1) Repeatable Science with Drones

Drones are revolutionizing the way natural scientists measure their study systems.  We are researching how measurements from small remote sensing drones, aka uncrewed aerial systems (UAS), can supplement existing data to answer environmental questions in new ways. UAS collect repeatable and reliable remote sensing measurements while saving time and money for researchers. Earth Lab is harnessing the versatility of UAS sensors to measure spectral and structural information across landscapes.

UAS extend field-collected data continuously across a landscape. By collecting images with spatial resolution as small as centimeters, UAS can capture fine-scale characteristics and phenomena captured on the ground. Such data can be scaled up to airborne or satellite imagery that are meters or tens of meters in resolution. Earth Lab is exploring what forest traits can be derived from UAS-based measurements and is working to build tutorials to make scaling UAS data to field-collected data an accessible workflow for all.

Earth Lab is committed to building a community of researchers who use UAS in their work. We aim to lower the barrier of entry to using UAS in research by sharing technical knowledge and best practices for UAS methods.

Similarly, there are issues with bridging differences of scale in remote sensing. The Earth is observed by many types of imaging technology with different spatial resolutions. From centimeter-scale imagery collected from UAVs, to airborne hyperspectral imagery at the meter-scale, to the 10’s of meter scale from satellite multispectral imaging systems, the diversity of data representing the Earth’s surface at different scales enables us to ask questions from the hyperlocal to continental and global scale. We combine these data to better understand the questions we can ask with the different remote sensing data sources, as well as processes and change occurring on the Earth in a variety of domains such as forest systems, the urban and built environment, and geomorphic processes.

 

2) Drones and CubeSats for Science

Moderate resolution satellite imagery can provide coarse-scale information on changes in the natural environment (e.g., percent forest cover) but cannot resolve finer scale changes (e.g., vegetation state changes from conifer to deciduous forest). The high resolution of CubeSats such as Planet’s Dove imaging technologies can capture environment responses at finer scales. At the individual tree-level, small unmanned aerial system (sUAS), or drone, technology enables rapid acquisition of scientific quality remote sensing data. Miniaturized multispectral cameras that retrieve surface reflectance products, and small form factor RGB cameras are now ubiquitous. By equipping such systems on drones, we can observe the natural environment at a much finer spatial and temporal scale than possible with moderate resolution satellite imagery. 

The difference in spatial scale from Landsat, Planet, and an example dataset acquired with a drone at Earth Lab are shown here.

(Left to Right) Color-Infrared representation of Landsat 8 (30m), PlanetScope (3m), and MAPIR Kernel imagery (0.03m, with PlanetScope in background) over a forested area near Gold Hill, Colorado.

Earth Lab uses drones to increase the temporal frequency and spatial resolution of ecosystem observations. Examples include...

  • We are upscaling multispectral UAS to unlock 40+ years of Landsat imagery to understand forest recovery after disturbance [BLOG]. Earth Lab deploys a sensor with spectral bands matching that of Landsat and integrated the sensor onto a multicopter drone [BLOG]. We are building hierarchical Bayesian models with these data to better understand which factors contribute to forest recovery following disturbance at a regional-scale [BLOG].

  • Earth Lab is quantifying quality and propagating uncertainty across scales by processing and comparing hyperspectral imagery from a joint flight between a UAS and an airplane, each bearing a spectrometer [BLOG]. 

  • By integrating surface elevation models derived from airborne lidar, drone-based Structure-from-Motion, and ground-based surveys, we are developing new methods for relating surface properties (e.g., the bedrock-soil interface) with topographic properties (e.g., slope, curvature, roughness). This work at Earth Lab directly bears on the challenge of constraining spatially heterogeneous hydrologic and geomorphic fluxes at the process scale in service to watershed-scale analyses [BLOG].

Earth Lab is working on methods to incorporate fine-scale drone data with large-scale satellite observations. Using Planet and Landsat surface reflectance data, we are linking vegetation properties and structure across scales through novel integration of radiometric and thematic data.