something in the five five ish kilometer And you can go extreme with that but So but you're getting pretty close At five kilometers you're getting pretty close to the right elevations and things like that So this is this is downscaled CCSM output at 48 kilometers I'm not going to make much of it This is from a test run that I did just to make things work and it's it's not quite right for interpreting but this is a cloud water mixing ratio and You know you can see it's it's making clouds on the right side of the island at this resolution So just feeling the topography to some extent, but you don't you don't get much out of that If you need the spatial variability that we have As you come down to 16 kilometers you start to do a little better. You know you're resolving the gap between the two mountains And again five kilometers Just as a sort of a side but a very relevant one We're not the only ones with scale issues Climate models in general not just warf or CCSM in general Have this kind of disparity of scales where you know at the low at the small spatial resolution the Dynamics are fully resolved. You're starting to you resolving the turbulent structures in the atmosphere explicitly From the Navier-Stokes equations And so the physics, you know work based on that Whereas at these big scales you can't resolve those those kind of cloud forming eddies and things that you need So there's there's parameterized physics out at those scales Those have been pretty well tested and calibrated based on modern observations So both of these work pretty well, but this this area in the middle which is importantly the kind of resolution that we want for Putting into landscape models. You're starting to resolve some of those eddies But you're not fully resolving them. So you have this weird mix of parameterized physics and resolved physics and it it's There's kind of a scale gap there and The the output is maybe not as reliable as it We would want it to be compared to one if running at these other scales, and it's also hard to go from something at giving boundary conditions that at these high scales or these these high DX's and Jump across that gap without the physics getting weird So we're not the only community with spatial scale issues um It's back to our yeah, so But these guys are all you know this modeling chain has been done can be done And so I'm gonna I'm gonna talk very briefly as I finish up about the this hydrologic gap Which I I don't really know how we're going to Go from there to there. There's a number of strategies out there I'll talk just briefly about that first the big ugly slide of bullet points The statistics of flow are important because the big events do the most work. I showed you the hydrographs Landscape evolution models tend to have basic hydrology in them, but it's not the you know It's not the fully resolved with groundwater and things like that and And you know all the water tends to run off instantly because you're running on a time step of a year Which is longer than it takes for the water to run off um You can use hydrologic models to do the hydrographs pretty well But they run it too small a time step time step to reasonably do a long le landscape run Including that But you do have to have feedbacks between a Between the channel network as it evolves and the hydrology And so you can't just take a static discharge map and run it forward and in an LEM and So as a my perch pointed out the other day statistics know nothing about time And so I think the answer to this at least for the moment is is to take a statistical kind of approach This is I couldn't find the original figure for some reason but You know so landscape evolution models tend to include something can include something like a stochastic rainfall approach where you put water on the landscape in Precipitation events that have a distribution of intensities and durations and in time between these storm events It's child works this way and so forth, but But it's not clear that in all settings the the kind of artificial hydrographs that this approach generates may they may or may not be realistic for a given Setting especially when the spatial variability is important. That's an area they might that's an active area of Work that needs to needs to be done We can take our real hydrographs. We can take or synthetic hydrographs that are you know tuned to these real ones From a hydrologic model and collect statistics This guy looks sort of log-normal collect statistics of discharge and so We want some way to map those into a landscape evolution model But you're not mapping it into just here's water on a landscape and it's going Through your LEM you're mapping it into the exist the drainage network of the LEM as it evolves Which is tricky and it's going to require some cleverness. I don't know the answer to this problem This is a gauntlet. I'm throwing to the community here Or maybe some of you have solved this I don't know come talk to me if you have But you can do something like you could build a probability distribution of discharge and then Parameterize that distribution Just simply maybe like I mean in a standard deviation, which and you could you could you could plot up how they how each of those parameters in the distribution should scale with nominal drainage area in your LEM or nominal discharge with a basic You know with a basic hydrology and then You know incorporate that so you cleverly map These discharge statistics into the channel network based on the properties of the channel network that the LEM is calculating And or make use of newer transport and erosion laws that have been developed in the last few years that explicitly incorporate discharge statistics and some of the Some of the parameterized distributions of discharge That's the kind of approach That I'm aiming for to bridge this hydrologic scale gap, but that's this is a very important aspect of Of this kind of work and if we want to keep doing this using these climate models to run LEMs We're gonna have to come up with ways to do this. All right, so I'm done I can get a bullet point There we go. So in summary new CCSM output For LGM in 20th century allows us to do dynamical downscaling for LGM We can look at LGM climate in the appropriate spatial resolution but Important scale gaps still exist between these different kinds of models and until we each have a petaflop iPhone in our pocket and We can integrate Million-year landscape runs at climate model time steps. We're gonna need to take some kind of statistical approach To crossing that scale gap and before I take questions. I'm going to take 30 seconds to pimp a my AG use session here We have some exciting invited speakers Nicole Gasparini is one Colin Stark been Crosby and Joel Petterson and a lot of non-invited talks that are going to be very exciting And so I invite you to check that out if you're going to be at AGU. Thank you Okay questions for Dylan who's got a solution to his Hydrogram problem any questions? No stomp stomp them. They're still writing their equations. Okay. Thanks. Thanks