 I want to start out by thanking James for inviting me to this really fun conference where I get a chance to hang out with people who think that modeling is actually unacceptable and maybe even a necessary scientific activity and not just video games for academics. And I also want to thank Marlene and Eric and Albert who've made this conference appear to go really smoothly and effortlessly, I think that's important. Modeling surface processes requires simulating changes across both space and time, of course. The work that we're doing focuses on the role of humans in transforming the Earth's surface. Like many of the presentations we've heard over the last several days, this involves integrated modeling of landscape evolution, sediment transport, terrestrial morphodynamics, but we explicitly seek to do this at a human scale. And so what are we talking about in terms of spatial and temporal scales that are most relevant to modeling anthropogenic surface processes and landscape change? Well, you can argue various ways, but I think really the beginning of major human impacts across many of Earth's surface systems really starts in the early Holocene with the adoption of farming and herding and settled village life. Effects at this time begin that span landforms, soils, sediment flocks, freshwater systems, and even atmospheric conditions. And these impacts have continued to grow quantitatively ever since, as we've just heard, such that today humans move more sediment than all natural processes combined in terrestrial environments. And so relevant spatial and temporal scales for human impacts are variable, of course, but they're much more limited than the range of potential range of other surface processes we've heard about. Timescales for surface dynamics can range from a few years to millennia, but most take place at the scale of decades and centuries. Similarly, spatial scales can vary from household to global, especially today, but for much of the human past, until maybe the 20th century, perhaps, most important spatial scales were very local to the level of regional watersheds. The research focus that we have is on the impacts of agro-pastoral landscape, land use, on landscapes and societies, and the landscape transformations that accompany the origins of these kinds of land use practices, farming and herding. Importantly, human-caused landscape dynamics also have important feedbacks back into human social decisions and the subsequent land use decisions that take place in it. So there's this feedback between landscape evolution, social practices, and back into landscape evolution. And these have formed complex coupled social-ecological systems that have become pervasive and global in today's world, and that's the world we live in now, with humans being keystone species. And so today I'm going to try and touch on some of the approaches that we've had in modeling these complex social-ecological systems and human-environmental interactions, focusing of course on surface process models. The modeling platforms we use are diverse, and we make extensive use of open-source technology. We couple numerical and numerical modeling with discrete event and agent-based modeling. We use GRAS GIS, Python, and Java primarily, with some other things thrown in. And because there's an interest including GRAS in the CSDMS tool set, James asked me to say a few words about GRAS and its potential for surface process modeling. GRAS itself is a wonderful 1980s acronym that supposedly stands for Geographic Resource Analysis and Support System. I'm not quite sure about that, given the dates. But anyway, it's a full-featured GIS, Spatial Modeling Platform. It's open-source, it's multi-platform, and Earth scientists are really common in the dev team and on the user base. It's an international project, so there's lots and lots of input from people in other countries. Among its many features, it has extensive raster and grid support within a GIS environment with many kinds of raster and grid-based tools. There's lots of very fast built-in tools for surface process modeling, things that, including surface transformations, hydrology, and surface flow, groundwater, solar insulation and radiance, interpolation. Has a really good map calculator with lots of functions from matrix operations. And all this takes place in C, so it's very fast. It also has tools for remote-sensing imagery manipulation, landscape analysis and geostatistics, and has good support for vectors with things like LiDAR processing, network analysis, and tessellation. One thing that's interesting for here, it has true voxel support, so if you want to do stuff in real 3 and 4D, you can do it in graphs or it's even a 3 and 4D map calculator from matrix operations in multiple dimensions. Just being released now are a new set of modules in temporal GIS that combine time and space in kind of an integrated way for analysis. All of these functions are available as command line modules, in spite of the fact that it has a really nice GUI, if I do say so, but because they're available in command line modules, that means grass is highly scriptable in many languages. There's especially strong Python support, and of course the source code is modifiable. It's in C with additional things in Python, so it's easy for people to work with. What I want to do now is turn to how we've used coupled grass GIS and other modeling platforms to create a computational modeling laboratory for studying the long-term interactions of agri-pastoral land use and landscapes, and we've focused in the Mediterranean. This is an NSF-supported project with high-resolution study areas at opposite ends of the Mediterranean. You can see more at the project website, if you'd like. As an overview, how we're dealing with surface process modeling, that's the part I'm going to spend most of my time on, is that we start with a DEM of topography. We calculate net erosion deposition for every cell, then we add that back into DEM and produce a new DEM that becomes the base for the next cycle, and then we can iterate this. We work at annual cycles, we can iterate this for decades or centuries and do it quite fast. As we all know, surface processes vary depending on where you are on the landscape, and so we use different algorithms for modeling processes in different topographic settings. So for drainage divides, ridge tops, and hill tops, we use the standard diffusion equation. I'm only going to put in a few equations here, as Rudy Slingerlin said the other day, just enough math to be dangerous. Diffusion will give us net vertical change. When we get to hill slopes, reels, and gullies, we use a modified version of unit stream power erosion deposition that's been modified for 3D landscapes by Helena Mitozova, one of my co-authors, and it has different exponents, the M&M exponents depending on whether we're in hill slopes or reels and gullies. When we get down to channels and channel areas, and these aren't the rivers that we've seen in some of the other presentations, we switch over to shear stress equations. We calculate these on the basis of individual storm events and then multiply times the average number of storm events in the year to annualize. In order to get shear stress, of course you have to know water depth, and so we get that from an idealized hydrograph that's estimated from accumulation, from each storm event, and with the base of it, we estimate it's a normal curve, and the base is the time of the event, which we calculate in hydrologic instance, so that's related to the grid cell size and the actual accumulation value. The height of this curve is what we use as water depth for estimating shear stress. When we put all this together, the basic assumptions are that in the system we're dealing with, most of the time water carries sediment at capacity, so we can estimate transport capacity from sediment flux in these systems, and so anything that changes sediment flux and hence transport capacity will change transport capacity, so if water slows down at the base of slopes capacity drops, if water speeds up when slope increases, or if it slows down when it hits vegetation or speeds up when it gets away from vegetation, this changes the potential ability of water to carry sediment, and so then water will either erode or deposit until it comes back to this dynamic equilibrium of its potential capacity, and so we estimate erosion deposition rates from the divergence of these transport capacities across a landscape from cell to cell, and then can calculate net vertical change from that divergence, so we can move. Okay, so if we're using different algorithms in different part of the landscape, we need to be able to switch from algorithm to algorithm in some automatic way. Traditionally, many people use the relationship between slope and accumulation to see where to switch between, say, diffusion and other kinds of processes that produces what's sometimes called a boomerang curve. We found that as being not very useful for what we're doing, so we've come up with another way we think is a better method for switching between different algorithms, and we look at profile and tangential curvature rather than slope. So if you can think about it, when curvature is highly positive in both directions, you're on top of hills, and accumulations are low, so it's positive and you're on top of hills. Let's see if I can make this go. There we go. So this is where we do ridge tops, hill tops, and then when curvature starts to, there's an inflection here and moves toward the negative, now you're moving toward negative profile curvature, but it gets to actually close to zero, so it's pretty straight. No changing curvature. Accumulation goes up, you're on hill slopes. This is supposed to happen wherever, I guess I can push the arrow button here. And then when it inflects again here, and you start to move back toward the positive from the negative, accentuating the positive, and accumulation goes on up, we move to, this is where we get to rills and gullies, this is where we have the maximum negative profile curvature at the base of slopes. And then finally, where you have an inflection in tangential curvature, cross sections like this, it's highly curved and starts to flatten out. We move into channel areas up here. So we turn to rule-based modeling at this point to switch between algorithms. And so what we do is we model each of these phases separately, and then we patch everything together automatically. Because this is all modeled in grass, it's easy to use in that. You can get the script. It's a Python script that runs in this environment. Because it takes advantage of grass's fast, sea-based modules and processing, it goes very fast. Our benchmark is a million cells because we model at high resolutions, like five-by-five meter cells. And we can do an annual cycle of a million cells in around 50 seconds. We have this tune for annual cycles. You could retune it for monthly, daily, or even event-based cycles. And we're talking about doing this on a desktop computer, something you could end up best buy and get. So here's what it looks like. This is a landscape in the Penalguida Valley in eastern Spain. It's what it looks like in real life. We've put a couple of little agricultural villages on here that have done a little bit of farming, a little bit of herding, and we can see what this looks like over the course of 50 years. So they haven't done a lot. Red is erosion. Blue is deposition. We can watch this change here. And we can see how rich tops start to show up with erosion. But also, we get erosion in the broncos here, starting at one of these villages, and we have incision going downstream from this village. We're also picking up deposition at the base of hills in these areas here. If we run this for longer, we get more severe effects, obviously. And of course, if people get a chance to do more things, we get more severe effects. And this is what it looks like in a nice 3D, eye-candy view. So this one script, which is called Our Landscape Evil for Landscape Evolution, is one piece, one component of this hybrid modeling laboratory that includes a Java-based, Asian-based model of human households and land use decisions, along with the grass GIS-based modules for landscape dynamics, and then regression-based models for climate and for vegetation. We're using, of course, open-source software for research transparency and for global accessibility, so it's available to people. And so this kind of a coupled modeling system, which is controlled from the Java-based GUI, where you can put in information on land use here, there's other pages. This allows us to carry out experiments in the complex interactions of socio-ecological systems and look at long-term anthropogenic change in Holocene landscapes. And this is giving us some new and sometimes counter-intuitive insights into emergent, coupled human and environmental properties and processes. I'll give you two quick examples here. One is some modeling we did a couple of years ago in Northern Jordan. We looked at the effect of community size in these Mediterranean landscapes, so we can start out. We have a tiny little hamlet that people do shifting cultivation. They have sheep and goats, they're herding, and they cultivate mainly in the wadi bottoms, and then they have run sheep and goats in the uplands. And of course, when we model this out over a period of time, it creates erosion. It creates mainly in the uplands, but most of the sediment from that erosion, which is this green, oops, backed up here, this is now working, it didn't work before. You get a chance to see these slides again. The erosion, which is the green line here, this is cumulative erosion, is not a whole lot more than the cumulative deposition, which is the green area. So over half the sediment that's eroded gets redeposited down in the wadis where people can use it to farm. And so the little bit of landscape degradation is actually beneficial and increases, at least maintains to increase productivity. So these people do fine with this strategy. But if people carry out the same strategy, exact same kind of land use, and just get a little bit bigger to a village, then the system goes through a phase shift and here's erosion here, and the redeposition is the yellow area here. And so erosion is several times greater than deposition. They started losing cultivatable land. And so the same practices end up in a loss of productivity, which gets worse over time. And so without changing anything, and of course, people's first response is, gosh, things are bad. We should do more of the same. It worked for grandpa, right? But it doesn't work. It gets worse. So this was interesting. Another study that's just coming out is the effect of community placement on the landscape. You put settlements in places on the landscape for various reasons. So we put four little villages, actually the same village, in four different places doing the same thing. So one was placed in flat air, kind of a nice terrace, which is a good place for agriculture. One was placed down in one of the broncos where it has access to water, and it's kind of hidden a little bit. One is placed up against the base of some really steep slopes where it's good for defense, and one we put up on top of a hill, a ridge, where it has good visibility over the landscape. So there's social and economic reasons for placing these villages in different places. We have them all do the same thing and see what happens. And I'm not going to give you the details of all of them, but the thing that's really interesting is that you would think that this one here, number four, which is up on top of the ridge, would have the highest risk for landscape degradation. In fact, it has the most sustainable agriculture of the areas. It's not as productive as the other sites initially, but what happens is they end up grazing in the areas nearest the site on steep slopes, and they farm. Whoops, ooh, here, let's push this again. Maybe I'll give up on the remote thing here. They farm down at the base of the slopes where the land is flat. And so there's a little bit of erosion on the slopes due to grazing. It re-deposits down at the base of the slopes, increases productivity, and they can keep this up for longer than any of the other sites. So not quite what you'd expect. You can read some of these in these papers. You got all that, right? You can go to our website and pick some of this up. What I want to do in the last few minutes I have here is talk about something quite different and appropriate for here, and that's a new community of practice called the CompSys Network that I and some colleagues have started. It's a network for building capacity and promoting best practices and computational modeling. That may sound familiar here. And our focus is on the dynamics of social and living systems, so as socially ecological and life sciences. We're doing a number of things to try and promote new kinds of interaction and information sharing. One of the things we're doing is creating a computational model library. That should sound a little bit familiar to people here. And again, these are for social and ecological models. We have about 100 models in the library currently. We're trying to link these models to journals so that journals require publication of the models in a library when you publish a paper on the journals in the journals. And at the same time, for models that are published, we're going to start offering permanent handles. This is like DOIs, so that there'll be resource locators for published models and ones that we go through a certification process that we're starting up. We would like to have a certification process, which is kind of a peer review process. So there will be panels that can review. You can optionally have your model certified. And we're hoping to at least be able to say that it works. It runs, and it does what it's supposed to do, and maybe even gives good results. That would maybe be a gold star certification or something. And again, those models would get citations. And so we want to also, as James said the other day, promote methods of citing models so people get academic and research credit for the work that they do, that this is evidence of research activity. And so we're trying to come up with citation protocols for models also as part of this. This is another look at the library here. Of course this partially meets NSF data sharing requirements. Some of the other things that we're doing, of course we're having workshops and presentations to professional audiences. And we're not going to try and be a data repository. There's a number of new data repositories opening up with NSF data net projects that are coming online. And so our plan is to try and link to some of these to identify standardized data sets that can be used for model testing and validation. So rather than do it ourselves, which was our original plan, we see these things coming online. I think we're going to try and become nodes in a couple of these data net projects as a way to access data sets for testing. We are also establishing standards for computational model metadata and description, focusing especially on Volcker-Grimman colleagues, ODD standards for agent-based model descriptions, providing educational resources to help people teach modeling, computational modeling, agent-based modeling, and social science and ecological science programs in higher education, and setting up a framework to have special interest groups in online communities. We're just opening up membership now for a lot of these things. We have an initial pilot program to try and provide high performance computing access to members. We'll see how that goes. I mean, it's not quite the same as you guys are doing here. And so I'm not sure whether it'll work well or whether it won't, but we'll give it a shot and see what happens. We started this project with a pilot project in 2006 and we're funded just a couple of years ago. COMPS just started officially last year. We're opening the membership now. We've just got a new membership statement that we'll want people to agree to. And so sometime in the next month, that'll come online. As you might guess, this community faces a lot of the same issues as CSDMS in terms of sharing knowledge and of modeling, which is, you can't do as well in journals in the regular journal format. So you have to come up with other ways of sharing knowledge about these new kinds of technologies. Trying to find a way to archive and make accessible models. We're trying to deal with ways to make sure people get recognition, proper recognition for model development and try to come up with ways from model validation and benchmarking. And I do have to say that the vision of James and others here at CSDMS has been inspirational in helping us. As we talk through some of these issues, it gives us a chance to look and see what other people have done. This community is similar enough that it's been very, very helpful. So in the future, given the importance of understanding the role of biological processes and social action and surface dynamics, we hope to have a kind of an active partnership between COMSIS Network and CSDMS. We think that's mutually beneficial to understanding these kinds of complex interactions between socio-ecological and biophysical systems that James mentioned on the first day here. Thank you very much. So we've got time for some questions. Yes, Rudy. Systems have struggled with the whole certification issue. It's something that we wanted to do, but we don't know how to do it. We don't know who should do it. We don't know who's gonna set the protocols. What do you do? Well, we haven't done it either. Well, part of the way we're going about it, we think, is to do it, to some extent, the way science always certifies research, and that is to have peer review. So our plan is to set up peer review committees from the community that would review models. You'd submit your model for certification, and then they actually have to work. Now, this is the issue. People have to work a little bit. They have to go in and look at the model, try it out, and see if it does stuff. We're gonna set the bar kind of low at first. Limited initial certification means that it works, and it works the way it's supposed to work. Validation is a more complex issue, and that requires standardized test data sets, and that's even more complicated in socio-ecological systems that it is in surface process, process modeling, because the potential is enormous. So that's another reason why we decided not to provide our own data sets, but to start making links to these other places that have data sets out there and start using them. So that's the plan. We haven't done it yet. Malish? I plan to set up some sort of system that would reward reviewing. I'm coming from a point of view. I review papers like 20 plus a year, and people talk about how much the peer review problem is in kind of limbo. It's not satisfactory, and part of the reason I see is that it doesn't have the rewards. Nobody else knows that I've reviewed 20 people of papers. How would you have some sort of system to reward those who review models? Well, the cynical answer is we would reward it in the same way you're rewarded academically for reviewing papers. But we actually have thought of this, a lot of us in this group so far do social modeling and think about social decisions. And so we've thought about using social networking tools to start giving some rewards, maybe not academic rewards, but rewards within the community. So there are social networking tools where you contribute different things and you actually get a membership rating. And the more good things you do, the higher your rating gets. And some extent, contributions would allow you to be more involved in the community at higher levels. This is a little tricky. There's ways of doing this that are out there now. We haven't, we've talked about them. I don't know whether this will work or not, but it seems to be incentive for people to, you know, like if you go on sites where people do reviews, tripadvisor.com, different people have different review credibility scores. So we're at least exploring those ideas. Guillermo? Quick comment. Just along those lines, there are 18, there is one agency in Canada that pays for reviewing proposals. I'm a living proof. Pardon? There's an agency in Canada that pays for reviews. Oh, okay. That's nice. Thank you, Michael. Wonderful.