IEEE Geoscience and Remote Sensing Magazine - March 2023 - 105

wilderness) to better protect the environment we live in.
Ultimately, we raise the rhetorical question " What makes
nature wild? " and invite the community to participate
in finding possible solutions to the open challenges we
have mentioned to get closer to discovering the concept
of wilderness.
THE MAPINWILD DATASET
CURATION
The cloud storage and computing facilities of Google Earth
Engine (GEE) [32] are utilized during the dataset curation
phase by keeping the reproducibility aspect in mind (i.e.,
defining fixed seed values to random number generators
and using definitive time frames for image queries). Given
the complexity of the task being addressed, MapInWild
consists of distinct sets of geosensors that aim to leverage
complementary information about the natural environment:
Sentinel-1, Sentinel-2, European Space Agency (ESA)
WorldCover, the Visible Infrared Imaging Radiometer Suite
(VIIRS), and World Database on Protected Areas (WDPA)
polygons; a sample from MapInWild is shown in Figure 1.
WDPA polygons comprise feature classes designating protected
areas and formed by the United Nations Environment
Programme's World Conservation Monitoring Center
and the International Union for Conservation of Nature
(IUCN) World Commission on Protected Areas [28]. The
complete list of WDPA classes is given in Table 1. From the
anthropogenic pressure point of view, in which the variations
in the diversity are explained along human influence
gradients, we reformulate these discrete sets of areas on a
continuous spectrum, where
the distinction among areas
is relatively ill defined. On
the grounds of the observation
that human presence
increases going from category
Ia to category II, we raise the
question: At what point of the
human presence continuum
should an area be considered
a wilderness area? Supported
by this theoretical research
question, we argue that category
Ia and category II could be beneficial proxies in interpreting
and explaining the concept of wilderness. Ultimately,
we form the MapInWild dataset with the following three
WDPA classes: strict nature reserve (category Ia), wilderness
area (category Ib), and national park (category II).
As for the Sentinel-1 image, a ground-range-detected
WE ADOPT A BOTTOM-UP
APPROACH AND ADDRESS
THE AMBIGUITY IN
NATURE'S VITAL ELEMENTS
(E.G., WILDERNESS) TO
BETTER PROTECT THE
ENVIRONMENT WE LIVE IN.
product with a resolution of 10 m is used. Sentinel-1 images
contain VH and VV polarizations and use the interferometric
Wilderness Proxy
Background
Tree Cover
Shrubland
Grassland
Cropland
Built Up
Bare/Sparse Vegetation
Snow and Ice
Permanent Water Bodies
Herbaceous Wetland
Mangroves
Moss and Lichen
FIGURE 1. A wilderness sample from the MapInWild dataset (Juniper Dunes, United States, WDPA ID: 555556115). The first row shows the
Sentinel-2 images of four seasons, from left to right: spring, summer, autumn, and winter. Second row, same order: Sentinel-1 image, ESA
WorldCover map, VIIRS Nighttime Day/Night band, and WDPA annotation. For visualization purposes, the VV and VH bands of the Sentinel-1
image are treated as red and green bands, and only visible bands (red: B4, green: B3, and blue: B2) in the Sentinel-2 image are used.
The ESA WorldCover map legend is given at the bottom the figure.
MARCH 2023 IEEE GEOSCIENCE AND REMOTE SENSING MAGAZINE
105

IEEE Geoscience and Remote Sensing Magazine - March 2023

Table of Contents for the Digital Edition of IEEE Geoscience and Remote Sensing Magazine - March 2023