 Today, I will be leading a virtual tutorial on the INVEST Coastal Vulnerability Model. We will be focusing on a brief introduction to the science underlying the INVEST model and then a live demo looking at the model inputs, the user interface, and also the modeling outputs. My name is Jess Silver. I'm an analyst with NatCap on the Marine team in Seattle, Washington, and I have been working with the INVEST Coastal Vulnerability Model as well as some of the other Coastal Marine models for a little more than 10 years at NatCap. So the INVEST Coastal Vulnerability Model is a model that is used to understand the distribution of risk of exposure to coastal hazards for people living along the coastline. So it's a model that incorporates a number of different aspects of coastal and near shore processes which include climatic forcing conditions along the coast, so exposure to waves and wind as well as potential exposure to storm surge, and then also natural characteristics about the shoreline, such as the shoreline type or geomorphology, the elevation, as well as the presence and extent of different kinds of natural habitats. So the main questions that the model is designed to help answer are questions such as, are there areas in my region that may be more exposed to coastal hazards, specifically understanding where people and infrastructure and other valuable assets are most exposed to coastal hazards in my region, and then also a focus on understanding where marine and near shore ecosystems can help reduce these impacts. The data inputs for the model are rasters and shapefiles that speak to a number of different shoreline characteristics, so I will go through these individually. But in brief, the model takes into account information about shoreline geomorphology, so the shoreline type, about the extent and type of biotic and abiotic habitats, exposure to wind, waves, and potential for exposure to storm surge, coastal elevation, also known as coastal relief, sea level rise, and then lastly, information about social and economic metrics along the coastline. So this is a relative index-based ranking model. So each of the model inputs are ranked from one, which is the lowest contributor to coastal exposure, so essentially the lowest risk category to high, which is a five, and that's the greatest potential for contributing to coastal risk. So we have low exposure, low risk with a one, high exposure, high risk with a five. And each of the inputs is assigned ranks, and the ranks are a mixture of absolute and relative values depending on the different input. And again, we will walk through each of these individually. So the first input is shoreline geomorphology, and so in this case, the model is trying to distinguish between shorelines that are relatively hard to road or wash away versus very soft shorelines. So in this case, we're really trying to distinguish between rocky, gravelly shorelines and sandy, muddy shorelines. And so here this is an input where the user is going to assign absolute ranks based on the different shoreline type. This is often a line shapefile where the ranks are assigned to line segments. The next input is coastal relief or coastal elevation, and in this case, the model is trying to distinguish between low-lying areas, which are going to be more prone to erosion and inundation, sorry, to flooding and inundation versus more steep areas, which are going to be less likely to flood in a storm. In this case, the model uses a digital elevation model or a topographic layer to calculate the average relief for each shoreline segment. And then it calculates the ranks based on the full distribution of elevation data for the region. The next input is natural habitats. So in this case, the model can take into account information about the extent and the type of both biotic and abiotic natural habitats. And it applies a user-defined rank that corresponds to the potential for different types of habitats to provide different levels of coastal protection. So for example, coral reef, mangrove forests, dense coastal forests might be given a rank of one, which indicates that they are more effective at providing coastal protection than habitats such as seagrass or kelp forests, which may provide some protection, but certainly less than the more rigid habitats. The model also can incorporate information about sea level rise, and it does this in a couple of different ways. So if there is a region such as the coast of the United States, the world, essentially if you're running the model over a region where you have spatial variation in relative sea level rise rate across the region, and you have the data to understand how sea level rise varies, you can capture that spatial variability in the model directly. So in this case, we have a number of different relative sea level rise rate trends, and we calculate the ranks based on the distribution of rates. You can also use the model to look at the difference between current and future sea level rise scenarios. So in some cases, either there isn't spatial variability or you want to incorporate both spatial variability and future net rise, and you can use the model to look at that. For reference, we have published a number of papers, including the Nature Climate Change paper here on the right that have extensive discussions about how you can use the model to look at different sea level rise scenarios. So the model also looks at exposure to wind and waves, and in both cases, it uses a globally available data set that is curated by NOAA. It's called the WaveWatch3 model, and it models wind and wave statistics for different points all over the oceans across the globes. And so we use this data set in the model, and you can exchange it for something more finer scale if you desire, but the global data set is totally sufficient. We use it to look at the wind and wave exposure for every shoreline segment along the coast, and then again, the ranks will be generated based on the full distribution of the wind and wave values. And so in this case, the model looks both at ocean waves, so oceanic generated waves, and then also local wind generated waves in sheltered areas. The model also looks at the potential for exposure to storm surges. So in this case, we use a proxy, which is the distance from the coastline to the edge of the continental shelf. The graphic below shows that when you have a long, shallow extent where waves can build up, you are much more likely to see storm surges, high storm surges, than if you have very steep drop offs. So in this case, the rank is generated based on the distance from the coast to the edge of the shelf, where you see a large, shallow extent contributing to greater risk, and a higher exposure rank, and short distances to the shelf, and steep drop offs contribute a lot less to exposure and get a lower rank. So this model is a relatively simple, transparent model, and it is also designed to be able to use somewhat coarse data sets. So there are a number of globally available data sets that can be used with the model, and it makes it very easy, relatively easy to run this model anywhere in the world. So we have included a number of the global data sets with the sample data associated with invest. These include the global data set for the WaveWatch 3 wind and wave statistics, a global map of the continental shelf, globally available DEM, and topographical information. There are other resources that we do not include with the sample data, but that are very good global data sets, including global habitat data sets curated by UNEP, the permanent service marine sea level rise has information about relative rates of sea level rise and tide gauges around the world. And then there are a number of different data sets that can be used for the population and demographic information. This is a schematic that shows a simplified version of how the model is working. So on the left, the first step is to grid up the coastline of the region that you're working with into a user defined segment size. So in this case, the coastline is gridded up to one kilometer shoreline segments. And then each of the different input data, which we just reviewed, are gridded up to the same resolution and they are ranked. And again, they're ranked based on the different approaches that we just discussed. So this could be absolute user defined ranks for some of the input variables and for others, the model is assigning the ranks based on the full distribution of values. One of the primary model outputs is something called the coastal exposure index. So in this case, this is the hazard index that's shown on the far right hand of the schematic. And the hazard index is the geometric mean of all the variable ranks for each of the inputs. So in this case, in the schematic, we've shown the hazard index categorized into high, moderate, and low. So when it comes out of the model, the hazard index is a continuous value from one being the lowest risk of exposure to coastal hazards to five being the highest risk of exposure to coastal hazards. And depending on your application, you can choose to define this categorically, such as high, medium, and low, or perhaps assign a threshold. So the second part of this model is to understand who and what is at risk along the shoreline. So when we talk about risk, we're really talking about the combination of exposure to coastal hazards and vulnerability to coastal hazards. So the components of the model that we just reviewed, which culminated in the calculation of an exposure index, these are the parts of the model that is going to help you understand the relative exposure potential for exposure to coastal hazards along the coastline. And then the second piece of the model is to understand who and what might be vulnerable. And these taken together are risk. So there are a lot of different potential social economic and demographic metrics that could be evaluated when thinking about vulnerability. So a simple place to start is population density. But depending on your application and on your region and on data availability, there are many, many factors that have been studied in great depth that help contribute to the vulnerability of different communities to recovering from and adapting to and preparing for coastal hazards. Now that we've done a review of the underlying science of the model, we'll look at a brief case study to get a basic understanding of how the model is applied. And then we'll do a live demo. So this is a case study from the Bahamas. This was a country-wide coastal hazard analysis done for the Bahamas. And in particular, this analysis was used in post-disaster reconstruction and coastal resilience building efforts following several years of very devastating hurricanes in 2015 and 2016. So ultimately, this analysis was used as one component to help inform the development of an IDB-funded climate resilient coastal management and infrastructure program. And the program involves funding a number of different climate resilient pilot projects, including restoration on several of the islands in the Bahamas. So the key questions that we're driving the work in the Bahamas are where are people at risk from coastal hazards in the Bahamas? And how might sea level rise change the distribution of risk across the country? And then understanding where coastal and marine ecosystems provide protection currently and under future sea level rise for the most socially vulnerable populations. So this is an example of the types of results that are produced by the coastal vulnerability model. So on the left is a map showing the distribution of relative risk across the Bahamas under the current baseline scenario. So in this case, you see three different categories and the red shoreline are those areas where the risk of exposure to coastal hazards is greatest, relative to the country as a whole. And one of the things that the model helps to understand is not only the distribution of risk, but what is driving risk. So in this case, in the gray, you can see the continental shelf, the extent of the continental shelf, and you may notice that areas in the Bahamas that have the greatest potential for exposure to hazards. So those are the red areas are often those areas that are sitting on the side of islands where you have long expanses of continental shelf. So in this case, the potential for the exposure to storm surge is a very important driver underlying the distribution of risk across the country. On the right is a graphic that is summarizing a couple key metrics for different scenarios that were evaluated. So in this case, we summarize the population, so the number of people living along the highest risk shoreline, so the red areas, and also the length of shoreline that is highly exposed. We evaluated this for both a current and a future sea level rise scenarios. And then we looked at a baseline scenario with all coastal and marine ecosystems intact. So that's the white and the gray speckled bar. And then we also looked at another heuristic scenario where coastal and marine habitats are removed entirely. And we assume that this is an extreme scenario that is unlikely to occur. But nevertheless, it's a very effective way of understanding the potential role that habitats are playing currently in providing protection and where. These are similar maps, but they show a little bit more of the detail provided by the model. So in this case, we're zooming in on two islands in particular. We're looking at the island of New Providence on the top. And we're looking at the island of Grand Bahama on the bottom. These two islands are two of the most populated in the country. And New Providence is the island, is the capital island. So Nassau is on New Providence. So on the far left panels, you can see the different coastal and marine ecosystems that are providing protection for these two islands. And then in the center panels, in the red areas, those are areas that are, so the areas in red, the areas in pink, are those areas that are currently at high risk with habitats intact. And then the areas in red are those additional areas along the shoreline that would become high risk if habitat was lost. So as you can see, for both islands, a considerable amount of shoreline would become high risk if you lost the protective benefits afforded by marine and coastal habitats. We are also showing you the distribution of population. So in the blue is the coastal strip of population values with the darker colors representing higher population densities and the lighter colors representing lower population densities. And so you can see that on both islands areas where people are living are also high risk areas, especially on Grand Bahama. And then you can also see that there are a number of different areas that would increase significantly in risk if shorelines, if habitat was lost along the shorelines where people are currently living. This information is also summarized in the bar chart on the far right. So in this case, we're showing the population at highest risk of exposure to coastal hazards for each island under three scenarios. The white bars show the current scenario with habitat intact. The solid black bars show the current scenario if habitats are lost. And then the furthest right bar with the white hatching shows a future sea level rise scenario without habitat present as well. In this case, the percentages refer to the fraction of the total island population at highest risk for each scenario. So for example, on Grand Bahama island, 4% of the population is currently at risk, at highest risk of exposure to coastal hazards with habitat intact. If you lose habitat, even under current sea level rise or no sea level rise, you see 24% of the population becomes highest risk of exposure to coastal hazards. And this increases further to 33% of the island's total population at high risk of exposure with future sea level rise. So some key results from this analysis. Nearly one-fifth of the coastline and nearly two in 10 Bahamians, country-wide, are currently at highest risk of exposure to coastal hazards. If you incorporate sea level rise, the extent of shoreline that's most exposed to coastal hazards would more than double, and the total population would nearly triple. So this would be more than 10% of the population and more than 40,000 people living in highest risk areas. In terms of coastal and near shore ecosystems, they occur along almost the entire coastline of the Bahamas, and there's often multiple habitats fronting sections of shoreline. So this could be coral reef backed by seagrass and then mangrove along the shoreline. And our results suggested that if these habitats are lost, even under current sea levels, the length of shoreline that is most highly exposed to coastal hazards throughout the country would quadruple. Lastly, with habitat loss and modelled sea level rise, the length of shoreline at highest exposure increases fivefold, putting an estimated quarter of the population at highest risk. So this presentation has been designed to give you a sense of the information underlying the coastal vulnerability model and also a sense of how it is typically applied and some of the questions and results that come out of the model. The next thing that I would like to do is to take a quick look at the model inputs, the user interface, and the model outputs. So as you probably know from watching some of the introduction to invest, invest is an open source suite of tools that are modular. So today we'll be working with one specific model, the coastal vulnerability model. Each of these models has sample data associated with the model that can be downloaded as part of the software or separately from the website that are designed to give you a starting place to understand how the model works. You can use, you need to use a GIS. You need to have basic understanding of using a GIS. You do not need to be an expert by any means and you can use ArcGIS or QGIS, which is a free open source GIS software. So in this case, I'm using invest 3.9 and I have downloaded the sample data for the coastal vulnerability model. These sample data are from the Bahamas case study analysis that you saw in the previous presentation and we'll just look at them very quickly. So one thing that you will need to create for every region that you do this analysis is an area of interest. So this is called an AOI and essentially an AOI is just showing the model the area over which you want it to run. So if I pull in my land polygon, I can see that here my AOI for the model is the northwestern corner of Grand Bahama Island. And because the coastal vulnerability is a relative model, the results are going to give you an understanding of the distribution of risk relative to all the other shoreline segments in the model. So other inputs include information about the coastal and near shore habitats. So in this case, we have information about the distribution of mangroves, the distribution of coral reefs, the distribution of seagrass, and the distribution of coastal forest. We also have in addition to shapefiles and rasters, the invest models often utilize CSVs to help give information about model parameters. So there is a CSV that is associated with there is a CSV that is associated with the natural habitats, which gives the model a number of different pieces of information. These include an ID and a pass for each shapefile. They also include the rank associated with each habitat. So as I said, some of the ranks are absolute values that are assigned by the user. In this case, natural habitats fits into that category. And then there is also a protective distance that corresponds to the distance over which the model should look out from the shoreline to find a given habitat. So this is essentially a computational shortcut that helps the model understand which habitat types are providing protection for a given coastal segment. And we have more information in our user's guide to help you understand how to assign both ranks and protective distances. Other inputs are the symmetry, information about the symmetry, information about the continental shelf. So in this case, this is one of the global datasets that's available with the invest model that you can use anywhere in the world. So it's a map of the continental shelf. There is also information about the topography. So in this case, it's the SRTM 30 meter topographic map. And this is what the model is going to use to calculate coastal relief. There is also information about the coastal geomorphology. So in this case, this is often a dataset that is either hard to come by or incomplete and requires some creation effort on behalf of the user. So in this case, the key, I'm going to try sharing my screen again. So in this case, the information embedded in the coastal geomorphology layer that the model is using is going to be information that you put directly into the shape file itself. So in this case, the important column is this rank column, and it's going to tell the model how to rank each shoreline segment. So you can see here, we have different shoreline types. In this case, mud and mud is assigned a rank of four. If we scroll down, we'll see that Sandy Beach is assigned a rank of five, and rocky shorelines are assigned a rank of three. Now, it may be the case that you live in an area with good information about shoreline geomorphology, and there are a lot of different classifications. The model really only needs to distinguish between very broad, categorically different shoreline types. So rather than trying to figure out how to reflect the difference between large gravel beach and medium gravel beach and small gravel beach, really what we want to understand is the difference between rocky shorelines, gravel beaches, muddy shorelines, sandy shorelines. So it's a fairly coarse representation of coastal geomorphology. The WaveWatch 3 dataset, again, another globally available dataset associated with the model. As I said, this is a grid for the world, and each of these shoreline point or each of these grid points has information about wind and wave statistics that the model uses to calculate the wind and wave ranks. And then the last input is the population. So again, this is going to be a raster dataset that's going to give the model an understanding of where people are living. And if we symbolize it slightly differently, it becomes clear where the populated areas are. So those are the inputs associated with the models. And now we will look briefly at the user interface and then we will come back and look at the model outputs to give you a sense of what that looks like. So in terms of the model interface, if you go to your start menu and scroll down to invest, I have a number of different versions of invest, you'll see each of the different models. So invest is modular. And so each of these models has its own user interface and its own modular workspace. And if we click on the coastal vulnerability, it takes a moment to open. In this case, I have some outputs pre-populated in from another analysis I'm working on. So we can clear those or not. All right, it's working just slowly. This is a very typical invest user interface. So it's going to give you a mix of spatial data, rasters, vectors, and then it's also going to ask for some CSVs. And then in some cases, it'll ask directly for some parameters. So there are a number of different ways to navigate to your inputs and input them into invest. You can click on the folder icons or the file icons that are on the right side of each input. And when you do that, it's going to take you to your folder directory and then you can navigate to an area that you want to save your results. So in this case, I might go to my invest folder and save it in the sample outputs folder. When you're working on a specific application, this might say, you know, my region baseline, no sea level rise, and the date or something. So your workspace is where the model is going to send all the results and the intermediates. And you can optionally add a suffix in to help with your organization scheme. So then each of these various inputs hold on. So each of the inputs in this case, we are going to load the area of interest. So we will navigate to our through our Windows Explorer, navigate to the invest folder. And then I can see here that I have the sample data, close to vulnerability. And then I'll be able to see all of the different files that are in my sample data. So in this case, I would load the AOI, that's my area of interest shape file can load my AOI. Everything you'll want everything to be in the same projection. Here we need to specify the model resolution. If you're running your model at over a country wide scale or a large region, I might recommend running it at a kilometer to start. If you are running it for a smaller area, we often run it at 250 meters. Ultimately, your resolution is going to be dictated by the quality of your data, the computational time required to run the model and the size of your region of interest. So for example, to run it one kilometer, I put 1000 meters. To enter in my land polygon, again, I'll navigate to my sample data folder. And I will look for my land mass polygon shape file. So each of the inputs can be input in the same way by navigating to the specific files. If you need information about a given input, you can click the information tab and it's going to point you to some information related to that input. Each of the inputs and the way to create them is described in the user's guide chapter for each model. There's also a way to save and load parameters by creating these JSON files. So in this case, with the sample data, with invest, the download of invest came these JSON files for each of the models. So I could load the coastal zone grand Bahamas invest JSON file and it would go ahead and populate the inputs for me in order to run the model. So I don't have to run each of them myself. You can also create a JSON file once you have loaded your whole, all the inputs for your particular model application. You can save as and you will be able to save your own JSON file so that you can always easily load your inputs without clicking through the interface. In this case, we've tried to make it clear whether or not an input has uploaded correctly by providing a red X if you need to revisit the input and a green check mark that shows if it's been correctly uploaded. Once everything has been produced or loaded into the window, you click run and the model should start running. So in this case, I have already run the model and we'll look briefly at the outputs in the NRGIS. And as you can see, my computer is very slow. I'm sorry about that. Let me put back up the LAN polygon and then we'll look at the outputs. So on the right here you can see this is what will be produced by the CV model. So there's a file here called or a folder called intermediate. In the intermediate are all the different individual inputs and associated information for each one. So for example, for all the for the wind wave, you can see a number of the different steps that the data produced by different steps associated with calculating the wind and wave parameter. The model produces something called a log file. And when you are trying to troubleshoot why you were getting an error or communicating with the invest software team for any reason, you'll want to refer to this log file. So it's going to tell me my inputs. So what I what my inputs were for this particular model run. And then it's going to give me a lot of information, which I don't need to understand. And then ultimately at the end, if you were to get an error, there will be some description here that may help you to understand what caused the error. So even if you have a run that ends with an error, a log file would be created and that is going to be one place that you're going to want to look to see if you can track there. So the outputs, the main coastal exposure outputs are going to be in this coastal exposure geo package. And if I pull in the information, what you will see is shoreline points that correspond to the segment size that you specified in the user interface. And when I open up the attribute table, you'll see there's a lot of different information in the attribute table. So there's a unique identifier, a shore ID associated with each point. And this is going to be very important if you want to do post processing and then join it back to the original table. And then you have the different information, the different ranks for each of the various input variables. So this is our underscore hab. So this is the habitat rank, the wind rank, the wave rank, third rank, relief rank and geomorphology rank. And as I said in the introductory talk, part of what the model is designed to be able to speak to is not only where is risk distributed along your shoreline, but what are the drivers? So understanding the underlying ranks for the input variables are going to be really important in figuring out the different risk drivers. You're also going to have information about the number of people. So in this case, this is where you have null, it is likely that there's no data for that cell, but the values are the average population density within a search radius for a given shoreline point. So as I said previously in the talk, the main output for the model is this exposure index. So this is the continuous variable ranging from one to five. So here we see we have an exposure index associated with each shoreline point. This is a metric which ultimately many times people choose to categorize further. For example, you could convert this or look at the different quartile breakdowns here and consider each quartile to be a risk category. And instead of showing the continuous variable, show something like high, medium and low, for example. The exposure no habitats is a scenario that is automatically run by the model. And this is the scenario in which we assume that all coastal and marine habitats are gone. So this is that without habitat scenario that I was showing you in the Bahamas case study example. And again, this is designed to be a an extreme scenario that is a bookend for understanding the protection provided by coastal and marine habitats. And then the last piece of information is the habitat role. And this is the difference between the exposure no habitats and the exposure index with the habitats intact. So for example, if I were to symbolize my sample data set by a few of these categories, I could symbolize it by coastal exposure. And so in this case, the purple colors, the darker purple colors are going to be areas where you see greater coastal exposure. So again, we're seeing it on this northern shoreline of the island. And my guess is that it would be driven by the fact that this is a slightly more low lying than the other shoreline. And it is also on a wide shallow bank with a high potential for storm surge. It also does not have the benefit of having a fringing coral reef close to shore. So if you'll remember the habitat data, this southern shore has a reef right offshore and the north shore does not have have a similar reef. If I were to instead plot by the habitat role metric, this is going to show us those areas where habitats are playing the greatest role in providing protection. So these are areas where you particularly see the importance for protecting existing habitat is in areas with high habitat role. And then if I wanted to look at each of the underlying driver, for example, maybe I wanted to see the storm surge ranks. In this case, the five again is the greatest contributor to storm surge potential, the greatest contributor to risk in this case, based on the distribution of storm surge. One is the lowest potential for exposure to storm surge. And here we see the importance of that underlying driver. So a much lower potential for storm surge on the southern coast and a much higher potential for storm surge on that northern coast. And again, that is driven by this continental shelf distance. So I'll end this tutorial here. Thank you for joining me. And please do not hesitate to get in touch with us if you have questions. Thanks so much.