Scale-dependency of Arctic ecosystem properties revealed by UAV
In the face of climate change, it is important to estimate changes in key ecosystem properties such as plant biomass and gross primary productivity (GPP). Ground truth estimates and especially experiments are performed at small spatial scales (0.01-1 m(2)) and scaled up using coarse scale satellite...
Published in: | Environmental Research Letters |
---|---|
Main Authors: | , |
Format: | Article in Journal/Newspaper |
Language: | English |
Published: |
Umeå universitet, Institutionen för ekologi, miljö och geovetenskap
2020
|
Subjects: | |
Online Access: | http://urn.kb.se/resolve?urn=urn:nbn:se:umu:diva-175081 https://doi.org/10.1088/1748-9326/aba20b |
Summary: | In the face of climate change, it is important to estimate changes in key ecosystem properties such as plant biomass and gross primary productivity (GPP). Ground truth estimates and especially experiments are performed at small spatial scales (0.01-1 m(2)) and scaled up using coarse scale satellite remote sensing products. This will lead to a scaling bias for non-linearly related properties in heterogeneous environments when the relationships are not developed at the same spatial scale as the remote sensing products. We show that unmanned aerial vehicles (UAVs) can reliably measure normalized difference vegetation index (NDVI) at centimeter resolution even in highly heterogeneous Arctic tundra terrain. This reveals that this scaling bias increases most at very fine resolution, but UAVs can overcome this by generating remote sensing products at the same scales as ecological changes occur. Using ground truth data generated at 0.0625 m(2)and 1 m(2)with Landsat 30 m scale satellite imagery the resulting underestimation is large (8.9%-17.0% for biomass and 5.0%-9.7% for GPP(600)) and of a magnitude comparable to the expected effects of decades of climate change. Methods to correct this upscaling bias exist but rely on sub-pixel information. Our data shows that this scale-dependency will vary strongly between areas and across seasons, making it hard to derive generalized functions compensating for it. This is particularly relevant to Arctic greening with a predominantly heterogeneous land cover, strong seasonality and much experimental research at sub-meter scale, but also applies to other heterogeneous landscapes. These results demonstrate the value of UAVs for satellite validation. UAVs can bridge between plot scale used in ecological field investigations and coarse scale in satellite monitoring relevant for Earth System Models. Since future climate changes are expected to alter landscape heterogeneity, seasonally updated UAV imagery will be an essential tool to correctly predict landscape-scale changes in ecosystem ... |
---|