IEEE Geoscience and Remote Sensing Magazine - September 2021 - 113

owing to the data's relatively fine spatial resolution and
proper accuracy [6]. Various operational LST products can be
obtained using effective TIR-based retrieval algorithms [6].
For example, LST products are available free of charge from
the FengYun-2 (FY-2) Visible Infrared Spin Scan Radiometer
(VISSR) [8], FY-3 Visible Infrared Scanning Radiometer
(VIRR), Moderate-Resolution Imaging Spectroradiometer
(MODIS) [7], and Meteosat Second-Generation (MSG) Spinning
Enhanced Visible and InfraRed Imager (SEVIRI) [9].
However, the spatial continuity of LST products is strongly
affected by pixels with invalid values caused by clouds or
cloud shadows, hereafter referred to as cloudy pixels. On average,
clear-sky conditions account for only about one-third of
the daily weather worldwide, with lower percentages occurring
in humid regions [10]. In particular, about two thirds
of MODIS LST data are contaminated by clouds [11]. Furthermore,
some cloud-free but naturally bright pixels are
frequently misclassified as cloud coverage, and their corresponding
LSTs are set to the missing pixel value [12]. In addition,
missing information caused by defective TIR sensors,
such as the Landsat Enhanced Thematic Mapper Plus (ETM+)
scan line corrector (SLC)-off data [13], also introduces spatial
discontinuity into LST products. Thus, spatial discontinuity is
a common limitation in LST obtained from TIR data owing
to cloud coverage, incorrect algorithm data, defective sensors,
and other problems.
PMW radiation, which can penetrate nonprecipitating
clouds, is minimally affected by atmospheric absorption
and water vapor [14], [15]; therefore, the LST retrieved using
this method has high spatial completeness. However,
the limitations of PMW measurement lie in the existence
of a wider scanning gap between orbits and lower spatial
resolution compared with those of TIR measurement. The
spatial resolution of PMW measurement, at 25 km for
the Advanced Microwave Scanning Radiometer (AMSR)
for the Earth Observing System (AMSR-E) and about 10 km
for AMSR2, is much lower than that of TIR measurement.
Wide gaps often exist between orbits over middle and
low latitudes [16], such as the 25% gap in the AMSR-E
image over the Chinese mainland. Therefore, LST with
low spatial resolution retrieved from PMW measurements
often includes spatial discontinuity, particularly
on a global scale.
Another limitation is the tradeoff between the temporal
and spatial resolutions of satellite-derived LST for a single
sensor, whether TIR or PMW, which often makes it impossible
to obtain LST with the high spatiotemporal resolution
required by some applications [17]. Generally, LST retrieved
from sensors with fine spatial resolution has poor temporal
resolution, resulting in temporal discontinuity and vice versa.
These two core limitations of satellite-derived LST products
are summarized in Table 1, and the corresponding sample
charts are shown in Figure 1(a).
The aforementioned issues severely hinder applications of
LST products in many fields [18]. For example, surface urban
heat islands (SUHIs) can be continuously observed with
SEPTEMBER 2021 IEEE GEOSCIENCE AND REMOTE SENSING MAGAZINE
cloud-free FY-2/3 or MODIS data; however, their associated
spatial resolutions are too coarse for revealing detailed SUHI
spatial patterns. Images with greater spatial detail are provided
by Landsat and the Advanced Spaceborne Thermal Emission
and Reflection Radiometer (ASTER). However, the long
revisit cycles of these two products, at more than 15 days, and
frequent cloud contamination cause significant differences
in the LST scenes these sensors acquire in space-time observation
[13]. Therefore, practical applications urgently require
LST products with spatially continuous high-spatiotemporalresolution
(SCHR)-LST, as shown in Figure 1(b).
Owing to the growing number of available satellite-based
LST products, many different approaches for generating
SCHR-LST have been proposed, resulting in abundant research
on these methods. Therefore, it is of great importance
to present an overview of the state-of-the art technology in
applied SCHR-LST techniques.
Although previous research of Zhan et al. includes a review
on the disaggregation of LST to finer resolutions [19],
it is presented from the perspective of thermal sharpening
and temperature unmixing. However, several recently proposed
methods (e.g., spatiotemporal fusion) for obtaining
LST with finer temporal and spatial resolutions should also
be surveyed. Similar to that in disaggregation, the purpose of
spatiotemporal fusion is also to improve the spatiotemporal
resolution of LST data using multiple sensors and multitemporal
LST data without the need for auxiliary data. Spatiotemporal
LST fusion has developed rapidly to become a research
hotspot.
Therefore, a review of recent research conducted on this
topic is needed. Furthermore, to the best of our knowledge,
a thorough review of methods for deriving spatially continuous
LST has not been performed thus far. The objectives of
this article are to review methods for generating SCHR-LST
products, describe the state-of-the-art technology, and identify
the most promising research fields to obtain results
that will ultimately benefit LST producers and developers of
SCHR-LST algorithms.
BACKGROUND
THEORETICAL CONCEPTS
LST is affected by incoming solar radiation, cooling by wind,
coastal effects, land cover, temperature inversion, and other
effects [20], making the data highly variable in space and
TABLE 1. THE CORE LIMITATIONS OF SATELLITE-DERIVED
LST PRODUCTS.
ISSUE
Spatial discontinuity type
TIR
PWM
Cloudy pixels, pixels
misclassified as clouds,
and defective sensors
Scanning gaps between
orbits
Tradeoff between temporal
and spatial resolution
Technical and budgetary
limitations
Technical and budgetary
limitations
113
IEEE Geoscience and Remote Sensing Magazine - September 2021

Table of Contents for the Digital Edition of IEEE Geoscience and Remote Sensing Magazine - September 2021
