IEEE Geoscience and Remote Sensing Magazine - June 2016 - 78

TABLE 1. AN OVERVIEW OF VARIOUS GROUND-BASED WSIs AND THEIR INTENDED AppLIcATIONS.
AppLIcATION

ORGANIzATION

cOUNTRY

WSI MODEL

air traffic control [18]

campbell scientific Ltd.

United Kingdom

ir nec ts9230

cloud attenuation [21]-[23]

nanyang technological University singapore

singapore

WaHrsis

cloud characterization [13]

atmospheric Physics Group

spain

GFat all-sky imager

cloud classification [12]

Brazilian institute for space research

Brazil

tsi-440

cloud classification [14]

Laboratory of atmospheric Physics

Greece

canon iXUs ii with FoV 180˚

cloud macrophysical properties [9]

Pacific northwest national Laboratory

United states

Hemispheric sky imager

cloud-track wind data monitoring [15]

Laboratoire de météorologie dynamique

France

nikon d100 with FoV 63˚

convection [16]

creighton University

United states

digital camera

radiation balance [17]

Lindenberg meteorological observatory

Germany

Vis/nir 7

solar-power forecasting [11]

solar resource assessment & Forecasting Laboratory

United states

tsi-440

Weather monitoring [24]

Pacific northwest national Laboratory

United states

tsi-880

Weather reporting [19]

ecole Polytechnique Fédérale de Lausanne

switzerland

Panorama camera

SatelliteS aS a Starting Point
Satellite images are commonly used to monitor the earth and
analyze its various properties. They provide remote sensing
analysts with accurate information about the various earth
events. Satellite images are available in different spatial and
temporal resolutions and also across various ranges of the
electromagnetic spectrum, including visible, near- and farinfrared regions. For example, multitemporal satellite images
are extensively used for monitoring forest canopy changes [1]
or evaluating sea-ice concentrations [2]. The presence
of clouds plays a very important role in the analysis of
satellite images. NASA's Ice, Cloud, and Land Elevation
Satellite (ICESat) has demonstrated that 70% of the earth's
atmosphere is covered with clouds [3]. Therefore, there has
been renewed interest among the remote sensing community to further study clouds and their effects on the earth.
Satellite images are a good starting point for monitoring the earth's atmosphere. However, they have either
high temporal resolution (e.g., geostationary satellites)
or high spatial resolution (e.g., low-orbit satellites) but
never both. In many applications, such as solar energy
production [4], local weather prediction, tracking contrails at high altitudes [5], studying aerosol properties [6],
and the attenuation of communication signals [7], [8],
data with high spatial and temporal resolution is needed. This is why ground-based sky imagers have become
popular and are now widely used in these and other applications. The ready availability of high-resolution cameras at a low cost facilitated the development of various
models of sky imagers.
A WSI consists of an imaging system placed inside a
weather-proof enclosure that captures the sky at user-defined intervals. A number of WSI models have been developed over the years, including a commercial WSI [Total
Sky Imager (TSI)-440, TSI-880] manufactured by Yankee
Environmental Systems that is used by many researchers
[9]-[11]. Owing to the high cost and limited flexibility of
80

commercial sky imagers, many research groups have built
their own WSI models [12]-[19], e.g., the Scripps Institution of Oceanography at the University of California, San
Diego, has been developing and using WSIs as part of its
work for many years [20]. Similarly, our group designed the
Wide-Angle High-Resolution Sky Imaging System (WAHRSIS) for cloud-monitoring purposes [21]-[23]. Table 1 provides an overview of the types of ground-based sky cameras
used by various organizations around the world and their
primary applications.
Machine learning for reMote SenSing Data
The rapid increase in computing power has enabled the
use of powerful machine-learning algorithms on large
data sets. Remote sensing data fill this description and are
typically available in different temporal, spatial, and spectral resolutions. For aerial surveillance and other monitoring purposes, Red-Green-Blue (RGB) images are captured by low-flying aircraft or drones. Multispectral data
are used for forest, land, and sea monitoring. Recently,
hyperspectral imaging systems with very narrow bands
have been employed for identifying specific spectral signatures for agriculture and surveillance applications.
In cloud analysis, one example of such remote sensing data is ground-based images captured by WSIs. With
these images, one can monitor the cloud movement and
predict the clouds' future location, detect and track contrails, and monitor aerosols. This is important in applications such as cloud attenuation and solar radiation modeling, which require high temporal and spatial resolution
data. The requirement for high-resolution data is further
exemplified by areas where weather conditions are more
localized. Such microclimates are mainly prevalent near
bodies of water that may cool the local atmosphere or
in heavily urban areas where buildings and roads absorb the sun's energy (Singapore, the authors' home, being a prime example of such conditions). These weather
ieee Geoscience and remote sensing magazine

june 2016

Table of Contents for the Digital Edition of IEEE Geoscience and Remote Sensing Magazine - June 2016