 Each presenter has 25 minutes with five minutes of questions. Our first presenter is Nate Roberts. Nate, if you can start loading up your presentation, that would be great. Okay, so Nate, I'll give you a private chat when you have three minutes and one minute remaining. And on that note, I will let you take over. Great, thanks, Justine. Hello, everybody. My capstone project, I am going to be looking at tracking pluvial related expansion and contraction of water bodies in the South Dakota Prairie Pothole region. So before I begin, just a brief outline on what I'm gonna present here this morning. First, I'll just present some basic background information. I'll cover a little bit of the study area and discuss the project goals and objectives, the preliminary methodology that we're using, some initial results that we have found and where we're gonna take this project, hopefully, to add its conclusion. So just to begin, my overall area of interest is the Prairie Pothole region. So what is that? It is about 900,000 square kilometers encompassing the map on the right of your screen. So covers North and South Dakota, Minnesota, some of Iowa, Montana, and three Canadian provinces. And what makes the landscape unique is it's dominated by these depressions and low spots that were formed during the last glaciation period. And 90% of the private land in this area is actually used for some form of agriculture. So be it row crops, corn, soybeans, that sort of thing, or for grazing pasture land, feed lots for cattle and so on. But in addition to that, it's also a pretty important habitat and flyway for birds. There's a, it's a pretty important flyway for Canadian geese and that sort of thing in the area. The landscape is also susceptible to exchanges in climate, especially extreme changes in precipitation. So just some background on the Pluvial years. And a Pluvial month is defined by Fiddle and Delworth as a time period where average precipitation exceeds 40% of the monthly average. So it's a really extreme climate event resulting in lots of day lose from rain and so on. And it's generally accepted that the Prairie Pothole region experienced this phenomenon in the early 1990s. And this has resulted in the formation and expansion of water bodies. So formation of new water bodies as well as expansion of existing ones. And what appears to have resulted is that farms, land and roads and towns have been flooded by this. So if you look at the photos on the right, I took those in the field a couple of weeks ago. So the top one used to be someone's farm place, which has now been taken over by this water body. And then in the lower right corner, this appears to have been a feedlot for raising cattle, which is now also flooded. And what's interesting about this area is past studies have examined the Prairie Pothole region in response to Pluvial precipitation, but on a much smaller scale. So past work has involved single lakes, Devils Lake in North Dakota, or a complex of lakes, the Wabbe complex in the eastern South Dakota. So at a much smaller scale that I'm hoping to do eventually. So current climate conditions within the Prairie Pothole region, it's generally accepted that we're now in a dry climate condition. In fact, current year to date precipitation amounts indicates the Prairie Potholes actually in drought conditions. So if you, the map to the right of the text there is from the National Drought Survey Monitor just a few weeks ago, and South Dakota especially is experiencing anywhere from moderate to severe drought. And field inspection suggests that these water bodies are actually retreating. So the photo on the far right I took again a few weeks ago and you can see that there was once a higher water mark at this water body than is currently being observed. So the study area that I'm going to hope to analyze is the South Dakota Prairie Pothole region. And that is dominated by row and crop and pasture land. The map on the right I made from the National Land Cover Database. So there is a fair amount of crop land and also pasture land. And this area interestingly enough composes about almost all the suitable ground for row crop production in the state. So to the west of this area beyond the Missouri River, which divides South Dakota in half, you have a noticeable lessing of average precipitation. So you'll have a lot of grasslands not suitable for row crops. So Ag has a $25 billion annual impact on the state economy. So any disruption to that could have a profound effect. And also they would be interested in new means of management in addition to that. So my overall research question is, are the areas that have been reclaimed or those areas that have seen water bodies disappear being farmed? And if so, what is being grown there? And to help, I guess, answer that question, I've come up with several project goals and objectives. The first of which is to quantify either farmland gained or lost to these water bodies within South Dakota prairie pothole region from 1986 to 2016. And like I mentioned before, it's generally accepted that climate conditions have resulted in these areas being flooded as of 1990s. So starting with 1986 should give us a good baseline to see what existed before that time and if that phenomenon is returning as we move more toward the present. Second, to quantify the spatial and temporal relationships between the expansion and contraction of poluvial water bodies and precipitation received from 1986 to 2016. So I want to see if I can statistically, if there's a significant relationship between a period of time where precipitation was received in the region and how much land area these water bodies have amassed during this time period. And third, again, to determine what crops are currently being produced in the areas that have seen these water bodies disappear or retreat. And the fourth goal is to develop an automated framework for future studies involving large data sets within this region. I'm going to need 30 years worth of imagery. So obviously it's quite a large data chunk. So I want to try to automate as much of the analysis as I can. So that is pretty important goal in my mind. So the data required for my project is primarily going to be based on Landsat imagery. And that'll be for the time period of interest, 1986 to 2016. And I'm going to use that to construct an annual, once a year, Landsat time series for the area of interest. And all the imagery will be the level one product, which the USGS and their publications has indicated that is suitable for time series analysis. And the scenes will have to be captured in June, July and August to correspond with peak growing season. And that being able to differentiate vegetation, healthy vegetation from water should help pick out those water bodies as they'll appear darker in the near infrared composite like the image on the right hand of the screen. And for processing sake, I required less than 10% cloud cover. Obviously, there'll be some cloud cover, but just for processing sake, the least amount of cloud cover, the better. So for my 30 years with the data, it encompasses three spacecraft systems, Landsat 5, 7, and 8. All of these have the 30 meter grid size. So additional data sets. I will be using the National Ag Statistics Service Crop Land Data Layer that describes crop cover in the area of interest in a 30 meter resolution. And that's available specifically for South Dakota from 2006 to 16 in a grid format. And the National Land Cover Database also will be used. That's a 30 meter grid also. And the most current on that one is 2011. Prior to that, it was 2006, but I decided to go with 2011 since it was more current. And then the climate precipitation data. I'm using a prism climate data source from that's built out of the University of Oregon. And they produce that in 800 meter grid. And they have many more years than I'm going to need. So 1895 to 2016. So my 30 years during my time period of interest should be easily attainable from that. So my methods on pre-processing this data. First, acquire the Landsat Scenes from the Earth Explorer, the USGS Spatial Data Portal. And then I need two different to create composite band rasters since Landsat 5 and 7 are thematic mapper or enhanced thematic mapper that by band combinations are the same. Whereas the Landsat 8 uses the OLI sensor. So there's a little bit different. So that's the divergence there is I need to just different band combinations required for creating the near infrared composites. And since the imagery doesn't cover the entire area, multiple scenes will be required. And I'll stitch those together using a mosaic method, clipping them to the South Dakota for a pothole region, the area of interest. Finally, the final output being a clipped and composite raster for the area of interest. Similarly, acquiring the National Crop Land Data Layer from the US or the AG Statistics Service, clipping that to the area of interest. And there's many different, I guess, classifiers within that. I'm just going to reclassify that to suit the needs of my project for open water crops. Wetlands developed land, grassland and pasture, and then clip that to the final output would be then the clipped layer for that. Then acquire the Crop Land Data Layer, clip to the area of interest, and then the final output being the clipped final product of that. So a big part of my project, the pre-processing will be done in ArcGIS, and I'm going to extract the water bodies from the Landsat imagery using an object-based image analysis approach, specifically using the eCognition software package. And I guess I just wanted to mention how that differs from any classifying that you may have done based on pixels is eCognition segments the image into these image objects, which are actually based on neighboring pixels with similar spectral values. So you're working with a higher unit of analysis, which I found to be pretty helpful in previous work. And while this method approach has been used extensively to detect change in forests and crop land, we have not seen anything in the literature that suggests it's been used to classify water bodies on a large scale. So just some background on how this works. You start with a source image that you then get into the segment. You segment the image or split the image up into those image objects, and then you can classify these image objects into groups. So the image on the right would be the final product of all these objects classified as water, where the green lines are water from the scene. So to calculate the Landsat to classify the water, I'm using a method called the Normalized Difference Water Index first proposed by McFeeders in 1996. And that calculates as the green bands less than near infrared divided by the green band plus near infrared. So what that does for you is any image objects with the NDWI of less than zero are bright objects, which would then be not applicable to being so sit with the water classification. Alternatively, objects with the NDWI of greater than zero would be dark objects, which would be water. So the image on the right is the final water body digitized with the objects merged into one seamless area. So just getting into the methods of how I'm going to classify the imagery. Again, just preprocessing done in ArcGIS, followed by the classification steps within eCognition, segmenting or creating those image objects, calculating the NDWI values, assigning objects with NDWI of greater than zero as water. And then in previous work, it's generally accepted that even though some neighboring objects may not have NDWI value appropriate to be classified as water alone, if they share a reasonable border, you can put them into that class with pretty good understanding of that they are actually water. And then you have to merge these objects into one seamless image. And then I'm going to try to separate out seasonal water bodies from permanent so you can compare the NDWI values from one year to an X in eCognition with the final product being a polygon-shaped file that you can then import into ArcGIS. So after importing into ArcGIS just in visually inspecting and doing any correction to the classification, I'll perform an accuracy assessment. Then I decided to incorporate just a field ground truth since I live fairly close to this area just seeing how my classification shakes out in person and that would result in the final water body classification. So bringing this all data together at a high level there's a lot more steps that are going to be depicted here but this is just a high level so you start with your water body polygons and then overlay that to the crop land data layer which would then tell you what crops are present in areas that had previously been flooded. And then to calculate any land area changes the national land cover database overlaid over the polygons to then determine what has existed what is now water or alternatively what is now not water. And then to correlate water body size with precip data I'll take the water body polygon surface area compare that to the prism climate data and run a regression analysis probably in the statistical package are to see if there's any significant relationship there. Some initial results just on the Landsat classification these three images are just simply digitized a digitized slew which is a shallow body lake over three years located near Langford South Dakota from 1986-95 in 2016 so just if you just visualize at the imagery here it looks that you can make a reasonable assertion that there is some expansion and then reduction from those three years and I should mention that all of these images are presented at the same scale so some initial results accuracy assessment of my three years I'm just going to highlight on three different areas of this table first the user's accuracy or the error of commission which means that these water pixels were included there are pixels included in the water class that evidently were not water so 2016 fairly low 33% with that the producer's accuracy or the error of omission or the water body pixels that should have been included in the class that weren't 50% for 95 and while these are sort of alarming I just kind of went back briefly and revisited my methods for creating this and I think I may need to revise that a little bit I don't think it is going to be as I guess unreasonable as these numbers suggest and the total accuracy seems fairly reasonable for this area so some initial results just on how these water bodies appear to be working out my test area near Langford South Dakota I calculated the surface area of this slu in 1986 is 318 hectares 1315 hectares in 1995 and 518 hectares in 2015 so you can see that with our 30 years here there would be a strong correlation to get some sort of baseline to how what was happening before these pluvial years or these years of intense precipitation in the area of the early 1990s so we have a reasonably sized water body a huge water body and then one that has shrunk considerably almost back to where it was in 1986 so the table below that graph is from the Prism Climate Data Series just a time series of the precipitation and millimeters for this area so just visually looking at it it seems pretty reasonable that our precipitation will mirror eventually what we are seeing in terms of these surface areas so some initial results on crop distribution these crops are now being grown in areas that were previously flooded in 1995 so I started with what areas were flooded in 1995 and what areas are now have crops in them in 2015 and this is a 10 kilometer study area near again Langford South Dakota I picked this area because I had spoken to somebody who lived in this area and experienced this phenomenon in the 1990s this is actually based around a farm place that he owns so in 1995 of 3,568 hectares of 41,756 hectares of my study area were flooded and then in 1998 by night by 2015 998 of these areas have been reclaimed or the water has disappeared so the polygons that are digitized here are the water bodies that existed in 1995 and in the center image this cyan colored areas are areas that no longer have any water present so this circle in the center of their highlights are just a small area of that so when you look at how what that does in terms of crops being grown in 2015 farm producers were able to add 24 hectares of soybeans that were to areas previously flooded in 1995 23 acres of hectares of hay and 20 hectares of corn so to quantify that in terms of money 24 hectares of soybeans is about 150 acres in standard units and based on yield average yield for this region last year at today's prices that actually would give the producer an additional $32,000 gross so when I calculated that I was like wow this is really cool this is amazing this is only 10 kilometers imagine what's going to be over the whole study area so I'm pretty excited to see how this shakes out and just based on knowledge I have of producers in this area they would be that took out two acre buffer of trees just to farm that so imagine the opportunity there if more of these areas can be farmed it's pretty significant future direction just further refined methods mask cloud pixels further automate processing correlate the water body advance and retreat with climate data and further predict crop distributions and with that I'd be happy to take any questions I'm just Dean asks out of curiosity why is the area called the pothole region and that correlates to that notion of when the glaciers were present here during the last glaciation period they've dug up the landscape and this has resulted in an area of kind of low spots and depressions throughout it if you move to the east of this area a little ways there's the Prairie Cato region that's where the deposits ended up it's really the highest point in eastern South Dakota so I guess that's why it's called the pothole region it's just dominated by all these little depressions and basins Jennifer asks is there costs to reclaim the land after it's underwater that is a interesting that is an interesting question we sort of went with the idea that the producer would know based on their experience when they should and should not replant something that had previously been flooded you know it doesn't have a propensity to flood does it have has it had a sufficient amount of time for the soil to dry out and etc we're just kind of looking leaving it up to nature I suppose just to dictate that and Michaela asks how I was going to use the national land cover data set versus my own results and yeah what I'm going to do with that is I'm going to use my results from the water bodies that were extracted from e-cognition from the Landsat imagery and the national land cover data layer will just be used simply to overlay against my polygons from my water bodies to determine what areas are now being I guess farmed so does the has the land class changed to something other than water in that area any final questions nope