 Hello. Welcome to the Invest Crop Pollination Model training video. This will be a training from me, Eric Lonsor, and Chris Newtonboom. I also want to acknowledge some of the original co-authors of the model and the kind of paper that this was based off of Taylor Ricketts, Neil Williams, Ray Winfrey, and Claire Kremmit. And so I'm going to talk really briefly about the importance of pollination, the theory about mapping the service, basically the production function, and how we value the pollinator providing habitats, a little bit how we've used the model, and then Chris will give a brief demo. Okay, so if you're working on this, you probably already know and recognize that pollinators are really important. They provide up to 35% of the global food supply by improving 75% of globally important crops. They provide biodiversity. They're also really important. They obviously support pollinations of many, many plants, 60 to 90% of all angiosperms require some form of external pollination. Most of the time this is going to be bees. And it's been shown to be a limiting factor of reproduction for many natural plant populations. Evaluate in a service. This is just another example of an ecosystem service or those benefits that nature provides people, nature's contributions to people or NCPs. It might be another way that you think of them. And there are two key steps to quantifying them that I'm going to go over this ecological production function, and then the economic valuation. The production function links the ecology of the system with the particular service. So in this case, if we're thinking about strawberry production, we're going to model just that plant pollinator interaction, the bees landing on the flower, providing pollination that leads to some sort of fruit. That fruit has value to people. So if you go to the market and purchase it, we can kind of then link that value that's going to be provided to society back to the service being provided by the bee itself and we do that for management decisions. We're interested in how changes in the landscape would translate into changes in dollar values or some sort of value. And we can break that down by understanding how the pollinator system changes with changes in the landscape. And if we can understand that we can connect it then with how changes in yield are affected by pollination and then obviously the changes in dollar and by putting that all together, we get the full component of changes in dollars to landscape. So that first part that I'm going to talk about is this ecological production function, which here is the model and you can read more about the details of that model in the invest guide. So here is just a training session of the bees to make sure they follow along exactly what the model is doing. We want to make sure that their ecology is, you know, the bees kind of understand what we're trying to do. I'm kidding of course what we're actually doing is trying to represent the ecology with the model. And so here's how we were thinking about it. So here is an example landscape. We can imagine bees in a nest right there and this is a representative landscape, the yellow kind of in the middle might be a crop of some kind that requires bees, and there's wildland in the surrounding so if we have a nest site where the bee is. We can represent that bees moving around the landscape we can assume that forage resources closer by are more important than those farther away and so we have this distance decay function. So we can assess the quality of that landscape. And we also know that sometimes bees vary or we know that these very some go really far, some go much shorter. And so for this one if it's a be that doesn't fly very far it may not be able to get into that wildland which may provide lots of great resources. If a bee can go farther it can it can access that. And so we can kind of model the fitness of the bees in that nest site by where they are with in relation to those resources and we've done a lot of tests in agricultural systems and I just want to talk a little bit briefly about how we've validated the theory in the model. This was there an original study in Yolo County California and the Central Valley where there's lots of different crop types. And here, my colleagues sample bees and watermelon fields that's where those black dots are. And the way the model works. As first we translate that landscape into different nesting guilds. So we can translate that land cover map into four guilds stem nesting cavity nesters would and ground. And then the here the green represents good quality and the reds are poor quality. All right, so these are, and we'll get into sort of the tables that are required to translate that. And then we convert that same landscape into floral resources and the reason it looks a little fuzzy is because we're modeling how bees are moving. And so this is a distance weighted convolution it's called that weights floor resources in different seasons so here you can have spring summer and fall. You could just have one season if that's all you know about. And this model then basically adds up the quality of those four resources in the landscape. Yeah, and then we put those two layers together to create this integrated assessment of the overall quality of the abundance coming from those nest sites. All right, and then we can predict based on the surroundings of each of those nest sites, how many bees might be visited. Now emphasize this is a relative score, we don't know exactly how many bees are out there that that's just sort of, that's some data that we just don't have we're not exactly sure what the true number is so it's a relative score from zero to one. So we call that a pollination service score. And here are three tests that have been done at first, and we showed that for modeling native abundance that we got pretty good fit in California and Costa Rica, but in some landscapes it does it doesn't always work in every landscape. So and this is in New Jersey and Pennsylvania and Eastern United States, we just have never been able to get a good fit there. And after that we wanted to make sure that it wasn't, you know, the model works more generally. And so we did a global test of this of the production function on many other crops and landscapes around the world. So this was a paper led by postdoc of ours Christina Kennedy, who looked at 39 different studies across 23 crops and six continents, and the model generally worked. There's a lot of variants but overall we do find this nice strong positive relationship between the pollinator landscape score and the abundance of bees on there. And I just want to point out here that it the management of those fields matters. And so if it's a simple lands conventional landscape. That's kind of a more intense there are generally fewer bees. So there's this interaction of local management with the landscape around it. And as the landscape quality increases, we tend to see more bees on those fields. Okay, so that's the kind of ecology of the bees kind of getting to those fields. The next part is the valuation. So how do we actually translate abundance into valuation. All right, and so here we have a bee drinking coffee. But really we're going to be talking about the, the benefits to coffee production of the bees themselves. I am colleague of mine Taylor Ricketts did some great work testing the model in Costa Rica. Here we see a relatively large coffee plantation in the middle of a landscape, and he sampled sort of from the edge, natural areas, deep into the coffee plantation to look at how distance from that edge affected production yield and he did a pollinator exclusion which we'll get to in a minute. And so to get to valuation. We have to kind of model the bees coming from those nest sites as I mentioned so in this case imagine that the yellow box here is kind of the crop we're interested in, and they're going to come from the surrounding forest. Okay, this is actually really good for resources for bees in Costa Rica. And so we first model the source abundance of those bees. And we imagine that they're going to fly to that field. And then we'll be able to kind of test that the effect of their abundance on crop yield, and then based on where they're coming from redistribute that value back to the landscape that's that natural capital. Okay. In the landscape. Okay, and so what Taylor Ricketts did. He actually looked at the abundance of bees visiting the crops and those sites again we're on the gradient and found that as the abundance index increases we see a really nice fit there and that was one of the original studies. And this was the actual effect on pollinate pollinator limitation. The production which is measured in this unit called a funnega in Costa Rica that the production of the coffee actually increases with increasing pollinator visitation. And so we can then translate that again this value back into the landscape, which we'll get to a man so here's then this abundance map in corn and coffee rather. This is the actual predicted visitation 107 bees from zero, and we can see there's a strong edge effect more wild bees are expected along the edge. Okay. And then if we can then retranslate in the actual dollars per hectare coming from wild bees. Okay. And then one of the things we can do with that is actually start making management decisions sort of what is the value of that pollinator providing habitat, who would be most vulnerable it's like landscape were to change. And then what might be the end of flip side of the value of restoration. And we can do that by destroying imagine that we're destroying forest pixels virtually converting them to pasture. And then looking at the change in landscape this is a marginal value approach. And we can see that in areas like this peninsula of forest here is probably the most valuable it's almost acting like a little natural beehive kind of inserting itself into coffee. And so that's where it would actually be $700 per hectare per year lost, if you were to destroy that forest. Okay, so high marginal value usually usually happens where there's already a good source of bees. There's lots of the demand for that being the certain terms of the coffee, and there's few other substitutes so there's no other landscape around there. And we can also do the same thing like if we were to lose forest, or convert them we can change production this is the change in value. So sort of the risk. This is where in the field, you'd actually be losing that value. And this is sort of the vulnerability of the crop, sort of the opposite of this there's lots of coffee and lots of demand. There is few opportunities for kind of saving that and other forest. All right so basically the point here is that the value of the service requires there to be some user, like demand is needed. Okay, and then we can also do this to look at the value of restoration. Maybe we could explore converting some of the crop into forest. Okay, it turns out for coffee that's probably not going to be great, but you can do this for other crops. Coffee does not highly dependent on pollinators, but some other crops are. So here we can start using these models to evaluate what the economic value would be so you can get up to $70 per hectare in your highest restoration values where there's going to be lots of coffee around and far from existing sources. Okay, so there's relatively low supply of the service. Okay, I'm going to pivot over to Chris to kind of talk about the required data how you run the model yourself using the invest tool. I'm going to talk about the land use and land cover map that's needed. The attribute table that is used to translate the land cover into habitat quality for bees, and then how we deal with multiple types of species. So I'm going to stop sharing and turn it over to you, Chris. All right. So, as Eric has laid out the invest pollinator model has a relatively simple list of inputs but a lot of complex ecological math underneath the hood. So we'll be walking through today, what you plug into invest, how you work with that, how you might find sources for that data, and what it looks like so that when you are taking this into your study area, it should be at least conceptually simple to access. So this is what the invest user interface looks on my computer it may look different for you depending on the version that you downloaded and are working with. But there should be a list of several files that you need to provide invest to run. I've already pre populated each of these with files on my machine, but you can simply click and drag different required data points onto this interface to make it work. I'm going through what each of these mean in a bit. I'll show you this is a the workplace a file that you want to drop things in when the model is done you can just click and drag and populate right there. So that's a interact with invest. But before we get to just hitting the run button down here in this nice little corner, we need to go through what each of these components of the model are. To start, we're going to look at our land use and land cover map this bit right here. We're going to go back to Yolo County in California. And this is what a typical land cover raster of crops in the US looks like this is called the cropland data layer. I think this is 2019. So you can zoom in you see this is a raster data set so it's a gridded continuous data set with different values representing different land covers on the map. In general the CDL is very detailed crop inventory. I would say for visual representation of colorful colors are typically crops you can see kind of the patchwork quilt here in the Central Valley and the more muted earth tones typically represent wild lands. So you can see on margins and in the hills on this area there's lots of wild lands surrounding the Central Valley of crops. So this is the sort of file that you can provide to invest to represent the land use. However, invest doesn't naturally know anything that relates this land cover map to pollinators in itself. You need to provide what is called the biophysical table that links grassland tree corn soybean almond whatever crops you have on this layer to expected resources for different pollinating species. So that brings us to this land cover biophysical table in the invest interface here. We have an existing biophysical table for the cropland data layer that we've used. It's been validated by several studies and so you can find this in the literature. And this is what it looks like for this version of the invest model. Again, sometimes the header column names might change in between versions of the investment model as code changes in the back end. So always check the user guide for the version you are working with. These are the titles for what we are currently working with here. And what this does is this shows the land cover codes so the value of that raster pixel linking to the name or description of that land cover type so we have lots of crops here. And you can see there are shrublins forests pasture lands. And we link each of those land covers to different types of resources for pollinators. We specifically have two main drivers of the model it's nesting resources and floral resources nesting being various types of pollinator nesting habitats. Here we have ground nesting cavity nesting stem nesting and wood nesting variables, and each of these numbers is an index of quality. So higher numbers from zero to one indicate higher quality nesting sites of these types of nesting habitats. You can collapse many of these categories into a single generalist nesting category if that's what's most important in your study area or you can expand it to be ground nesting cavity nesting etc. The model will respond to correctly labeled headers in that case. Another resource that this table links land cover to is floral capacity on the landscape so each of these land covers has a different amount of floral resources that bees can use to feed themselves and to promote additional be abundance of landscape. Again it's a zero to one index one being the best amount of floor resources the landscape could provide. And each land cover will have a different amount of floral resources. These, you notice that there are three different columns here. These are different time periods across the year. So we have spring summer and fall floral resources for each of these crops and land cover types. And this has been parameterized for the continental US. You can also collapse this to a single, single column for an entire annual representation of pollinator floral resources if you want. But the model will respond in granularity and either synthesize it up to one generalist nesting and floral resource category or do this much more explicit temporal and nesting categorization. So this plugs in right here to the land cover biophysical table. I can see that there are nice green checkboxes on the right here. That is the model testing to make sure that your data tables and data inputs are formatted correctly. If they're the right file type this was a CSV so a comma separated Excel file. This is a tip for a raster file. This is a little info boxes that tell you what should go in these categories and it will give you a big old red X if you've done anything the model won't accept. The final table that we need to talk about before we can hit run is this guild table. So now that we have none of the model knows how the landscape intersects with resources for pollinators. We need to know how those pollinators use those resources. So the guild table is a little bit more simple than the biophysical table. And what it does is it takes a list of these species here we have bomb is called out, and then a couple of generalized species here either a generalist species that specifically rely on small stem nesting or cavity nesting resources. And then for each of these species you can link their dependence on these different kinds of resources using the columns here. The generalist species will happily use any of the nesting resources that we provided the model so we have ground nesting cavity nesting stem nesting etc. However the stem nesting species that we've called out in row four will only use the stem nesting resources here as indicated by this one column. This gives you an ability to map the different ways different species will use the landscape so you will see different results for these different species depending on how the landscape is providing these different kinds of nesting resources. The same can be said for the foraging activity here on the right. So this is linking the expected ways these different species forage on the landscape what they're looking for in terms of floral resources. These are different times of the year. So you can see that for the generalist species we've averaged it out it's an even foraging across all three seasons that we've prepped the model for. However bomb is here is only really foraging in the spring and then a little bit in the summer before not doing any foraging in the fall. So this is a way to look at how different species will rise and fall across a year in terms of their ability to forage. The last two parameters that you need to link each species to our alpha and relative abundance alpha is a very key parameter because this is telling the model how far each of these species is going to travel across the landscape to forage for floral resources. This will be in the units of the raster that you provide. So here it's in meters. And here you can see that bomb is is flying a significantly further distance than the small stem nesting the that we have right here that's only traveling about 300 meters. So that gets back to that animated diagram Eric showed earlier that some bees might not be able to forage in the wild lands further and further from their nesting site. So this helps you really target any restoration that needs to be done in smaller areas for smaller foraging rabies. The last parameter is the relative abundance of each of these species on the landscape. So in general this should sum to one across the column. So here we've just made it an even waiting for each of these species because this is an example. This can be taken from literature from your study area if you're really focused on bomb is saying you can up this to point eight and the others could be point one point zero five, etc. And all of these nice green checkboxes marked here you can click run. We're not going to run this for you it's it's rather boring to watch two researchers watch a model run on screen. So we've already run this to show what the results look like for this study area. So if we click into the workspace that we provided the model this results folder here. You can see a list of files going to take a look at them this way. A list of files most of them are rasters, along with a log telling you what the model did and then some intermediate outputs in a separate folder. Each of these can be visualized in a GIS. Here we're using ArcGIS but you can use Q or you can do it natively in a coding environment like Python or R. And we can take a look at there are two kinds of maps here there's pollinator abundance and pollinator supply pollinator supply let's take a look at that for bombs. You can throw it on the map and you see right here it's a gradient from zero to about point one to again as Eric said these are relative numbers so we're not expecting point one to be on the landscape this is just a low to high abundance ratio. And you can see pretty much immediately that the cropland in the Central Valley is much less pollinator heavy than the wildlands to West in white here. I just wanted to jump in and say that this setup for this county and all the bees took about a minute to run. This took about a minute to run my computer, it'll take longer for you I don't have the stats on how big this raster is, we're looking at 30 meter resolution so it's not extremely high input raster, but again, it depends on the specs of your machine and the capacity that you have to run the model. But it shouldn't be taking dozens of hours unless you're really running a very high resolution large land area. So, for pollinator supply, this is showing where the pollinators exist in the landscape where they're nesting where where they are starting from. So this is just showing not where you would expect to see them during the day, just where pollinators exist on the landscape in Z2. Pollinator abundance is often the map that we typically link to crop production to indices of value etc and that if we look at bomb this I've already loaded up onto the map is slightly different so if we toggle these on and off you can see the supply is a lot more crisp. And the abundance is a lot more faded in and that's because abundance is taking into account that traveling distance where are these pollinators coming from their nesting sites or the supply map on to the rest of the map in an expected way. Again, this is a zero to it's about again one point zero point zero four is the ratio here again it's just a relative index, not a number of views. In the West in the wild lands again you see an expected higher abundance, but there is some higher amounts of expected abundance on some of these farm fields in the Central Valley representing that foraging into farm lands from more protected nesting sites in the wild land. So that is the general look of the results of the invest model again you can see over here to the right is a lot of different rasters here, and that's because the math or the model will spit out the pair wise combination of the different kinds of nesting different kinds of species that you you entered into the model in the different temporal scales of foraging that you expect to see so we can see we have a bit generalist abundance in fall in spring in summer, we have bombast supply generalist supply etc so there's different ways of slicing the data that you have provided the model. So with that I'll return it to Eric to wrap us up. Great thanks Chris. And I think, just in wrapping up, I will say that if you're more information on the invest user guide and the website and literature you can find if you have any questions. Feel free to contact the kind of the help line. That's provided the website to. Alright, thank you. Thank you so much.