 First speaker this morning is Allison Long. Allison Long is a science fellow with Minnesota North Dakota South Dakota units of the Nature Conservancy. Her work involves many aspects of conservation science including communicating science, conservation planning, managing data and traditional field science and grassland monitoring projects. She's also involved in work around sustainable grazing, climate adaptation and metrics around benefits to people from conservation interventions. Allison earned her master's of science degree in environment and conservation science from the North Dakota State University. Studying community composition of my micro rise of fungi across the Northern Great Plains. She also has a bachelor's of science degree in psychology. She currently resides in Fargo, North Dakota where she is perfectly located to travel, perfectly located in Fargo. To travel to all beautiful nature areas of the Dakotas in Minnesota to offer. Her title today is adaptive management on tall grass prairie for native species composition. Let's welcome Allison. Well, as Sarah mentioned yesterday, I'll be talking to you about an adaptive management project similar to MPAM called the Grassland Moderate Team or GMT which has a focus on improving native species cover and composition in our tall grass prairies. Like Sarah said about MPAM, we certainly don't have all the answers but I'm excited to share with you a little bit about the project and the work that we've done so far. So to give you a quick summary of what I'll be talking about today, I'll just give an overview of the project. I'll give some results of previous retrospective analyses. I'll talk a little bit about trends in abundance of smooth brome and Kentucky bluegrass that we've seen and then share some conclusions. So first I wanna highlight and acknowledge that this project is a collaborative effort mainly between the nature conservancy and US Fish and Wildlife Service and Minnesota Department of Natural Resources but also in partnership with the US Geological Survey, University of Minnesota, University of Colorado at Boulder and South Dakota Game Fish and Parks. So the Grassland Moderate Team or GMT began in 2007 among a multi-agency group of grassland managers and scientists. The group felt that a cooperative strategy and like a standardized monitoring effort would improve our effectiveness at resolving uncertainties about grassland management and would also facilitate comparisons of data across ownership and throughout the tallgrass prairie region. So as we've talked a lot about in this workshop, remnant and restored prairies in the Northern Great Plains are threatened by encroaching invasive species, especially cool season introduced grasses and woody vegetation. And as you know, the main focus of grass and monitoring efforts, or sorry, management efforts is to protect and enhance the competitive ability of a native plant species. And for this project in particular, more focused on the native tallgrass prairie communities in the Western Minnesota and the Eastern Dakotas. And the purpose of the GMT project is to evaluate which management actions have the greatest likelihood of improving cover and composition of native plant communities and also recommending management actions. So the aim of the adaptive management approach is to evaluate the best management actions and their frequencies to maintain and enhance native tallgrass prairie plant communities in the region using real-time management information. So similar to NPAM's mixed grass prairie management treatments, the management actions that we use in the GMT project are burning, grazing, burning and grazing in the same management year and rest. So for this project, a management unit is considered burned if greater than or equal to 50% of the vegetation has been exposed to fire. A management unit is considered grazed if greater than or equal to 25% of the area or the biomass has been impacted by livestock in some way. And there's no limits on the number or type of animal or length of grazing. And then we also wanted to look at burning and grazing within the same management year, which is defined as October 1st to September 30th, separately from those actions separately. And the last option is rest or no management action. And because managers still face a lot of uncertainty over which actions to take and how often to take them, we've employed this adaptive management model to help us learn. So adaptive management is a structured iterative process of decision-making in the face of uncertainty with a goal of reducing uncertainty over time via monitoring. And the grass and monitoring team uses double loop learning with iterative and deliberative phases. So the first deliberative phase identified stakeholders, project objectives and other decision components and that occurred in 2007 and 2008. And then from 2008 to 2020, the team has been in an iterative phase of the cycle. So implementing management, monitoring, using data from monitoring to evaluate management and update the model and adjusting management recommendations. And then in 2020, the team started the double loop learning phase. So re-entering the deliberative phase of the cycle. So focusing on reassessing and updating the model. And then the iterative phase resumed again in 2020, following that double loop learning and updates to the adaptive management model. So I'm not gonna go into much depth on this, but I want to highlight that the model, the outputs are management recommendations is run every year and it uses a state transition model which has been mentioned several times in this workshop. In GMT, the states are defined by native cover and plant community. And each state receives its own management recommendations from the model each year. And then really briefly in case anybody is interested, the process use a random forest model to calculate transition matrices. And then data are used to inform stochastic dynamic programming. And then finally, the model generates management recommendations dependent on the current conditions of the site, such as level of invasion and predominance of herbaceous versus shrub cover. And then those management recommendations are provided to land managers. So this is a map of our sites with the orange polygon showing the distributions of sites across the western, across western Minnesota and the eastern Dakotas. And then that gray outline is the northern tallgrass prairie ecoregion boundary. The sites span to ecological provinces, the tallgrass aspen parklands in the north and the prairie parklands in the south and the west. And the sites in the model are all unplowed remnant prairies. And prairie community types include wet prairie, music prairie and dry prairie. The dominant vegetation varies by community type and includes big bluestem, yellow and gene grass, prairie cord grass and prairie drop seed in wet to music prairie and little bluestem and porcupine grass in dry prairie. The most common invasive species are the non-native cool-season grasses, Canada thistle and sweet clovers. And I should mention that the sites in the model would only unplowed prairie, but their condition really ranges from having a high diversity of native plant species and low abundance of invasive all the way to the other end of the spectrum. Okay, so now I'm gonna share with you an overview of the monitoring that we do at these sites. So the protocols we use were designed to measure baseline plant community composition as well as changing composition and structure in response to the selected management actions. They're also developed to be simple, fast, robust and effective for large-scale monitoring. So the sample units are permanent transects distributed randomly within a management unit at a density of one transect per 10 acres with a minimum of five and a maximum of 15. And they're generally revisited about every three years. So most transects are not monitored every year. And the transects are 25 meters long, a 10th of a meter wide, and they're subdivided into 50 half-meter plots. So you can see in this picture that red outline kind of gives you an idea of what those plots look like. So every time I say plots, think of that. And we also collect data on vegetation structure by using a robel pole placed at the center point of the transect. And we use the robel pole as is shown in that photo to get a visual obstruction reading or VOR. And we collect as well a litter depth measurement every five meters along the transect. And then the other important part of our monitoring is plant community composition data, which is collected two ways. We record any native or invasive indicator species observed within a three-meter wide belt transect, so a meter and a half on either side of the meter tape all the way along the transect. And then also within each half-meter plot along the transect, we record any invasive species that are there and whether they are present or dominant. Then the other aspect of this is for each half-meter plot along the transect, we assign and record a plant species code, or I'm sorry, a plant group code, which reflects three different dimensions, native to invasive, herbaceous to low shrub to tall shrub and grass to grass form to forb. So you can see here the decision tree that we use in the field. So for example, if there's more than 50% native cover in a plot, that's the first step here. And then the next step would be to decide if it's all native or mostly native and these classes are determined by the percent of cover, then you would decide if it's herbaceous or low shrub or tall shrub. And then the last step would be if it's grassy, grass form mixed or just forb. And at the end of that process, you get a three-digit plant group code that reflects all those different decisions. So after we have all that monitoring data and all of our management data, we put all that into the model and to just give you an example of what the model output might look like, we put in all the data into the model and then the model will assign each transect to one of six states and then gives a recommendation. So for example, a transect with high native cover and herbaceous community would be assigned to state four, high herb. And then for that state, the model would say in the first year burn, in the second year graze and then third year rest. So now that you have kind of a general understanding of the GMT project and the model that makes management recommendations, I wanted to share some of the retrospective analyses that have come from the work that this team has done. So one of them is this paper by my supervisor, Marissa Allering as well as Sarah and others. And so this analysis uses data from 2008 to 2016 and overall, this analysis showed that native plant cover increased at low quality sites. So that was one of the big takeaways of this. And just diving a little bit more into those results. So the analysis modeled relationships between management actions and native prairie species cover and frequency. And so another important result from this work was that across all the different model configurations that were tested in the analysis, the team found that the strongest predictor of whether the percent of native cover management goal was met was the initial state of the transect. So management goals were different depending on the initial condition of the transect. So for example, for units that began in a high quality state, the goal would just be to maintain that state. So you want no change. And then on the other hand, sites that were initially in a low quality state would have a goal of improving to a higher quality state. So this figure shows the relationship between whether the percent of native cover goal was met. So zero would be that met and one would be met on the y-axis. And then the initial state of the transect is on x-axis. And the bars along the top and bottom show the density of data points that meet the management goals. And there's no trend line here, but you can see that the higher the initial state, the higher the percent of native cover goal met. And again, the point here is that the strongest predictor of whether a transect met its native cover goal was its initial state. So transects that had high native cover to start with are the ones that are most likely to have a high native cover later. And then this figure is the same as the previous slide, except that it shows the proportion of years burned on the x-axis. And again, there's no trend line on this figure, but you can see that the higher proportion of years burned is associated with higher percent of native cover goals met. And one of the big findings from this analysis was that burning was the most effective treatment for enhancing the native plant communities. And then when plotting this relationship for the proportion of years grazed, you can see that a greater proportion of years grazed is related to lower percent of native cover goals met. So actually this analysis found that grazing negatively affected the native plant community, especially at low quality sites. However, I should point out that the effect is weak and the data set is actually a bit biased because not many high quality sites get grazed in this project. So pivoting from that analysis on native cover, I just wanted to share some findings related to cool season invasive grasses. And as we've discussed a lot in this workshop, Kentucky bluegrass and smooth brome are some of the biggest threats in this region. And so we have some data investigating their effects. And I wanted to mention that in addition to Kentucky bluegrass and smooth brome, our data set also includes observations of Poha compressa or Canada bluegrass. So both Poha species are lumped together in our monitoring protocol. And so therefore they're also lumped together in this analysis. So I was curious to see if the GMT data set would show any trends in cover for these species. Oh, sorry. So this analysis included data from 2008 to 2022. And not included in the analysis were transects outside of the northern tallgrass peri-eco region as well as transects that were only monitored once were excluded. And transects that were monitored only twice were excluded if those two monitoring events happened for your fewer years apart. So for each transect, we calculated the proportion of half meter plots where smooth brome is dominant, which is greater than 50% cover. Where smooth brome is present, which is one to 50% cover. And then where Kentucky and Canada bluegrass are dominant, which is 52, or sorry, greater than 50% cover. And the proportion of half meter plots where Kentucky and Canada bluegrass are present, one to 50% cover. And then a one sample Wilcoxon test was used to evaluate if proportion of plots with brome and bluegrass changed over time. So these figures just show the proportions of plots per transect where each species was present or dominant over time with black diamonds representing the means. And one thing to keep in mind is that we can't make conclusions about trends over time from these figures because different transects are monitored each year. But this does give you kind of a sense of the scope of the issue. And you can see that on average, bluegrass is present in a large proportion of plots per transect. So each transect in this analysis has been monitored multiple times over the years. And from each survey, we have the proportion of plots where brome and bluegrass were observed either present or dominant. So in this figure for each transect, I subtracted the first observation from the most recent observation to obtain the overall change in proportion of plots within each transect where each species was present or dominant. So you can see there's a lot of variation there. And you can't tell from this figure, but from some other analyses, I know that brome and bluegrass are increasing in some transects, they're decreasing in some transects, and there's a lot of transects where there's no change. And the black circles on this figure show the means. And if we take a closer look at those in this figure, there's a horizontal line at zero change in proportion over time. And you can see that the three categories above the line are all significantly higher than zero. So brome has become both dominant and present in a greater proportion of plots over time, and bluegrass has the largest increase in presence over time. But across all plots mean proportion of plots within a transect where bluegrass is dominant has not changed over time. So building off of that, I wanted to share one other retrospective analysis using the GMT dataset, looking at invasive species cover and management. So this work was done by Hugh Ratcliffe, Marissa Allring, Sarah and others. And the analysis includes data from 2010 to 2019 at Minnesota sites only. It estimated the effects of burning, start time of the growing season and their interaction on invasive species relative cover and frequency. So overall they found that burning reduced the abundance of cool season grasses, leading to a reduced abundance of invasive species as a whole. The reduction persisted over time for invasive cover, but quickly diminished for the frequency of occurrence. So these plants quickly started to emerge again, even if they didn't have as much cover. So as part of this analysis, it was hypothesized that with earlier growing season starts or GSS invasive species like Kentucky bluegrass and smooth brome shown in red in this figure would be better than natives shown in blue at shifting their phenology and using that new empty niche space, allowing them to increase in abundance either through growing larger or increasing their number of individuals. And the ability for these highly invasive species to grow earlier could then result in seasonal priority effects where their greater abundance earlier in the season reduces access to light and space for later growing species, which tend to be the native species. However, the results did not support this hypothesis at all. Invasive species abundance was not positively affected by earlier growing season starts. Instead, the team actually found evidence that later growing season starts increase the abundance of some invasive species. However, the effects of burning on plant community were mostly unchanged by the timing of the growing season. Although earlier growing season starts did weaken the effectiveness of burning on Kentucky bluegrass and smooth brome. So this figure displays some of those results. It shows the estimated effects on species frequency and cover of burning at the top, growing season start day in the middle and then the interaction of the two at the bottom. The error bars are showing 95% confidence intervals and orange indicates a negative effect, green indicates positive effect and black indicates no effect. You can see on the left-hand side, T minus one means management actions happening in the prior year and T minus two means two years prior. So the two rightmost columns that are highlighted here are Kentucky bluegrass and smooth brome. You can see the negative effects of burning and growing season start date separately, but the positive effects on those two are combined in the bottom. So to summarize all of this, the GMT data show that smooth brome is becoming present or dominant in a greater proportion of plots per transect over time and Kentucky and Canada bluegrass are also becoming present in a greater proportion of plots over time, but there's no change in dominance there. We also saw that burning is most effectively enhanced, sorry, that burning most effectively enhanced the native plant community and increased the dominance of native indicator species. And the earlier growing season starts weakened the effectiveness of burning on Kentucky bluegrass and smooth brome. And then a few thoughts about the project and the process in general. So ecological change is slow. As we know, these ecosystems developed over thousands of years and they're composed of long-lived perennial plants. So it's going to take a lot longer than 15 years for us to dramatically change the system. Also learning from adaptive management is a slow process, although the double loop learning process allows us to reassess and make adjustments to the model and framework as we learn. And then last, I just want to reiterate the benefits of collaboration. So documenting both management actions and monitoring information in a standardized way allows us to learn faster from each other than we can working alone. So that's all I have and happy to take any questions.