FOPHeaderColorLogoWhiteAndGreen
Search

The Pando Photographic Survey

A Methods Review of a Photographic
Survey Using 360° Imagery

Dont have time to read this now?

Overview

The Pando Photographic Survey is a first -of-its-kind, baseline re-photography project to document Pando, the world’s largest tree. Spanning 106-acres of the Fishlake Basin in the Fishlake National Forest in Southern Utah, Pando is the world’s largest tree by weight and by land mass, as well as the largest aspen clone. A relatively recent discovery, whose size was only definitively verified by genetic testing in 2008 (DeWoody, Mock et al, 2008), Pando has never been fully documented. The goal of the project was to create an immersive baseline photographic record that land managers and scientists can use to study the tree and later, replicate our work, to monitor changes in the tree over time. We argue 360-degree imagery offer advantages and efficiencies over 2D photography especially for documenting large scale subjects in a rigorous way. We also discuss 360-degree imagery’s value as an way to study subjects remotely either adapting existing field methods, or, emergent methods. This review details the rationale for the project and the methods used to map, name, record, process and publish this record which captures 8,593 locations within the tree in high-resolution, 8K, 360-degree still images. An effort that can be replicated with ±30cm accuracy between planned and recorded location points and no more than ±3m along the ground which, volumetrically translates to about the size of small stem on the smaller size or, two medium sized branch bodies on the larger size.

The Use of Photography in Land Management and Research

The emergence of photography in the 1830’s marked a new era in both creative and scientific documentation of natural subjects. When used rigorously, photographs provide a way to document unique subjects and observe how they change over time. While likely unfamiliar to the general public, re-photography plays a critical role in the protection and care of many public lands we enjoy today, especially immersive photography. By way of example, Carleton Watkins’ stereographic images of Yosemite National Park, (analogous to today’s 180-degree and 360-degree photography), played a critical role in efforts to raise awareness and protect what became Yosemite National Park (Tyler, 2020). Creating a rigorous photographic record involves the creation of a baseline; a first record of location data and images that describe them. Once a baseline is established, the record can then be used for study of the given subject and, used to re-record the subject to observe changes over time. Just as photography has played a role in the study, monitoring and care of other natural subjects, this project’s aim was to create a record that could be used to study Pando remotely via any number of human or machine methods. For use in research planning and, as a portable reference for field work. Just as Watkins work inspired awareness and shared understanding of what became Yosemite, the effort is also a tool a tool to educate the public on a natural wonder without compare largely unknown and often misunderstood despite some 40 years of work to study and protect the tree.

The Re-emergence of Virtual Art and 360 ° Photography

The availability of equipment that can take 360-degree images has rapidly advanced in the past decade with innovations in materials design, integrated circuitry, lens manufacturing and software. While those advances arguably, have led to a resurgence of interest in immersive art forms, they are not new. Immersive art forms and strategies have been used in science, art, conservation, education and theater dating back to ancient times (Grau, 2004). Today, documenting large scale natural subjects typically involves one of three methods: Satellite imagery (more generally, aerial methods), Survey plots and lastly 2D Photography. Each has their drawbacks as it relates to the study of Pando. Aerial imagery cannot penetrate the canopy and provide a ground level detail which contains vital information about the tree. Survey plots rely on both statistical concepts and field observation to generalize claims, but are limited in that they also must maintain, a group of plots in a forested landscape should look and behave in a consistent manner, which we do not see with Pando. At the time of this writing, we lack a standard model or peer reviewed research study that has definitely described Pando’s expected or actual rate of regeneration, which varies from clone to clone. 2-D photography can be used, but compared to 360-imagery, is lower quality, more time consuming and more error prone. What’s more, 2-D photography produces images that lack qualities consistent with how field work is conducted on Pando today, immersively.

 

360-degree photography produces high-quality images that describe defined location points in the Pando Tree from the ground level into the canopy. Carried over to the land scale, location points used to document the tree can serve as a master plot point map or, can be used along with other methods to interrogate and support emergent or previous. As Pando is a remote subject only accessible for study a few months a year, a rigorous 360-degree record provides a way to study the tree in detail, year-round. Lastly while 360-degree photography may a be novelty to many, its use is not unprecedented and we will show, offers numerous advantages over traditional documentation at the same time, it provides a robust and highly felxible framework that serves as a robust analog to in-person field study, especially for remote subjects most would be unable to study otherwise.

Mapping Pando

In order to create a document that can be replicated over time, we first had to describe Pando in terms of location points we would record and capture with our cameras. All GIS data products from this project relied on the WGS84 datum. Grid points were sets up based on Corner ID: UT260260S0020E0_100300.

Figure 1: Aerial View of the Pando Photographic Survey Location Map
Figure 2: Google Earth Image of Pando Photographic Survey Grid with Overlap

Creating A Grid

To create a grid of locations to document, an area map of Pando land mass was needed. While multiple area maps exist, we relied on a land mass map developed by Paul Rogers and Darren McAvoy – itself based on the 2008 study that first verified Pando’s size via genetic testing (DeWoody, Mock, Hipkins, Rowe, 2008). Pando’s stated size is 42.896 hectares (106-acres) while the Rogers/McAvoy map of the tree shows an area of 104 acres. Thus, we divided the map into 8,589 grid points offset 7m on center, based on the 42.896 Hectare (104-acre) map. While we do not count it in the record, a quick estimate suggests that overlaps along the area map boundary would cover another 3.6 acres of Pando’s land mass (3.5m per one half shot x 4217m perimeter). Thus, our claim the record serves as a comprehensive document of the tree and the land it calls home.

The 7-Meter Grid

The choice to use the 7-meter grid was reasoned. The choice was made based on equipment capabilities, Pando’s physiology the safety of crews and equipment. Field studies by the lead photographer showed that an 8K, 360-degree camera at chest height, could reliably capture high quality image data up to 3.5m (11.4ft) in each direction with a strong bias for ground cover. As mentioned, ground cover and sub-canopy study is critically lacking today. The 7m grid also provides meaningful overlap with nearby locations; each of the camera’s six-lenses captures 3.5 meters of high-quality information (barring obstructions) outward along the ground and well in to the canopy. The cameras provide medium and lower quality information out to 5m. What’s more, each location point overlaps with 3-to-8 additional locations points on the planned location grid. In terms of the tree’s physiology and the humans who worked to document it, each of Pando’s estimated 47,000 branches can reach diameters up to 1m (3.28 feet), so if crews encountered a cluster of mature trees, 7 meters provides ample workspace for 2 crew members and equipment (around 4m of space) to safely enter, work and exit.

pando photographic survey grid map
Figure 3: Detail sof 7m grid detailing relationship between each point
002SampleAreaMap
Figure 4: Deatil showing area capture by each location dcoumented

Exempted Coverage

While the plot map and grid describes 8,593 locations points, Utah Highway 25 bisects Pando’s upper and lower elevation regions. Locations points in the highway, whose speed limit is 55 Mph, were exempted from capture out of concern for the safety of crews and equipment. Exempting road locations, left 8,542 unique locations to document. In addition, leaders exempted additional routes (described below) which were simply unsafe to enter and work in. In addition, while not explicitly exempted, as part of field crew safety training, team leads discouraged crew members from entering any given point where they, the equipment or their field mates were likely to physical harm (snag piles, larges juniper bushes etc). In the general record, if a given location was unsafe to enter, they were simply marked as obstructed or unsafe.

Organzing Work

In order to organize effort to document Pando, each location defined by the grid needed a unique name. In addition, field crews and scientists would need a common language to work from as reference. Minding that, we developed the following system for naming location points, regions, sections, and routes.

Naming Location Points

The center of each grid point Pando was given a “Letter + Number” combination. The location name defines the point in relation to North, East, South, and West in Pando’s land mass. The top-most row of grid points to the north start with the letter “A” and continue “B, C, D…” as you move toward the south side of Pando’s land mass. Letter sets were doubled as needed, for example “AA” ultimately reaching the conclusion of row “EF”. Numbers are incremented one-at-a -time as each the sequence advances from west to east. For example, A66 is 7m (22.96ft) west of A67. A68 is 7m (22.96ft) west of A67 and so on. It is important to note, as recorded locations are 360-degrees and in sequence, you can head in any direction relative to the initial location. Adjacent routes also provide image data moving towards or, away from any given point. This allowed field crews to work more efficiently, while the final product would allow researchers to construct any number of unique work paths heading along a single path, or creating a unique sequence with points adjacent to one another, ex: BX64, BY64, BZ64.

Regions

For field work purposes, we describe Pando in terms of two-prefixed, regional names. The Upper Pando and the Lower Pando. The “Upper Pando” refers to all locations above Utah Highway 25 working toward Pando’s highest elevation point, approximately 2802m (9,200 ft). All points below Utah Highway 25 are referred to as “Lower Pando” as they move toward Pando’s lowest elevation point, approximately 2712 (8,900 ft).  It is important to emphasize, Upper and Lower are common names. There is evidence no Pando has been broken up by the Highway, which was built well before Pando’s discovery and, the 2008 paper that first described the tree as a genetic clone spanning 106 acres.

Figure 5: Region Map of Pando used by team

Sections

Once work regions were defined, we further divided the land mass into “Sections”. A “Section” earns it common name its relation to primary landmarks in the area. In all, we defined 9 logical sections.

Figure 7: Simplied version of shoot map showing boundary lines of eachj section

Upper Pando Sections

  • Cabins (“CR”): An area populated by Cabins operating on easements generally built before Pando’s discovery.

  • Upper Pando Exclosure (“UPE”): The largest portion of the tree protected by “ex-closing” fences.

  • Upper Pando Hill (“UPH”): A section spanning approximately 6 acres along Pando’s highest elevation point

  • Upper Pando Clean Up (“UPC”): A small section of the tree outside the fence along the Upper Pando Exclosure Section

Lower Pando Sections

  • Dr Creek North (“DCN”): A small portion of the tree that runs adjacent to Dr Creek Campground
  • Dr Creek Campground (“DRC”): A portion of the tree that features an area set aside for recreational uses
  • Lower Pando Exclosure(“LPE”): The second largest portion of the tree under protection using “exclosing” fences
  • Lower Pando, Lakeside (“LPL”): A small section that runs outside of the Lower Pando Exclosure protective fence along Coots Slough
  • Lower Pando South (“LPS”): A large portion of the tree along Pando’s southernmost extent

Organizing Effort and Creating Logical Sequences Using “Routes”

Each Section of Pando was further divided into “routes”. Routes contain a series of logically related location points in order of the named location point. For example, Cabin Route 1 follows points A66, A67, A68 west through point A91 in the east.

Figure 7: Route 31 in Upper Pando, which only includes one row inside the section boundary.

For efficiency’s sake, in some cases, routes could also move “out and back”. For example (as shown below), Lower Pando Exclosure Route 20, sequences locations CF52 through CF74, then turns around and sequences CG74 through location CG51. As the images are shot in 360-degrees, the final result is not altered in any way by this approach to field work. As discussed, each point is bounded by at least 3 and, up to 8 adjacent location points while the sequence moves toward and away from any given location in 2D space. 

Figure 8: Route 20 in the Lower Pando Exclosure, with two rows of a single route within a section boundary

Survey and Camera System

While we explored both one and, two person teams to create the record, the most efficient and accurate method was found to be two-person teams: one surveyor and one camera operator. Once trained on use of the survey and camera systems, a given team could capture one shot every two minutes on average once they arrived at the starting point.

Survey System

In order to record location data in the field, we used an Arduino-based RTK Base and Rover Survey system. Each team was given a rover pole and battery units also connected to an Android phone (various models) via USB connector. In addition, the Phone was loaded with GPX location waypoints and, the freely available SWMaps App which surveyors used to get headings and document locations.

Figure 9: Image showing Survey and Camera Rig used by field crew
arduion rtk survey rover poles
Figure 11: RTK Rover Poles
arduino rtk base station
Figure 10: Arduino Based RTK Base Station

Survey System Set Up and Operation

The RTK Base station was set up each day at a fixed location. Once calibrated, surveyors would connect their mobile devices to the rover unit via USB Cable and search for signal using the SWMaps App. Once signal was established, surveyors could then use the preloaded GPX maps to get their heading to reach the first location on the assigned route. The mobile devices also recorded “digital” waypoints whose location was attained by the rover, and labelled by surveyors and saved in the SWMaps App. That data was retrieved and used to determine accuracy comparing planned location coordinates with recorded location coordinates.

Recording Locations

Each location was defined by its common location name and a Latitude and Longitude coordinate. Once a surveyor arrived at a given location, for example “BX59”, they dropped a “digital pin” named “BX59” in the SWMaps app, then placed a temporary flag in the ground which the camera operator would use to set the camera down to record the location.

Survey Accuracy

After recording, the coordinates of planned locations were compared with coordinates recorded by crew surveyors. Comparing planned location data against recorded location data, we observed ±30 cm accuracy under field conditions. The variance is likely explained by the tree itself and human error as the unit has a stated accuracy of ±2 cm. For example, the tree’s aerial mass could interfere with signal while surveyors could tilt the pole marking a location point.  We now turn to a sample of how that variance looks in 2D space.

Planned-locations-map
Figure 12: Image of Planned Locations in LPE Route 1
Planned-and-actual-locations
Figure 13: Image of Actual Recorded Location in LPE Route 1

It is important to note, while the planned locations are mapped in 2D, images are documented in 360-degrees with meaningful overlap between each point. Thus, ±30 cm in 2D, only translates to an object about 3.6 inches in volumetric space. An object, it is important to note, could very well be visible in 3 to 8 adjacent points.

Camera Systems

As already discussed, 360-cameras offers advantages over 2D Images for documenting large scale subjects. While any 360-camera with 8K image capability (7680x4320px) could be used to replicate our results, we chose to use Insta360 brand of professional cameras. The cameras are lightweight, portable feature good battery life, and offer many software and hardware features that simplify their operation by non-professional photographers. In addition, we chose these cameras as they also feature inbuilt gyroscopes that allow images to be automatically leveled when stitched, automating an otherwise time consuming process. While they are not cheap to own, Insta360 camera set-ups are affordable to rent from places like Lensrentals.com (one of our sponsors).

>>>Authors Note: Learn more about Insta360 One X2 for smaller scale projects.

For this effort, we primarily used Insta360 Pro and Insta360 Pro2 camera, whose feature set and operation were nearly identical for use. In addition, for one small portion of Pando we found too dense to enter with the larger camera rigs, we used the pocket sized Insta360 OneX2 cameras on slim microphone stands oversampling by a factor of 2 as they only cover about 1.5 meters of high quality image data on any side of the location. While our work was primarily shot using the Pro line of of Insta360 cameras, we encourage others to consider use of “pocket sized” 360-cameras like the OneX2 for work on smaller areas.

Field Rigging Cameras

Pando’s remote, high mountain home features a rugged landscape that demands advanced rigging as we found through trial and error. A landscape that ground is soft and level in places, but, by and large, littered with piles of downed trees, lava fields, car sized juniper bushes and countless fox holes all of which can be hazardous to crews and equipment. To keep the camera systems stable, we explored two approaches. A standard microphone-stand with an 8lb base with microphone-to-1/4 20 adapter used to attach cameras. An arrangement suitable for the lightweight One X2 cameras, but as we found, not durable when working with the larger Pro line of cameras. For the Pro line of cameras, we used Rocky Mountain Style C Stands with a Baby Pin-to-1/4 20 flat based adapter to mount cameras. We found this arrangement optimal as Rocky Mountain C Stands have legs which feature high arches that provide overhead when they are placed over large objects. They also feature a quick release adjustable leg that allows you to adjust the height of that leg and  quickly level the stand on uneven surfaces.

Camera Control Units (Trigger and Time Stamping to Develop Sequences)

In addition to the Rocky Mountain C Stands, camera rigs were outfitted with an Android phone (Various Models). Those devices were then mounted to the the Rocky Mountain C-Stands using bicycle cell phone mount with a ring clamp. The mobile phones were loaded with Insta360’s camera control software to allow operators to control the camera via the mobile device. While you can trigger the camera manually without the use of the additional phone and control software, we chose the software approach as it allows full access to the camera’s settings and provides a familiar interface for those familiar with use of cell phone cameras.

 

It is important to note for those unfamiliar with the Insta360 Brand Pro and Pro2 cameras, that it is strongly recommended you work to limit sequences of shots to around the time one of one battery’s lifecycle; 70 minutes. This as timer units may fail from time to time (with or without the app as we found) throwing off sequences as we found in 2021 and 2022. We also recommend you replace the SD card after each battery change or, each time you power down the camera noting the SD card label and the time. This will provide a way to sequence images based on the default time stamp should you also see this problem arise.

Camera Hardware and Capture Settings

Insta360 Pro and Pro2 camera’s both produce high quality, 8K, RAW 360-degree images which can be processed with great latitude just as any RAW image can be. As the team was primarily made up of non-professional photographers, Lead Photographer Lance Oditt developed a series of “base” hardware and “Capture” settings. The camera rigs were set up as follows.

  1. Image Type: 8K RAW
  2. Fan: On (To prevent camera processors from overheating)
  3. Delay: 10 seconds (to allow crew to move out of field of view)
  4. Speaker: On (to ensure crews could hear “ding” once the camera has recorded an image)
  5. Audio (Microphone): Off

>>>Authors Note: Learn More About Operating Insta360 Pro Cameras

Capture Settings

Both Insta360 Pro and Pro2 feature six lenses which shoot at a base aperture of f/2.4. While the camera allows for fine grain control of aperture, ISO, White Balance and Exposure, Pando’s and its homeland provide unique physical challenges to consistent quality image capture; wind. Pando’s location in a high mountain basin with steep sides means the tree is near constant motion at the ground level or, in the canopy above. The challenge then with camera settings, was how to avoid blurring which cannot be easily remedied in 360-degree images. Generally speaking, the smaller the “f/number”, the faster the lens responds, gathers information and closes again. Thus, we undertook a number of tests and actually found the camera’s deafult image controls to be sufficient for producing consistent and good quality images.  Remarkably, we have yet to find any substantially blurred images despite the fact effort was carried out under intensely windy conditions. For non-technical photographers, this means the process can be replicated without the need for specialized knowledge so long as images are captured in the RAW format using the cameras base settings.

Equipment List

Surveyor
  • RTK Base Station
  • 1 Rover Pole
  • 1 Bicycle Style Mobile Phone Mount with clamp to attach to the Rover Pole
  • 2 USB Cables (one for antenna and one for battery connection)
  • One Android Mobile Phone loaded with SWMaps App
  • One mobile phone power bank
  • One USB Cable splitter to connect mobile phone battery and Rover antennae to phone
  • 25-40 Survey Flags with a flag scabbard
  • Clipboard and Pen
  • One laminated slate card and whiteboard markers
  • Route Sheets
  • Safety Equipment: Water, Food, 2-Way Radio, First Aid Kit, Hat, Sunglasses, Long Pants, Close Toed Shoes
Camera Operator
  • Insta360 Pro (or Pro2) Camera
  • 8 Insta360 Pro Batteries
  • Rocky Mountain C Stand
  • 1 Bicycle Style Mobile Phone Mount with clamp for Rover Pole
  • 1 USB Cable (to connect mobile power bank and the phone)
  • Android Mobile Phone with Insta360 Camera App loaded
  • 8 Pre-Labelled SD Cards
  • 1 Mobile phone power bank
  • Blank 3.5 x 5.5 Field Book and pens
  • 1, Microfiber Cleaner Cloth to wipe lenses
  • 15-20 Alcohol Prep Wipes to clean lenses at regular intervals
  • Safety Equipment: Water, Food, 2-Way Radio, First Aid Kit, Hat, Sunglasses, Long Pants, Close Toed Shoes

Quality Assurance & Processing

Route Sheets and their related images were collected from field crews at the end of each day. Images were reviewed for slate shots (captured to mark the beginning and end of an assigned route), date/time sequence (timing), whether there were missing or obstructed shots and, overall image quality. If the documented routes passed quality checks, image sets were labelled and stored on two back-up systems for safe keeping until image processing work could begin. 

Location Data

As discussed above, the surveyor phones stored documented location data which allowed us to compare the planned shot location against the location recorded by surveyors. In post-production, a cross sample of planned versus actual locations demonstrated high accuracy; ±30 cm. Accuracy made more meaningful by the fact that each location can also be cross-referenced by data in at least 3 and, up to 8 adjacent locations.

Processing

Processing 360-degree images is a relatively easy process for anyone familiar with basic image viewing and editing software. The process we undertook involves Sequencing, Stitching, Processing and Metadata Injection. Each step offers varying degrees of automation depending on your level of expertise, the desired results and, time constraints.

Sequencing

As discussed above, locations were documented in order of their appearance on a given Route. Recording the start time and end time of when a sequence of shots was captured along a route, we were able determine that each shot correlated with the previous and next via time stamp, by total number of shots, and by visually comparing each image in the sequence. Once the sequence of the image set for a route was validated, images were given their location label and stitched for manual processing.

Stitching

360-images are comprised of 6 images which are “stitched” together to create a single equirectangular image. Software freely available by Insta360 allows fine grain control of stitching quality while automating work to bring the six images together, level them and output a single Adobe RAW .dng file with the metadata tag of “360”. Insta360 offers their stitching solutions to anyone with a camera serial number. If you rent your cameras in order to replicate this work, or, use an Insta360 Pro line camera on other projects, feel free to use our Serial Number IPE0918NTSW8S6 to download software for your project to replicate results detailed below.

screeen grab of the insta360 stitcher software

Settings to Stitch Images for the Pando Photographic Survey

  • Content Type: Monoscopic
  • Stitching Mode: New Optical Flow
  • Sampling Type: Fast
  • Blander Type: Auto
  • OpticalFlow Stitching Range: 20
  • Template Stitching Range: 0.5
  • Use Default Circle Position: Selected
  • Smooth Stitch: Selected
  • Gyroscopic Stabilization: Selected
  • Use Nadir Logo: Unselected
  • Resolution: 8K (Will Output 8K RAW Adobe DNG file based on camera setting at time of capture, “8K RAW”)

Image Processing: RAW and JPEG Outputs for Different Uses

Image processing is an interpretive process as cameras can see more than the human eye. For this effort, we recorded images in RAW format to ensure all information the camera sensed was recorded at the time of capture and maintained throughout the process. In technical terms, 360-degree cameras create advantages in the field documenting large subjects, but also create unique challenges in image processing because of how they document light; every image can and often does, capture the sky and sun which can skew exposure, color quality, highlights and shadows leaving images “washed out”. Thus, for viewing and immersive usage purposes, each stitched image file requires global adjustments to re-balance visual information. Thus, we used Adobe Photoshop CC to balance color, exposure, shadows, highlights, saturation etc. optimizing for ground cover for scientific use.  Conversely, images processed for educational uses focused on consumer device color models.

Once we made global adjustments, we then saved a copy of the RAW image file as a high quality JPEG file. To provide the most interpretive flexibility, we make both RAW and JPEG versions of each image available in the freely available download of each route data set.

Metadata Injection: General Information about Each Image and GPS Data for Each Image

Metadata allows creators and those that use images, a way to share general information about the image, its origins unique characteristics, and permissions. In the case of this project, we used the IPTC Core and IPTC Extension as our base metadata schemas. We also utilized the “GPS” portion of IPTC Core  provide location data for each image. While locations were recorded in decimals, GPS location data from was converted from decimal to minute format. While many programs are available to write, modify or amend metadata, having 8,542 images that each have a unique GPS coordinate associated with them, meant batch processing was of limited value. While we could process images in bulk for general information, we would have to manually inject GPS information for each image or, develop a bulk method to inject GPS data correlated with the given location name.  To solve this problem, production team volunteer Cade Wolcott developed a batch metadata injection script to simplify insertion of GPS LAT/LONG coordinates into each image. In addition, developed a machine based quality assurance script to confirm a location, for example BX59 matched the recorded coordinates captured in the field. The result was evaluated by the lead photographer and chief scientists for precision.

 

As part of our commitment to open science ethos, you can download and use the script to replicate our results, or for other projects using the link below.

Review and Download Our Batch GPS Metadata Injection Tool: https://github.com/wolcade/pando-metadata

Publication of Images and Data Sets

Publishing the document involved three streams of effort; packaging, display and distribution.

Packaging

Once image processing was complete, we compiled each route into its own folder. Each route folder includes a sub folder containing the RAW stitched images and the processed jpeg images. The route folder contains a digital version of the route sheet which includes location data and general notes and information on its capture. Finally, each package contains a shape file of all location waypoints, a large-scale version of the complete shoot map and lastly, a text file featuring general notes on the effort, usage rights, usage, citation, the team, and our thanks to various people who supported this first of its kind effort.

Display

While RAW 360-degree images are viewable using a variety of image viewing software packages, in order to view them online they generally must be in JPEG format. Thus, the processed JPEG images from each route were uploaded to virtual tour host provider Kuula. Once uploaded, Kuula allows you to gather and “share” a snippet of code that can be pasted into an HTML document allowing people to view the given image set in their viewing interface.  While JPEG is a ubiquitous format, we chose a third-party provider to display images out of concern for server loads; each image can between 5-20MB, while a route could contain upwards of 60 images.

Publication

Each section of Pando Photographic Survey hub (“Explore Pando in 360”) features a listing of routes published to date. For each route, we offer two options for use. First, a way to view images using the aforementioned Kuula Interface on a unique web page. Second,  a link to download the entire package of the data associated with the route using large file transfer service “WeTransfer”.  Thus, we provide everyone a means to view the images, or, gather data for offline uses.

Bringing Together the Parts and the Whole

For this effort, we mapped, documented, and recorded individual locations of Pando in a way best suited to equipment and organizing field work. The results, however, do not demand that the record be used in the way we organized effort. Thus, while we cannot anticipate all uses, here are some of the ways the record and freely available data sets could be used or modified for use in remote research, or, in support of field work.

  1. Each route can be used as singular path for study
  2. Each route could be used for planning field work in a given area or location
  3. Two adjacent routes could be joined together into a unique “sub section”
  4. Any number of locations could be used to create “virtual” paths or plots
  5. Adjacent points could be sequenced to provide a virtual trail or to create plot lines EX: BH74, BI74, BJ74, BK74 and so on
  6. An image (RAW of JPEG) or, a series of images could be processed using machine learning models to isolate information on canopy cover, disease or, to simplify physical measure of volumetric space
  7. Any route, self-defined sub-section or self-defined path could be used for observation of general ecological characteristics of the larger landmass such as forbs, wildflowers, fungus, geologic features.

While some of the processes and approaches may be new to those not familiar with 360-degre photography or photographic publishing, we believe they do not vary so much from more traditional data publication practices; gather, package, upload, report and share. In terms of uses, while our work was constrained to approaches that favored efficiencies in data gathering, it is our conviction, the system provides enormous flexibility for remote study, or to support field research. While we are only able to publish about 20% of the record per month based on the size of our team at the time of this writing, we worked to ensure that there are adjacent routes for each section to start, while the larger record will gain power and value with each new addition over the coming months.

Margin of Error

Photography and re-photography are descriptive approaches. In many ways, either the image was captured or, not. In the case of this effort, there is no ready way to isolate nor, easily transliterate errors between the 2D Grid and a volumetric image. Despite this, we offer that the Pando Photographic Survey accurately records 8,589 locations of the tree with ±30cm margin of error from the base location to the recorded location. Further, each image records the location described with up to ±3m margin of error along the ground. Finally, we offer that volumetrically, the variance described could be described in this way. On the lower edge of the margin of error, ±30cm, the effort could miss the size of a very small stem in terms of volume. On the larger extent, ±3m, accuracy could be described in terms of 2 larger mature branches per location point. Having said that, those who follow the plot map and our methods will have an advantage we did not have as this is a baseline study. Future researchers will have both the planned and recorded location data while we, simply had the base map. In short,  meaning, those who replicate the effort on any scale, will have the averages to their advantage. Any effort to replicate a location or the entire record, will have 3 data points while our work started with one.

As a first of its kind document of Pando we hope will serve future research, we offer these notes on efforts we undertook to ensure quality, data integrity and reduce the influence of human error.

  1. Each route sheet includes instructions on how to prepare, document and capture the given route along with room for notes where there was any variation between tested method and variation due to field conditions.
  2. Crews were instructed to skip and note any location where the space between planned and actual location exceeded 2m.
  3. SD Cards were individually labeled. Labels were noted on  field sheets at time capture of a given route was undertaken.
  4. Each route begins and ends with a slate image taken in 360-degrees with each crew member holding up a placard or notebook with the route name, time and heading information.
  5. SD cards were only used once per day and locked and stored away until quality assurance review.
  6. Each SD Card of field data was individually reviewed for veracity and quality and compared against the route sheet it claimed to be associated with at the end of each shoot day by Lead Photographer.
  7. In post-production, images were compared to the field sheet to ensure they matched the associated route in number of shots, sequence and location
  8. In post-production, images were evaluated individually to isolate unseen camera errors
  9. Duplicate images, or locations that were noted as “obstructed” were documented before sequencing and stitching
  10. After sequencing, we reviewed sequences visually to ensure image provided a logical flow
  11. In the case that a recorded location data was lost or corrupted, we defaulted to the planned location coordinates and offer a margin of error of ±3m in 2D space. A wide berth in 2D space and volumetrically.

In all, while Friends of Pando is not a scientific institution, we offer the findings with the hope that what we have worked to develop, can serve as a credible resource for remote research and a valuable reference. Hope that the information we provide, would allow researchers to use data and that our work provides meaningful consideration for use in peer review models of research as well as emergent models of inquiry based on open science and citizen science models.

Limitations

Translating Accuracy and Errors Between 2D Space and 360° Images

The use of a shoot map in created in 2D, to capture images in 360-degrees can pose challenges for quantifying errors. In 2D, ±30 cm (0.98ft) could compound along horizontal plane, while in 360-degree spaces, an area so small hardly registers volumetrically. Thus, one limitation of the record is that, where margin of error can impact total coverage area, how error translates volumetrically is not easy to transliterate. In this way, one could argue that we should make a choice if we are to declare margin of error; in 2 degrees or, 360 degrees, which today, is simply not possible by any measures we have explored.

Missing Locations

In cases where a location could not be recorded as it was an unsafe to enter, or, a flag and camera unit could not be placed within 2m from the planned location, the shot was marked as “Missing” or, “Obstructed”. While missing data on a small scale could affect outcomes significantly, in this case, the scale of the effort offers compensation which may need adjustment on larger scales. Generally speaking, every point on the grid is adjacent to at least 3 and up to 8 additional location points with overlap. Thus, a limit of the record is that not every location was recordable. Here, others may need to make another attempt at capture, or simply work around the location when creating virtual plots.

Variance in Ground Cover and Obstructions

While the effort was developed and undertaken with the general concept that every image could document a 49m (22.97ft) area in 2D space, that is not always the case. After all, Pando is a tree whose 47,000 branches vary in size and  along with other natural features of the landscape and can obstruct part or most of the view of a given location. While an image filled with one branch is still an accurate record of the tree, another arguably limitation, may be that one will not always be able to move through the physical plane, or observe all locations in equal measure. Despite this, we believe where views are obstructed, it provides an analog to how one would experience the tree in field research conditions.

Virtual Plots and the Problems of Remote Work and Transliterating the Physical and Virtual

Work began on the Pando Photographic Survey in the Summer of 2019. Work continued through the COVID Pandemic when western cultures saw renewed interests in remote work. While most work in natural sciences and forestry happens in the field, we have received questions and concerns about how the record could be used, physically removed from the subject. While there are many tools and applications available for VR headsets, and other ways still, to translate equirectangular images using applied mathematics or advanced imaging models, it is worthwhile to address that our methods and models are novel compared to physical field work and for that matter, traditional re-photography. In short, another limitation could be that the 7m grid and, the the use of 360 degree imagery complicate applications using more familiar approaches. We welcome researchers to feedback for ways to improve on our models for remote research, help us update data packages and usage models for use in  traditional or, emergent research methods.

Images in Time

Another limit worthy of discussion is how the image was developed and the value of the provided record over time. The lead photographer has referred to this as the “Dynamic Pando” problem or, the “David Bowie Problem” of trees; how we think of them. Pando is not static, the tree has invented and re-invented itself from the ground-up many times over the course of its estimated 9,000-year lifespan. Generally, a new stem (what becomes a branch, and appears to many as “trunks”) can grow up to 0.91m a year and is generally considered a good candidate to maturity once it reaches about 2.5m, or approximately three years of growth. Thus, a limit of the record is that the “shelf life” of the effort for hard count regeneration studies spans from September of 2022 through September of 2025. Through the course the effort, we have seen advances in imaging and image processing technologies from a picture of a Black Hole, to advances in AI that can isolate visual patterns mathematically. It is our hope that by making the document freely available, it will open Pando to study using methods we could not anticipate and could not imagine. Thus, it could be said, some of the value of the record, may only come from creative uses of existing technology or, emergent methods and approaches the record might not be able to fully consider as it is a record of a given time.

The Virtual and the Real

We have fielded many questions over the course of the project about whether or not, documenting Pando using 360-degree images is the same as we, trying to virtualize the tree – that is, aiming to provide an analogous that can replace what the tree means to us or, can mean for others. The answer is simply no, but it merits acknowledgement and humility as we believe, it is a bit of good skepticism about art and science both. The land Pando calls home, has been used by humans for at least 1,500 years. Well before the the initial observation of the tree in 1976. Humans have, and continue to shape the land Pando calls home and its future is inarguably in human hands. Thus, in the record, you will find, we have purposefully left camera stands in the shots, after all, they could serve as a guide for human scales of space. You will find crew members lurking on the edge of the frame which came to be known as “Bigfooting” amongst the team. You will find two people who just met, making silly faces in a slate shot, their shirts stained with the salt from the sweat of lugging 40 pounds uphill all day as Pando, and all trees, can be a source of joy and comfort. What’s more, in this record, occasionally, you will see crew members front and center, but maintained; not only because “some” data is better than “no” data, but as we maintained that as part of nature, it is the human effort and imagination that make natural wonders “real” and worth knowing. Be that approximate, be that upon inspection. Be that in detail. Be that real. Be that memory, but no less real as a hope for memories and future generations.

Conclusions

Building on the history of photography in conservation and land management, 360-degree imagery shows great promise for documenting large-scale forested landscapes like Pando. 360-degree photography can produce high-quality image data from the forest floor well into the canopy in challenging physical landscapes. 360-degree photography is more efficient than 2D processes in the field, while processing 360-degree imagery involves less work that can be automated using freely available software. As is applicable to research, 360-degree photography offers opportunities for immersive observational experiences year round with the subject at the same time it is less expensive in terms of the time and the expense of field work alone. As our subject was the world’s largest tree, the methods and show enormous potential for use in land management, ecology and forestry on much smaller scales, and could arguably be adapted to larger scales. While many questions remain about the potential value and uses of 360-degree photography, we believe this approach shows enormous potential for use in creating inventories or, to document other landscapes working on variety of scales.

Disclosures, Citation and Usage

This project and these findings were developed by Friends of Pando in collaboration with our partners Fishlake National Forest of the U.S. Forest Service, Department of Agriculture and, Snow College, Richfield. It was made possible with additional support from sponsor Lensrentals.com along with private donations from fellow Pando lovers like you. Friends of Pando is a registered nonprofit in the State of Utah and is recognized by the IRS as a 501(c)3 organization focused on the forestry and beautification. This project and its results were developed with the intent to educate the public, support research and preservation efforts and inspire stewardship of the Pando Tree.

Friends of Pando began publishing the results of the Pando Photographic Survey on March 28th, 2023 on its website,  www.friendsofpando.org. Friends of Pando makes all data, images, models, maps and methods freely available for study, replication and enjoyment for educational and noncommercial purposes on an as is basis. We will continue to work to publish the record through September of 2023 and offer additional educational experiences using throughout 2023 and, beyond.

References

  • October 2020, Tyler Green, “Carelton Watkins: Making the West American”, University of California Press
  • 2008, DeWoody, Jennifer; Rowe, Carol A.; Hipkins, Valerie D.; and Mock, Karen E., “Pando Lives: Molecular Genetic Evidence of a Giant Aspen Clone in Central Utah”. Aspen Bibliography. Paper 3164. https://digitalcommons.usu.edu/aspen_bib/3164
  • 2004, Oliver Grau, Virtual Art, From Illusion to Immersion. The MIT Press

Team

Leadership

  • Lance Oditt, Lead Photographer and Executive Director of Friends of Pando
  • Ryan Thalman, Chief Scientists and Professor at Snow College, Richfield

Field Crew, Production Team, Volunteers (2019 - 2023)

  • April Anderton
  • Dave Anderton
  • JD Anderton
  • Kaci Anderton
  • Kaylee Carlson
  • Janis Connell
  • Michael Dalton
  • Jason Dilworth
  • Corban Gibson
  • Brady Gordon
  • Heidi Johnson
  • Tonia Lewis
  • Lena Lindsay
  • Nathan Lindsay
  • Lindy Madden
  • Robert Press III
  • Andrew Russon
  • Ngawang Salaka
  • Kyden Saner
  • John Shattuck
  • Hope Smith
  • Tiesha Smith
  • Stuart Smith
  • Wilson Thorpe
  • Carson Utley
  • Jenny Rioja
  • Jakob Watson
  • Cade Wolcott

Lance Oditt Would Like to Offer His Gratitude and Thanks To:

  • Rich and Lisa at Lensrentals
  • The City of Richfield, Utah
  • The Staff and Volunteers at the Richfield Visitor Center
  • Staff and Leadership at Snow College, Richfield
  • Billy Kanaly of Denver Botanic Gardens
  • Liz Davy
  • Paul Rogers
  • Simone Friedman and Manny Friedman of EJF Philanthropies
  • Anthony Fredericks
  • Stephanie Moulton
  • Don Gomes of Entrada Institute
  • Melindiana Jones

Friends of Pando is dedicated and working to educate the public, support research and preservation efforts and inspire stewardship of Pando, the world’s largest tree.

 

Friends of Pando is a proud partner of Pando’s public land stewards, Fishlake National Forest of the U.S. Forest Service, Department of Agriculture. Learn more about our partnership.

 

Friends of Pando and its partners are equal opportunity employers.

fopemblemfooterwidth

Just $14 a month supports work to ensure Pando can be enjoyed for generations to come. Make a one-time or, recurring tax deductible donation today.

Friends of Pando
PO Box 12
Richfield, UT, 84701
Phone: 435-633-1893
IRS EIN: 87-3958681