Yellow sweetclover (Melilotus officinalis; YSC), an invasive biennial legume, bloomed throughout the Northern Great Plains (NGP) following greater-than-average precipitation during 2018-2019. YSC can increase nitrogen (N) levels and potentially cause broad changes in the composition of native plant species communities. There is little knowledge of the drivers behind its spatiotemporal variability, including conditions causing significant widespread blooms across western South Dakota (SD). We aimed to develop a generalized prediction model to map the relative abundance of YSC in suitable habitats across rangelands of western SD for the recent sweet clover year 2019. The following research questions were asked: 1. What is the spatial extent of YSC across western SD? 2. Which model can accurately predict the habitat and percent cover of YSC? and 3. What environmental drivers affect its presence across western SD? We trained machine learning models with in-situ data (2016-2021), Sentinel 2A-derived surface reflectance and indices (10m and 20m) and site-specific variables (e.g., climate, topography, land cover, and edaphic factors) to optimize model estimates. Our study identified the most suitable drivers to explain the variability in YSC presence and its percent cover through data dimensionality reduction techniques. Our research demonstrated how machine learning algorithms could help generate valuable information on the spatial distribution of invasive rangeland plant species. We found major YSC hotspots in Butte and Meade counties of SD. The floodplains of major rivers in SD, such as the Cheyenne, White, and Bad Rivers, also showed a higher occurrence probability and percent cover range. These prediction maps could aid land managers in devising strategies for regions that are prone to YSC overruns. This management workflow can serve as a prototype for mapping other invasive plant species worldwide.

Yellow sweet clover (Melilotus officinalis; clover hereafter) is a biennial legume native to Eurasia that is now present in all 50 states. Clover can grow 2 m tall and achieve high densities across large areas in the Northern Great Plains when conditions are conducive, such as in 2019. Clover is highly efficient at fixing nitrogen in soils which reduces the abundance of native grasses, while simultaneously facilitating invasion of non-native grasses, which may alter fire regimes. In contrast, clover provides considerable forage for ungulates, attracts a wide variety of insects that, along with clover seeds, are important to waterfowl, gamebirds, and songbirds, and supports numerous pollinators. Little is known about the extent of clover in central Montana and northwest South Dakota and this study represents the first known attempt to map clover in these regions. In 2019, the Bureau of Land Management conducted Assessment, Inventory, and Monitoring (AIM) surveys at 10 sites in central Montana (defined as the approximate geographic extent of Musselshell County) and 24 sites in northwest South Dakota (defined as the approximate geographic extent of Butte and Harding Counties). Concurrent Unmanned Aerial Vehicle (UAV) flights were conducted at 22 sites: 6 in Montana and 16 in South Dakota. We created orthoimages from the 22 UAV surveys as well as clover maps for the 19 sites with clover. Percent clover cover from the UAV-derived clover maps closely matched percent cover from AIM data along surveyed transects. The UAV clover map with the greatest percent cover in each region was then used to identify pixels comprised of clover in National Agricultural Imagery Program (NAIP) imagery: 5,000 pixels in Montana and 2,500 pixels in South Dakota. We used separate MaxEnt models to classify clover across 1 NAIP tile in central Montana and 2 NAIP tiles in northwest South Dakota. Next, for each region, we calculated the percent of classified NAIP pixels within each Sentinel-2 pixel and selected 1,000 pixels from each of 2 fractional cover (FC) bins representing 20% increments from 10-50% cover and 1,000 pixels from each of 5 fractional cover (FC) bins representing 10% increments from 55-95% cover. We also selected 1,000 pixels in each region from dense clover strands visible in Sentinel-2 imagery representing pure (i.e., 100% cover) clover areas. Separate MaxEnt models were run in each region for the pure clover areas and each FC class. We fixed the pure clover area for each region and added fractional coverage components outside this consistent pure clover area by thresholding each of the 5 FC models using 5 common MaxEnt thresholds and merging results using 3 classification approaches for pixels classified by multiple FC models: minimum, mean, and maximum cover predicted. Accuracy of the 15 FC maps were validated by comparison to AIM survey data (30 m buffer from AIM plot center) and UAV-derived clover maps (300 x 300 m grid of 900 Sentinel-2 pixels centered on AIM plot centers). Datasets in this release include the following items in associated zipped folders: 22 UAV orthoimages of which 19 have embedded clover maps aligned to Sentinel-2 imagery (MT1-6_Sentinel_proj and SD1-16_Sentinel_proj). 2 Classified NAIP images aligned to Sentinel-2 imagery. (MT/SD_NAIP_Sweet_Clover_Sentinel_proj) 15 Fractional cover maps for both central Montana and northwest South Dakota (MT/SD_Sweet_Clover_Fractional_Cover_Maps). 2 Point shapefiles of AIM plot centers and 2 polygon shapefiles for Sentinel-2 to UAV comparison extents (MT/SD_Sweet_Clover_Shapefiles). Seven .csv files (Sweet_Clover_csv) that contain 1) Green Leaf Index reclassification values for UAV clover classifications (S1_GLI_Reclass.csv); 2) Clover cover from AIM surveys and UAV-derived clover maps along AIM transects and sites (S2_UAV_AIM_Comparisons.csv); 3) Center locations for all pixels used to classify clover with MaxEnt models (S3_Training.csv); 4) MaxEnt variable permutation importance

Yellow sweet clover (Melilotus officinalis; clover hereafter) is a biennial legume native to Eurasia that is now present in all 50 states. Clover can grow 2 m tall and achieve high densities across large areas in the Northern Great Plains when conditions are conducive, such as in 2019. Clover is highly efficient at fixing nitrogen in soils which reduces the abundance of native grasses, while simultaneously facilitating invasion of non-native grasses, which may alter fire regimes. In contrast, clover provides considerable forage for ungulates, attracts a wide variety of insects that, along with clover seeds, are important to waterfowl, gamebirds, and songbirds, and supports numerous pollinators. Little is known about the extent of clover in central Montana and northwest South Dakota and this study represents the first known attempt to map clover in these regions. In 2019, the Bureau of Land Management conducted Assessment, Inventory, and Monitoring (AIM) surveys at 10 sites in central Montana (defined as the approximate geographic extent of Musselshell County) and 24 sites in northwest South Dakota (defined as the approximate geographic extent of Butte and Harding Counties). Concurrent Unmanned Aerial Vehicle (UAV) flights were conducted at 22 sites: 6 in Montana and 16 in South Dakota. We created orthoimages from the 22 UAV surveys as well as clover maps for the 19 sites with clover. Percent clover cover from the UAV-derived clover maps closely matched percent cover from AIM data along surveyed transects. The UAV clover map with the greatest percent cover in each region was then used to identify pixels comprised of clover in National Agricultural Imagery Program (NAIP) imagery: 5,000 pixels in Montana and 2,500 pixels in South Dakota. We used separate MaxEnt models to classify clover across 1 NAIP tile in central Montana and 2 NAIP tiles in northwest South Dakota. Next, for each region, we calculated the percent of classified NAIP pixels within each Sentinel-2 pixel and selected 1,000 pixels from each of 2 fractional cover (FC) bins representing 20% increments from 10-50% cover and 1,000 pixels from each of 5 fractional cover (FC) bins representing 10% increments from 55-95% cover. We also selected 1,000 pixels in each region from dense clover strands visible in Sentinel-2 imagery representing pure (i.e., 100% cover) clover areas. Separate MaxEnt models were run in each region for the pure clover areas and each FC class. We fixed the pure clover area for each region and added fractional coverage components outside this consistent pure clover area by thresholding each of the 5 FC models using 5 common MaxEnt thresholds and merging results using 3 classification approaches for pixels classified by multiple FC models: minimum, mean, and maximum cover predicted. Accuracy of the 15 FC maps were validated by comparison to AIM survey data (30 m buffer from AIM plot center) and UAV-derived clover maps (300 x 300 m grid of 900 Sentinel-2 pixels centered on AIM plot centers). Datasets in this release include the following items in associated zipped folders: 22 UAV orthoimages of which 19 have embedded clover maps aligned to Sentinel-2 imagery (MT1-6_Sentinel_proj and SD1-16_Sentinel_proj). 2 Classified NAIP images aligned to Sentinel-2 imagery. (MT/SD_NAIP_Sweet_Clover_Sentinel_proj) 15 Fractional cover maps for both central Montana and northwest South Dakota (MT/SD_Sweet_Clover_Fractional_Cover_Maps). 2 Point shapefiles of AIM plot centers and 2 polygon shapefiles for Sentinel-2 to UAV comparison extents (MT/SD_Sweet_Clover_Shapefiles). Seven .csv files (Sweet_Clover_csv) that contain 1) Green Leaf Index reclassification values for UAV clover classifications (S1_GLI_Reclass.csv); 2) Clover cover from AIM surveys and UAV-derived clover maps along AIM transects and sites (S2_UAV_AIM_Comparisons.csv); 3) Center locations for all pixels used to classify clover with MaxEnt models (S3_Training.csv); 4) MaxEnt variable permutation importance

,Phenotypic evaluation of 37 crimson clover (Trifolium incarnatum L.) accessions from the US National Plant Germplasm System. Focus of the trial was on traits important for cover crop performance, including fall emergence, winter survival, flowering time, biomass, nitrogen (N) content in aboveground biomass, and proportion of plant N from biological nitrogen fixation (BNF). Experiments were conducted at the Beltsville Agricultural Research Center (Maryland, USA) across three growing seasons (2012-2013, 2013-2014, 2014-2015).,The field design was a randomized complete block design (RCBD) with four replications in each year, except for five accessions planted in 2015, which only had three replications due to limited seed availability. Each plot was a single row 0.6 m in length and 1.5 m between plots. Between 37 and 45 seeds were planted per plot, depending on seed availability in each year.,Fall emergence was evaluated in late October of each year by counting the total number of plants in each plot. Winter survival was determined by counting total number of plants per plot in late April divided by the total number of plants counted in the fall.,Flowering time was evaluated by recording percent flowering on a per-plot basis on a scale from 0% (no flower buds present) to 100% (all flowers dried up entire length of head). Flowering evaluations took place periodically between late April and early June. In 2013, evaluation took place on six dates: 23 Apr., 9 May, 15 May, 24 May, 30 May, and 4 June. In 2014, evaluation took place on five dates: 28 Apr., 6 May, 13 May, 19 May, and 27 May. In 2015, evaluation took place on eight dates: 25 Apr., 29 Apr., 4 May, 7 May, 11 May, 14 May, 18 May, and 21 May. Frequency of evaluations and total duration of evaluation period varied from year-to-year primarily due to the effects of year-to-year weather variation on the rate of growth and development.,Once an accession was rated at 50% or greater for flowering, biomass was collected. All plants in the plot were pulled up with roots attached. Plants were counted and the roots were clipped. All plants within a plot were placed in the same brown paper bag and dried. Dry weight was recorded and plants were ground for laboratory evaluation of nitrogen content, proportion of nitrogen from BNF, and metagenomic analysis.,The crimson clover biomass samples were separated into shoots and roots. Shoots were oven dried (60 °C) for approximately 72 h, weighed, and ground to pass a 1.0-mm screen. Tissue C and N concentrations and 15N natural abundance were determined for the shoot material of each accession using a Thermo Delta V Isotope Ratio Mass Spectrometer (Thermo Scientific, Waltham, MA) and Carlo Erba NC2500 Elemental Analyzer (Carlo Erba, Milan, Italy). Isotopic abundance data were expressed as δ15N in parts per thousand (‰), representing the abundance of plant tissue 15N relative to that of atmospheric N2.,,

Geographic distribution data were collected based on county level occurrences (or converted from point occurrences to county level occurrences) within the five focal states (Minnesota, North Dakota, South Dakota, Nebraska & Iowa) and each U.S. state or Canadian province bordering those focal states (Wisconsin, Illinois, Missouri, Kansas, Wyoming, & Montana in the USA and Saskatchewan, Ontario & Manitoba in Canada).

Geographic distribution data were collected based on county level occurrences (or converted from point occurrences to county level occurrences) within the five focal states (Minnesota, North Dakota, South Dakota, Nebraska & Iowa) and each U.S. state or Canadian province bordering those focal states (Wisconsin, Illinois, Missouri, Kansas, Wyoming, & Montana in the USA and Saskatchewan, Ontario & Manitoba in Canada).

Invasive annual grasses, such as cheatgrass (Bromus tectorum L.), have proliferated in dryland ecosystems of the western United States, promoting increased fire activity and reduced biodiversity that can be detrimental to socio-environmental systems. Monitoring exotic annual grass cover and dynamics over large areas requires the use of remote sensing that can support early detection and rapid response initiatives. Here, we integrated in situ observations, weekly composites of harmonized Landsat and Sentinel-2 (HLS) data, maps of biophysical variables (e.g. soils and topography) and machine learning techniques to develop fractional estimates of exotic annual grass cover at a 30-m spatial resolution from 2016 to 2018. Comparisons with Bureau of Land Management Assessment, Inventory, and Monitoring (AIM) field data (2016 and 2017) indicate good agreement between observed and mapped values (n = 1700; r = 0.83; mean absolute error [MAE] = 11), as constructed from an ensemble of regression tree models, with slightly lower agreement between mapped values and independent field observations (n = 112; r = 0.65; MAE =14). Geographic coverage of the study area includes portions of Oregon, California, Idaho, and Nevada.

Invasive annual grasses, such as cheatgrass (Bromus tectorum L.), have proliferated in dryland ecosystems of the western United States, promoting increased fire activity and reduced biodiversity that can be detrimental to socio-environmental systems. Monitoring exotic annual grass cover and dynamics over large areas requires the use of remote sensing that can support early detection and rapid response initiatives. Here, we integrated in situ observations, weekly composites of harmonized Landsat and Sentinel-2 (HLS) data, maps of biophysical variables (e.g. soils and topography) and machine learning techniques to develop fractional estimates of exotic annual grass cover at a 30-m spatial resolution from 2016 to 2018. Comparisons with Bureau of Land Management Assessment, Inventory, and Monitoring (AIM) field data (2016 and 2017) indicate good agreement between observed and mapped values (n = 1700; r = 0.83; mean absolute error [MAE] = 11), as constructed from an ensemble of regression tree models, with slightly lower agreement between mapped values and independent field observations (n = 112; r = 0.65; MAE =14). Geographic coverage of the study area includes portions of Oregon, California, Idaho, and Nevada.

Invasion of exotic annual grasses (EAG) and increased wildfire have led to an emphasis on managing rangeland plant communities for resistance to invasion and resilience to disturbances. In sagebrush steppe and similar rangelands, perennial bunchgrasses and particularly their roots are hypothesized to be primary contributors to the resistance and resilience. We asked how bunchgrass root abundance relates to annual grass invasion and aboveground indicators of bunchgrass vigor that are more readily measured, such as plant height. We used a standardized USDA protocol for root measurement in 445 excavations made in 2016-2018 across a topographically and ecologically varied region of sagebrush steppe burned in the 2015 Soda fire in the Northern Great Basin USA.

Invasion of exotic annual grasses (EAG) and increased wildfire have led to an emphasis on managing rangeland plant communities for resistance to invasion and resilience to disturbances. In sagebrush steppe and similar rangelands, perennial bunchgrasses and particularly their roots are hypothesized to be primary contributors to the resistance and resilience. We asked how bunchgrass root abundance relates to annual grass invasion and aboveground indicators of bunchgrass vigor that are more readily measured, such as plant height. We used a standardized USDA protocol for root measurement in 445 excavations made in 2016-2018 across a topographically and ecologically varied region of sagebrush steppe burned in the 2015 Soda fire in the Northern Great Basin USA.