Welsh's milkweed (Asclepias welshii) is a federally listed threatened species that is endemic to active aeolian sand dunes in southern Utah and northern Arizona. The plant is a robust herbaceous perennial that spreads via underground rhizomes to form patches that migrate with the dunes. Although plants can potentially reproduce from seed, most spread is by clonal propagation. There are eight known populations, but ca. 98% of plants are in the population at CPSD. Primary threats to this population include OHV impacts and dune stabilization, but population size based on stem counts has increased through time there, prompting managers to consider delisting as a possible option. In the latest five-year review, the USFWS (2015) recommended more extensive surveys and improved monitoring as the highest priority needs for this species. Specifically, monitoring protocols need to be refined to account for the shifting nature of the dunes, to address the impacts from OHVs, and to distinguish or define individual genets versus stem numbers. We have demonstrated that drone-based protocols can be effective for survey, census and monitoring of plants in sparsely vegetated habitat (Rominger and Meyer 2019, Rominger et. al. in press). Drone technology provides a time-efficient and economical way to carry out surveys to locate new populations, to determine current population status through census, and to monitor to detect population trends through time that are critical to extinction risk assessment. Another high priority need for Welsh's milkweed is to acquire a better understanding of its genetic structure and levels of genetic diversity. Without some assessment of the level of genetic diversity within and among populations, it is not possible to evaluate extinction risk. In a strongly clonal species like Welsh's milkweed, where genetic individuals cannot be identified morphologically, molecular markers are the only nondestructive way to accomplish this assessment. An earlier effort to use microsatellite markers from another milkweed species was unsuccessful, but next-generation sequencing technology will make a straightforward genetic evaluation readily achievable.

Our one-year study will be structured around five objectives and associated tasks: 1) Adapt our existing methodology for drone-based survey, census, and monitoring to accommodate the specific attributes of Welsh's milkweed and its habitat, 2) Carry out a first year of drone-based monitoring that is inclusive of areas monitored annually on the ground at CPSD and compare the results of drone-based and on-the-ground monitoring, 3) Carry out a complete drone-based census of the Sand Hills population, 4) Conduct drone surveys at 5-10 aeolian dune areas in potential habitat to look for new populations, and 5) Use next-generation sequencing technology to examine population genetic structure and levels of genetic diversity in the CPSD population.

Our work will contribute substantially to meeting the management objectives for this species identified by USFWS (2015), as well as providing valuable baseline data for land managers of both BLM and state park land as they implement their management plans for the dune ecosystems they manage. Development of drone-based monitoring could lead to a more accurate methodology that will aid in evaluating the relative threats of OHV use and dune stabilization to this species, which relies on the active dune habitat for persistence. Molecular genetic studies will enable evaluation of the threat presented by reduced within-population genetic diversity, a key component of any listing decision for the species, and will also provide a tool for evaluation of among-population genetic diversity in the future.

Drone pilots engaged in this project are fully licensed by the FAA to operate UAVs and will abide by all regulations that govern their use. Permission to use drones for data collection over both state and BLM land occupied or potentially occupied by Welsh's milkweed has been obtained, as well as permission to carry out concomitant on-the-ground data collection as needed. We will obtain permits from USFWS to collect tissue samples and seeds of this listed species for genetic analysis.

Objective 1) Our methodology for image acquisition, processing and analysis as well as our experimental design approach to census and monitoring is described in the attached documents (Rominger and Meyer 2019, Rominger et al. in press). We will carry out exploratory image acquisition and analysis in summer 2021 at CPSD to adapt our existing protocols and make sure they will work for this species. We will fly at multiple altitudes to determine the best altitude for resolving maximum detail and obtaining the best information on stem status (vegetative vs. reproductive) and spatial configuration. Objective 2) To carry out a first year of monitoring we will obtain detailed location information for existing monitoring plots at CPSD and design flight plans that include these plots as well as extensive surrounding areas, including stabilized dunes versus active dunes and areas open to and closed to ORVs. Using drone imagery obtained during flowering in 2022, we will determine the density and map the spatial distribution of vegetative and reproductive stems across much larger areas than those included in the existing monitoring plots. We will include validation on the ground of the accuracy of our image interpretation. We will also compare our results directly to on-the-ground monitoring, assuming the monitoring data are made available to us. We will also work on methods for following dune movement and stabilization through time as well as for evaluating ORV impacts from the imagery. Objective 3) We will obtain drone imagery at the optimal altitude for census over the entire Sand Hills population in mid-summer 2021 and quantify the current number of stems in the population. We will also map the spatial distribution of Welsh's milkweed stems relative to each other and to environmental variables such as dune configuration and total plant cover at this population. Objective 4) In consultation with the Kanab Field Office of the BLM, we will identify 5-10 priority dune areas to survey for new populations of Welsh's milkweed. Our goal will be to determine whether the species is detectable in the imagery, based on our evaluation of imagery taken at the same altitude in known populations of the species (i.e., Sand Hills). Depending on accessibility, any new occurrences can potentially be verified through ground reconnaissance or through drone flights at lower altitude. Objective 5) Conservation-based genetic studies of rare clonal species have been recently accomplished by sequencing a reduced representation of the genome (Amor et al. 2020), through a process known as double-digest restriction-site associated DNA sequencing (ddRADSeq). RADSeq and ddRADseq are techniques that efficiently and inexpensively identify genetic variation across any genome, including those of non-model organisms (Peterson et al. 2012). The techniques are effective because they combine restriction enzymes that systematically cut the DNA with next-generation sequencing that identifies the sequence of base-pairs in the cut fragments. The data are analyzed with software such as GENALEX 6.5 to remove any SNPs (single nucleotide polymorphic loci) that deviate from Hardy-Weinberg equilibrium, and BAYESCAN v. 2.1 to remove SNPs under balancing and divergent selection (Nazareno et al. 2017), ensuring that the analysis is based solely on neutral genetic variation. Population structure, inbreeding coefficients, and heterozygosity are calculated with GENALEX 6.5, and estimated pairwise genetic differentiation (FST) with SPAGeDi software (Hardy and Vekemans, 2002). We will use this approach in our molecular genetic study of Welsh's milkweed population structure. Sampling for our study of clonal structure and genetic diversity in Welsh's milkweed will involve the creation of a grid over a single "population" of milkweeds at CPSD followed by tissue sampling of leaves from stems within each cell of the grid. Grid cell size will be determined by estimations from previous studies of approximate milkweed runner lengths, and sample locations can be marked and recorded by drone. Tissue samples will consist of 1 cm2 of green non-senescent leaves dried on silica gel and stored at room temperature. DNA will be extracted with a Qiagen DNeasy Plant Pro kit according to the manufacturer's instructions and shipped to an external provider such as Floragenex or Oregon State University for specialized ddRADSeq analysis. Jacqualine Grant and students at Southern Utah University will perform bioinformatics to assess clonal structure and determine next steps for more in depth and/or wider scale genetic analyses across multiple populations.

As a primary goal of our project is to determine the feasibility of improving monitoring through the use of drone imagery, a logical follow-up to our one-year study would be to carry out a second year of monitoring to determine how well it will work across years. If successful, drone-based monitoring could become the accepted protocol for this species. Any loss of resolution will likely be more than offset by the larger areas that can be monitored and the ability to locate stems precisely in the imagery relative to each other and to processes that affect their relative success in different parts of the dune system. We intend to provide detailed flight plans and image analysis protocols that can be used by managers or their contractors to continue drone-based monitoring if it is deemed a suitable approach for Welsh's milkweed.

Principal investigator on the project will be Dr. Michael Stevens of Utah Valley University, who will oversee all aspects of the work. Co-principal investigators include Dr. Susan Meyer of the USFS Shrub Sciences Laboratory, who will be involved with the design and reporting of the drone research and will be responsible for conducting associated fieldwork, and Dr. Jacqualine Grant of Southern Utah University, who will carry out the population genetics component of the project. Project Manager Kody Rominger of Utah Valley University will carry out the drone and GIS components. Our other partners include Aaron Roe, BLM Utah State Botanist, Lisa Church of the BLM Kanab Field Office, Josh Hansen, manager of CPSD State Park, Mindy Wheeler of the Utah Department of Natural Resources, and Jena Lewinsohn of the USFWS Salt Lake Field Office.

This project will provide information on species distribution, population status, and genetic diversity that will be valuable in the process leading up to consideration of a delisting decision for this species. However, conservation management for species persistence will remain in place for many years even if delisting were to be recommended, so that the methodology and new knowledge we provide will continue to be useful to management over the long term.

07/01/2021

06/30/2022

2022

A principal objective of the project was to use Welsh's milkweed as a second test case for the feasibility of using drone imagery for efficient census and monitoring of a rare plant. We carried out census flights over the entire area of Coral Pink Sand Dunes (CPSD) in summer 2021, a total of 131 flights at 40m agl (above ground level) across 1400 ha. This required 10 days in the field for a single drone pilot. Our methodology involved use of a DJI Phantom 4 Pro quadcopter drone for image capture, Pix4D for image processing, and Adobe Photoshop for color correction. The resulting orthomosaics were then subjected to inference using a customized version we developed of the YoloV5 deep learning (artificial intelligence) program that had been trained on known Welsh's milkweed images to produce counts and map locations for Welsh's milkweed clumps across the dunes. We used validation imagery obtained at 15m agl to verify identification accuracy. For the Sand Hills population, we used a drone survey methodology to determine which areas were occupied by Welsh's milkweed, then performed census flights for the occupied areas. This survey methodology involves systematic manually controlled flights across the survey area during which the imagery is examined in real time. We have not yet specifically developed a methodology for population monitoring, but examination of validation imagery obtained as part of the census protocol showed that this approach would be highly feasible. The survey methodology will be useful as we expand our efforts to census other known populations and to locate new populations using a fixed wing VTOL mapping drone we will acquire later this year.

This project as originally conceived had five principal objectives: 1) Adapt our existing methodology for drone-based survey, census, and monitoring to accommodate the specific attributes of Welsh's milkweed and its habitat, 2) Carry out a first year of drone-based monitoring that is inclusive of areas monitored annually on the ground at CPSD and compare the results of drone-based and on-the ground monitoring, 3) Carry out a complete drone-based census of the Sand Hills population, 4) Conduct drone surveys at 5-10 aeolian dune areas in potential habitat to look for new populations, and 5) Use next generation sequencing technology to examine population genetic structure and levels of genetic diversity in the CPSD population. For Objective One, we determined that Welsh's milkweed was an ideal choice for drone census and monitoring, as it was readily identified in drone imagery and not easily confused with any other species present. For Objective Two, we opted to complete a census of the entire CPSD, a task not in the original proposal, rather than developing a specific monitoring protocol. We made this decision when we realized how feasible this was and how useful it would be for comparison with the 2003 census map produced by Kneller. While we did not formally develop a monitoring protocol, examination of 15 m agl validation imagery from the census made it clear that it would be very straightforward to develop such a protocol. We plan to complete this part of the project with BLM funding we have obtained for summer 2023. For Objective 3, we carried out survey flights to identify occupied areas at Sand Hills and obtained census imagery for occupied areas. This census information will be provided to the BLM Kanab Field Office later this fall. For Objective 4, we obtained permission to use a drone to survey for new and existing populations in the area along the Utah-Arizona border late in the summer of 2022, but the stormy weather did not permit us to complete this work prior to plant senescence. The survey work will be completed in summer 2023 as part of our project with the BLM. For Objective 5, we obtained ca. 50 tissue samples for genetic analysis from across both ownerships at CPSD in July 2021. We also systematically sampled three apparently clonal patches at northern, central, and southern locations within the dunes. We used drone imagery to design this sampling protocol, which involved sampling clumps at known distances from each other to determine whether they were clonal, possibly closely related by descent, or only distantly related. This resulted in the collection of ca.105 additional tissue samples, i.e., 35 clumps sampled per patch. DNA extraction/quantification was completed in spring 2022, and resulting DNA was determined to be of generally high quality. Unfortunately, due to logistical difficulties with funding and facilities, these samples were only recently sent for sequencing, so that we do not yet have data to address our research questions. We expect to be able to provide a supplemental report on this part of the project by the end of the calendar year. We also report here on an additional study we carried out on the seed and seedling ecology of Welsh's milkweed that was not part of the original proposal. We compared seeds and seedlings of Welsh's milkweed with those of the common roadside species Asclepias speciosa. We found that seeds of Welsh's milkweed were >5 times heavier than those of the common species and also had a much smaller relative investment in the hairy coma that aids in wind dispersal. These features could limit its dispersal by wind, which would reduce the likelihood of dispersal beyond its island-like habitat. This low dispersal capacity could also explain why apparently favorable dune areas have not been colonized. These very large seeds also have advantages in the dune environment relative to those of the common species. They produced absolutely larger seedlings than those of the common species in spite of a much slower relative growth rate, a common adaptation to stressful environments. Seedlings of the common species could not emerge at all from sand at depths >2 cm, whereas Welsh's milkweed seeds emerged to 56% from 10 cm depth. Please see the attached document for a full report on our results to date on this project.

The key management question to be addressed for Welsh's milkweed is whether it requires protection as a threatened species in order to have a high probability of persisting into the foreseeable future. Our CPSD census shows that the large population there has likely increased in abundance in the last 20 years and has certainly colonized some new areas within the dune system. The effects of ORV traffic and dune stabilization on patch persistence, as well as movement of both sand and milkweed patches through time, could be evaluated at relatively low cost with census-level imagery and analyzed with a deep learning approach. This could be combined with on-the-ground reproductive output monitoring that cannot be achieved with drone imagery. The wider distribution of the species, and factors limiting its occurrence and abundance on other dune systems, also need to be examined. Lastly, as we hope to accomplish later this year at least for CPSD, the level of genetic diversity across the species range still needs to be evaluated. Even large populations can succumb to catastrophic events if diversity is too low, and the degree to which this species relies on clonal reproduction for persistence is still not known.