HiLDEN working group: Arctic drone synthesis to bridge the gap from plants to pixels

From the 24th to the 28th of March, 13 researchers met in New Hampshire to advance Arctic tundra research.  Attendees (plus three remote participants) tackled one of the largest drone-imagery datasets ever compiled for high latitudes – all contributed from field teams using similar protocols – to commence the High Latitude Drone Ecology Network (HiLDEN, https://arcticdrones.org/) data synthesis.


Drone data are allowing scientists including members of the HiLDEN team to capture a landscape-level perspective on the rapidly changing Arctic.

The Arctic is warming rapidly and satellites indicate a greening of the these high latitude ecosystems, yet available datasets do not capture the spatial detail required to understand how warming (or other environmental changes) are actually causing this landscape greening. The longest running satellite datasets either use grainy pixels the size of Manhattan, or take snapshots so infrequently that clouds block out a substantial portion of each growing season.  In contrast, Arctic field researchers have extremely deep understanding of what happens in garden-sized vegetation plots, but it can be difficult to link these to regional patterns. The HiLDEN network is a big step towards filling this gap. It provides a new opportunity to explore tundra vegetation change from a fresh perspective, and to better link satellite and on-the-ground observations by capturing high resolution data of tundra landscapes with drones.


Drone data can help us test the question of whether climate warming and resulting vegetation change on the ground are the major driver of Arctic greening signals observed by satellites.

Climate is changing rapidly at the extreme latitudes of our planet. Warming temperatures, melting sea ice, and thawing permafrost are transforming Arctic environments. In some places, bare ground is becoming vegetated.  Elsewhere, plants are now growing larger. These simple changes – if widespread – can alter the trajectory of global patterns of climate change. Up to a third of the world's soil carbon is stored in frozen soils, and if released would accelerate climate warming for the entire planet. However, scientists remain uncertain about how and where these feedbacks might unfold in the tundra. A critical step towards resolving this uncertainty is to understand the patterns and causes of changes in tundra vegetation – and this is where HiLDEN data synthesis comes in.


Drone data provide local information that can be translated to regional patterns to improve our understanding of tundra biome-wide greening trends.

The group assembled at Pierce's Inn in the wooded hills of New Hampshire and at the Neukom Institute for Computational Science on the Dartmouth College campus. Fireside discussions led in to data synthesis powered by high-powered computing, both on the cloud and at Dartmouth College.  Using the Google Earth Engine (https://earthengine.google.com/), we explored patterns of satellite data in conjunction with drone records at 42 sites around the Arctic.  Throughout the working group, we identified our major research questions, outlined a manuscript, and discussed future avenues for the network.


The HiLDEN data synthesis team meeting to discuss how drone data can help us to better understand Arctic greening trends.

Our dataset included over two hundred thousand aerial images from across the Arctic including tundra landscapes in Europe, Siberia, Alaska, Canada and Greenland. What immediately stands out from looking at these pictures is the stunning diversity of tundra landforms – from wetlands to highlands, shrubs and trees to barren grounds, from treeline to the northern reaches of tundra biome.  All of these patterns are lost in poor resolution satellite data. Beyond traditional color photographs, the group analyzed imagery taken from discrete regions of the radiometric spectrum to generate specialized vegetation indices (like the Normalized Difference Vegetation Index – NDVI) while also creating three dimensional surface models of these landscapes.


Drone images from across our 42 field sites indicating the tremendous uniqueness of ecosystems around the Arctic.

Based on these data products, our preliminary analyses resulting from the working group indicate the following: 1) The HiLDEN dataset encompasses tundra sites that span the diversity of biome scale greening trends. 2) Various satellite observed greening trends correspond with one another across our sites, but the years of high and low greening do not always track well across datasets. 3) Local-scale variation and features of the land-surface (captured in drone data) seem to have an influence on various characteristics of greening trends observed by satellites.


Drone data overlaid on satellite imagery collected as a part of the 2017 HiLDEN field campaign in Siberia (Kumpula, Kerby et al) illustrating how drone data can provide a higher-resolution perspective on Arctic greening patterns.

Over the coming months, analyses from this first HiLDEN working group will be refined and shaped into a data synthesis manuscript.  This working group has demonstrated that collaborative science with common data collection protocols and a common purpose can yield exciting new perspectives on the ecological responses to climate warming at the cold extremes of our planet.


A drone image collected as a part of the HiLDEN 2018 field campaign on the Yukon Arctic Coast of Canada of permafrost thaw cutting into a tundra landscape.

We thank the Neukom Institute for Computational Science (https://neukom.dartmouth.edu/) for supporting this workshop and HiLDEN data servers since the network was founded.  We also thank the Scottish Alliance for Geoscience, Environment, and Society (SAGES) and the Institute of Arctic Studies at Dartmouth for their critical support in the preparations of the dataset and in planning the working group. Finally, we thank all of the HiLDEN members for contributing data and facilitating what we think might be the largest ecological drone data synthesis to date!

Summarized on behalf of the working group by Jeff Kerby, Jakob Assmann, and Isla Myers-Smith