The COVID-19 pandemic forced most higher education courses to use virtual delivery modes for part or all of 2020 (Ali, 2020), which posed a challenge for all disciplines. This change was particularly challenging for the many United States (US) undergraduate geoscience programs, which require field camp or a field course for degree completion (Wilson, 2016). The majority of these field courses had been planned for in-person implementation and were quickly redesigned for remote delivery. Most US universities closed campuses in March 2020 and did not return to in person until fall 2020 or later, whereas the field courses needed to occur May through August 2020. In response to this crisis, geoscience field instructors worked together with the National Association of Geoscience Teachers (NAGT) to develop and share remote field teaching resources through the “Designing Remote Field Experiences” project (Egger et al., 2021).

This paper describes one such impacted course that pivoted to remote teaching, “Geoscience Field Issues Using High-Resolution Topography to Understand Earth Surface Processes”, taught through the University of Northern Colorado (UNC). It was originally planned as an in-person course with Structure from Motion photogrammetry (SfM), terrestrial laser scanning (TLS), and Global Navigation Satellite System (GNSS)

GNSS (Global Navigation Satellite System) is the general term that refers to all Earth's satellite navigation systems. Most people are more familiar with the term GPS (Global Positioning System), which, technically, only refers to the US satellite constellation. Hereafter, this paper will refer to GNSS or GPS/GNSS.

Bywater-Reyes was the primary course designer and instructor for the course and led the adjustments to remote teaching. UNAVCO runs the National Science Foundation (NSF) and National Aeronautics and Space Administration (NASA) geodetic facility (GAGE: Geodetic Facility for the Advancement of Geoscience). Its mission includes providing educational support to the broader geodesy and geoscience communities; thus, UNAVCO staff collaborated on the prepared data collection. The teaching activities developed for this course were adapted from UNAVCO's GEodesy Tools for Societal Issues (GETSI; https://serc.carleton.edu/getsi/index.html, last access: 28 March 2022) modules: Analyzing High Resolution Topography with TLS and SfM (https://serc.carleton.edu/getsi/teaching_materials/high-rez-topo/index.html, last access: 28 March 2022) and High Precision Positioning with Static and Kinematic GPS/GNSS (https://serc.carleton.edu/getsi/teaching_materials/high-precision/index.html, last access: 28 March 2022)

This course and the activities it included contributed to the NAGT Designing Remote Field Experiences collection (https://serc.carleton.edu/NAGTWorkshops/online_field/index.html, last access: 28 March 2022) (Egger et al., 2021). The overall course is at https://serc.carleton.edu/NAGTWorkshops/online_field/courses/240348.html (last access: 28 March 2022), and the individual activities are linked within the course page, as well as contributing individually to the “Teaching with Online Field Experiences” collection (https://serc.carleton.edu/NAGTWorkshops/online_field/index.html, last access: 28 March 2022).

High-resolution topographic datasets (SfM and ground-based and airborne lidar) are valuable in disciplines ranging from geomorphology and tectonics to engineering and construction (Bemis et al., 2014; Passalacqua et al., 2015; Robinson et al., 2017; Tarolli, 2014; Westoby et al., 2012). Use of high-resolution data in Earth science education allows students to quantify landscapes and their change at sub-meter resolution (Pratt-Sitaula et al., 2017; Robinson et al., 2017). Understanding surface processes is listed as very important in the recent “Vision and Change in the Geosciences” with the objective “Students will be able to recognize key surface processes and their connection to geological features and possible natural and man-made hazards” (Mosher et al., 2021, p. 17). Furthermore, use of multiple types of data allows students to practice critical thinking skills such as assessing which acquisition method is appropriate for different scenarios and what errors are associated with different methods. Critical thinking, integrating diverse data sources, and strong quantitative skills were all identified as very important skills for undergraduate students to master (e.g., Kober, 2015). Similarly, making inferences about the Earth system; making spatial and temporal interpretations; working with uncertainty; and developing field, GIS, computational, and data skills were all listed as very important skills for geoscience students to demonstrate (Mosher et al., 2021). Furthermore, learning to collect, post-process, and analyze large datasets is a marketable transferable skill that prepares students for the job market, with cartography and photogrammetry job prospects being “excellent” according to the Bureau of Labor Statistics. For historically marginalized students, high-paying job prospects are particularly important (O'Connell and Holmes, 2011).

Fieldwork, while valuable to building students' self-efficacy and problem-solving skills (Elkins and Elkins, 2007), can pose a barrier to diversifying the geosciences because of ableism (Carabajal and Atchison, 2020), cost (Abeyta et al., 2020), cultural factors (Hughes, 2015), racism (Abbott, 2006), and sexism (Fairchild et al., 2021) in the field. COVID-19 forced the geosciences to develop virtual field experiences, with a positive side effect of removing many of the aforementioned barriers to fieldwork completion. For example, the computer-based nature of remote field learning removes many physical accessibility issues present for typical field courses. The option to learn from home may make the remote courses more feasible for students with family or work responsibilities, as well as reducing real and perceived safety issues related to gender, sexual orientation, and race that may occur in tradition field camp settings. Although remote field courses are not necessarily the most desirable for all students, the development of high-quality remote field options can be one component of diversifying the geosciences (Egger et al., 2021).

The objective of the course was for students to learn manual and remote sensing methods of topographic data collection, including (1) GPS/GNSS, (2) SfM, and (3) TLS surveying and airborne lidar use. GNSS uses ground-based receivers to trilaterate positions calculated from signals sent by orbiting satellites (to accuracies of a couple of centimeters in this use case). SfM is a photogrammetric technique that uses overlapping images to construct three-dimensional models with widespread research applications in geodesy, geomorphology, structural geology, and other subfields in the geosciences (Passalacqua et al., 2015; Westoby et al., 2012). Lidar also generates three-dimensional models valuable for the same range of applications, but it uses laser scanners to send out thousands of laser pulses per second, measure the return time, and calculate distances. Scanners can be ground-based (TLS) or airborne. SfM requires less expensive equipment and less field time but more processing time than TLS. In low-vegetation field areas, SfM can yield similarly valuable high-resolution topographic models with point densities usually hundreds of points per square meter (depending on instrument-to-object distance; Westoby et al., 2012); however, TLS is much more effective in areas with dense vegetation. For both methods, ground control points (GCPs), usually measured with GNSS, are needed for georeferencing the topographic model. For SfM, they are also critical for reducing distortions and errors (James et al., 2019). One of the key outcomes for students was to understand the benefits and challenges of each method and how to determine the most valuable in different circumstances.

Course content focused on Earth-surface process applications but could be adapted to other geoscience topics. The course was taught workshop-style, composed of multiple synchronous work sessions with asynchronous work time in between. The bulk of the instruction occurred within a 2-week period during the summer. Synchronous lectures were conducted via Zoom and course content distributed via Canvas. The class used Slack as an asynchronous way to exchange questions, comments, and solutions amongst the students and between the students and instructor. During the course, students worked with three different analytical software packages: Agisoft MetaShape, CloudCompare, and ArcGIS Map. Five students attended an optional in-person field collection campaign (one student traveled from out of state, and the remainder were UNC students). The course was divided into two units: Unit 1 focused on the SfM workflow, including integrating GNSS and point cloud processing, and Unit 2 on lidar products and workflows, including TLS, topographic differencing, airborne lidar, and method comparison. Each unit ended in a unit report, with the second providing students an opportunity to improve workflows and explore additional data sources and analyses.

The course-specific learning outcomes were that students should be able to do the following: A.

Make necessary calculations to determine the optimal survey parameters and survey design based on site and available time.

The course field site was the Cache la Poudre River at Sheep Draw Open Space (City of Greeley Natural Areas) in northern Colorado. It was selected for the following reasons: (1) the site shows both standard river features and evidence of extreme flooding, (2) the Poudre River is important to several local communities, and (3) the site is proximal to the UNC campus. According to the Coalition for the Poudre River Watershed, “The Cache la Poudre River Watershed drains approximately 2.735 E9 m2 above the canyon mouth west of Fort Collins, Colorado. The watershed supports the Front Range cities of Fort Collins, Greeley, Timnath and Windsor. In an average year, the watershed produces approximately 3.38 E8 m2 of water. More than 80 % of the production occurs during the peak snowmelt months of April through July” (https://www.poudrewatershed.org/cache-la-poudre-watershed, last access: 28 March 2022). In 2013, the Front Range and plains of Colorado experienced extensive flooding. The region received the average annual rainfall in 1 week (Gochis et al., 2015). There was extensive damage to infrastructure and in some cases the erosion of a 1000 years' worth of weathered material (Anderson et al., 2015). Near Greeley, significant portions of the Poudre Trail were impacted as the river topped its floodplain and eroded its banks. The study site is adjacent to the Poudre Trail, with portions of the former trail eroded into the river and the current trail rerouted around the 2013-developed river course.

Data for student use were collected from the Poudre River by a joint UNAVCO-UNC team in May 2020. The types of data included were as follows:

UAS-collected photographs for SfM point cloud generation (DJI Mavic 2 Pro)

Course Unit 1: SfM and GPS/GNSS started out on Day 1 with an introduction to the SfM method. The day's activities were the first step in addressing course outcomes A (survey design) and C (point cloud data). After an overview presentation, students used smartphone cameras to take ∼20 overlapping photos of an object of interest (e.g., sofa, shed, or berm). For simplicity and to learn about local reference frames (rather than global ones from GNSS), they took compass bearing, inclination, and distance measurements and used trigonometry to calculate x–y–z coordinates for the ground control points (GCPs). Students used Agisoft MetaShape software to post-process their photos and create the 3D point clouds. The software was available on their personal computers through a 30 d trial licence. Students then evaluated the performance of their model by considering data quality in different model regions and what method changes might improve their product. They also made recommendations for how SfM could be applied to different fields in the geosciences. The assessment of student learning was based on successful production of a locally referenced point cloud and the data quality analysis.

In the Day 1 activity, students used a relative local coordinate system to produce an accurately scaled model. However, for real-world applications a global coordinate system is frequently preferable, which can be achieved with survey-grade GPS/GNSS, so Day 2 was focused on course outcomes A (survey design), B (GNSS and geodetic survey), and E (justifying techniques). Day 2 morning activities were adapted from the GETSI module High Precision Positioning with Static and Kinematic GPS/GNSS. First, students learned about the method through a lecture. Next, they worked with data collected using different types of receivers and resulting accuracy and precision. Assessment included a concept sketch of GPS/GNSS systems, quantification and evaluation of accuracy and precision of different grades of GNSS, and recommendations for appropriate applications of each.

In the afternoon of Day 2, students were introduced to the field site and methods used for data collection at the Cache la Poudre field location (described above in Sect. 2). Students watched a video (Video 1; https://youtu.be/EZ5I8Ge8YjI, last access: 28 March 2022) about the field site and a video introducing the GNSS methods (Video 2; https://youtu.be/Xpj1QJf8AkY, last access: 28 March 2022). Then, using the pre-collected base-station data, students completed the Post-Processing GPS/GNSS Base Station Position assignment. Students submitted the base station file to the Online Positioning User Service (OPUS), the National Geodetic Survey (NGS)-operated system for baseline processing of standardized RINEX files into fixed (static) positions. For the assessment, students wrote a paragraph explaining their procedure, interpreting the results, describing the difference between ellipsoid height and orthometric height, and highlighting anything that was surprising or confusing about the results.

On Day 3 students combined skills learned in the previous 2 d in order to create a georeferenced point cloud from the field site (course outcomes A–C) and started to consider relevant geomorphic analyses (outcome D). The morning exercise was “Ground Control Points for SfM” at the Cache la Poudre site. This began with a group discussion on where ground control points at the site should be placed within the field area (Fig. 1). Students were then given a text file of the x, y, and z coordinates (UTM) collected by the UNAVCO-UNC team, and had to import them into ArcGIS to create a ground control point map. In a follow-up discussion, students compared the ground control point locations actually used in the prepared data with the locations they discussed for placement in the initial discussion. They were asked to summarize the strengths and weaknesses of the implemented ground control point plan at the site, which helped to assess learning related to both survey design outcomes.

The afternoon exercise was the “Structure from Motion for Analysis of River Characteristics”. Students picked either Area of Interest 1 or 2 for their SfM workflow (Fig. 1). Students with adequate computing power could choose to do the entire study region. Using resolution and height information about the UAS-collected photographs, students first calculated the expected resolution of the final point cloud. They were then asked to assess what types of features or processes at the Cache la Poudre study area they expected could be resolved; from there, they discussed the types of geomorphic questions they could feasibly expect to answer with the dataset of that resolution. Next, students followed a more detailed Agisoft MetaShape guide to construct a georeferenced point cloud of their area of interest. As they were familiar with MetaShape from Day 1, students were able to work through the procedure independently. Once their model was complete, students were asked to answer a series of questions related to error analysis of their model and to reassess appropriate geomorphic applications and design of the ground control point network used. Finally, students were asked to formulate a testable hypothesis related to processes on the Cache la Poudre River that they could answer with their dataset. For example, students could investigate cutbank stable bank heights and angles. The completed exercise was the summative assessment and particularly revealed student accomplishment of SfM point cloud creation and geomorphic analysis.

On Day 4, students used the open-source software CloudCompare (http://www.danielgm.net/cc/, last access: 28 March 2022), which allows for the viewing and manipulation of point clouds. This was a continuation of the same learning outcomes as the afternoon of Day 3 (outcomes C and D) and continued on to some justification of methods (outcome E). Students learned the basic operations used in CloudCompare, such as importing point clouds, classifying the points, and taking measurements that allow for hypothesis testing. They also incorporated an open-source plug-in called CANUPO (http://nicolas.brodu.net/en/recherche/canupo/, last access: 28 March 2022) that facilitates additional point cloud classification (Brodu and Lague, 2012), such as distinguishing between vegetation and ground. Students create a digital elevation model (DEM) from ground points and export it for use in ESRI ArcGIS Map. In ArcGIS, students familiarized themselves with viewing 3D data in 2.5D and created hillshade and slope maps. Then they were asked to retest their hypothesis with tools available in ArcGIS and 2.5D (e.g., measure tool and raster values). Students compared and contrasted applications with the three-dimensional point cloud versus 2.5D raster and summarized the appropriate uses and applications of each in the day's assignment.

The summative assessment for Course Unit 1 was the SfM Feasibility Report, which included assessment of all five course outcomes. Students were to imagine themselves as natural resource managers and assigned the task of investigating the feasibility of using SfM to study geomorphic processes on the Cache la Poudre River. They were asked to summarize the SfM workflow and present the outcomes, limitations, and suggested applications of their SfM model of their Poudre area of interest. On Day 5, students were given a work day to complete the report.

Day 6 consisted of an optional field demonstration during which students completed a GNSS ground control survey, and Bywater-Reyes and colleagues collected UAS images at the Poudre Learning Center (https://youtu.be/s5CGhk8GIOUBrodu; Bywater-Reyes, Sharon: Poudre Learning Center Project. 10.5446/54388).

Day 7 was the start of Course Unit 2: TLS, Topographic Differencing, and Method Comparison and began with an introduction to TLS methodology through a video and lecture. The exercise used pre-collected TLS data that the students were asked to compare and contrast with the SfM point cloud they had developed in Unit 1, which was collected from the same geographic location (Cache la Poudre River) on the same day (Fig. 3). The learning outcomes primarily focused on outcome C (point clouds) but also laid the groundwork for more advanced method comparison to come (outcome E). Students visually inspected the datasets for similarities and differences; then they measured geomorphic features in the scene and compared their measurements for the two methods. Using skills gained in previous class activities, students classified the TLS cloud into vegetation and ground, exported the ground cloud as a text file, and created a raster that matched the specifications of the one made in the SfM activity. This prepared for 3D (cloud-to-cloud differencing) and raster differencing on Day 8. Assessment (mostly formative) was based on their completion of measurements and a discussion of method comparison, including a group discussion.

On the morning of Day 8, students used the concepts of point cloud and raster differencing to further compare their SfM and TLS results and interpret differences between the methods (outcomes C and E). After a lecture on point cloud differencing, students proceeded with differencing of the SfM and TLS data for their area of interest using CloudCompare with the M3C2 Plugin (Lague et al., 2012). Since these datasets were collected at the same place on the same day, differences between the datasets were due to errors or uncertainties in one or both of the models. Students were asked to interpret the 3D differences between the datasets. The second lecture, on raster differencing, discussed best practices in preparing rasters for differencing (Wheaton et al., 2010). Students then used the ArcGIS Raster Calculator tool to subtract one raster from the other. Students interpreted the results and compared the differences between 3D (point cloud) and 2.5D (raster) differencing. The summative assessment was the assignment in which students interpreted their results as an error analysis and discussed which dataset they think is more accurate (and why) and which method provided the most robust error analysis.

So that the students could gain experience with airborne lidar data and with actual geomorphic change detection, during the afternoon of Day 8 they were given two lidar-derived raster datasets collected before and after the 2013 floods of the Colorado Front Range on a river (South St. Vrain Creek) that experienced substantial geomorphic change. In the exercise “DEM of Difference”, students practiced raster differencing skills in the context of geomorphic change detection and also characterized their detection limit with a simple thresholding approach. This helped to further address outcomes C and D as students answered questions in the assignment about the differencing method and made a series of calculations that pertained to geomorphic change.

To broaden student knowledge of data availability, Day 9 focused on additional high-resolution (usually lidar) data sources. After a lecture, students conducted an assignment using existing high-resolution datasets housed within OpenTopography (OT; https://opentopography.org/, last access: 28 March 2022). First, students practiced downloading and viewing data from OT; second students conducted a topographic differencing exercise (Crosby et al., 2011), complementing the point cloud and raster differencing students conducted on Day 8. As with the afternoon of Day 8, the learning outcomes primarily focused on point clouds and geomorphic analysis (C and D). The learning assessment was done via the student assignment, in which students determine erosion and deposition in a dune field and analyze error and detection thresholds.

The summative assessment for Course Unit 2 and the course as a whole was the final “Method Comparison Report” and presentations in the last 2 d of the course. Students picked from a variety of options including improving methods from Unit 1 (SfM and TLS methods), adding new elements to Unit 1, choosing an additional exploration with the datasets collected on the optional field day, or using a different dataset such as airborne lidar. As the course summative assessment, the report pulled together student learning on all five course outcomes. The presentation (Day 11) additionally gave students practice in oral presentation of scientific findings.

This section provides a variety of examples of how students met the different course-specific learning outcomes. It is not intended to be exhaustive but to provide general illustrations of student learning drawn from both assignments and Slack daily reflections and discussions.