IEEE Geoscience and Remote Sensing Magazine - September 2022 - 15

well-known drawbacks. The spatial resolution of the imagery
is very coarse, which limits the images' use to large
lakes and floodplains as opposed to small water bodies,
such as rivers and wetlands [52], [75], [77], [87], [89]-[91].
Additionally, quantitative estimates of the water subpixel
percentage are difficult, particularly when passive microwave
signals are contaminated by vegetation [74], [92],
which can seriously distort the estimation of water surface
area [30].
As reviewed in the preceding, it is difficult to select a
unique sensor and methodology to obtain consistent surface
water measurements, regardless of the environment
[76]. Addressing the complementary strengths of different
types of sensors, the methodology of combining multisatellite
observations has been especially recognized in recent
years [12], [30], [76], [92]-[94], contributing remarkably to
surface water delineation across all-time, all-weather, and
high-spatiotemporal-resolution series. Through the years,
various global, continental, and regional surface water
area products have also been developed, showing individual
characteristics and providing valuable data sources
for different hydrological applications [30], [82], [95]-[97]
(see Table 2).
SATELLITE-DERIVED WATER LEVEL
So far, several accepted satellite-based methods for retrieving
water levels have been proposed [16], [98]. Depending
on different water bodies, these methods can be categorized
into three types: the area-elevation relationship,
a land-water contact approach based on digital elevation
model (DEM) data, and satellite altimetry.
The area-elevation relationship method involves a
number of surface areas to derive an empirical equation
that statistically relates area to elevation and then uses the
relationship along with area measurements to estimate
water levels [24]. Since the surface area is relatively easy to
measure directly from satellite observation with high accuracy,
this technique has been more commonly applied in
poorly gauged reservoirs and lakes. For example, Tan et al.
[99] used multisource satellite images combined with hydrological
data to measure water level changes in Poyang
Lake. For rivers, many researchers successfully estimated
water levels based on the relationship among water level
measurements and satellite-derived river width and discharge
instead of water area [100], [101]. This method looks
simple and efficient, while the problem associated with it
is that we often do not have enough observed data to construct
area-elevation curves [24]. Furthermore, the derived
area-elevation relationship is essentially empirical: it is
useful for individual water bodies, and the transferability
from one hydrogeomorphological setting to another has
not been shown [102].
The land-water contact method is suitable for the determination
of water levels in lakes and reservoirs, relying on
the combination of surface water extent and bottom topographic
data [24], [103]. The principle of this method is as
SEPTEMBER 2022 IEEE GEOSCIENCE AND REMOTE SENSING MAGAZINE
follows. First, water-land borders derived from satellites
are transformed into vertices. Height values along those
vertices are extracted from the matched bottom topographic
data. Then, the spatial distribution of the water level
can be derived through the interpolation method from
the height values of the water-land borders [104]. Several
studies have been carried out to estimate water level variations
by using this approach, and the results showed that it
could capture spatial variations in water levels and achieve
good performance [102], [103], [105]. However, this method
also poses some disadvantages. Most importantly, the
absolute accuracy depends too much on DEM uncertainties
and errors in both the horizontal and vertical directions
[106]. Small deviations can lead to large errors in the
computation of variations in volume [107], [108]. Despite
increasingly accurate geopositioning techniques, the approach
is still limited to shallow and favorable water conditions
[109].
Satellite altimetry is a new and promising technique
for directly detecting the elevations of surface water from
space [50], [110]. It is based on measurement of the two-way
travel time of a radar (or laser) signal sent by a satellite and
reflected by a water surface [26]. Several altimeter satellites
have been launched since the early 1990s [17], [111]. Table 3
provides an overview of the main satellite altimeter instruments.
These sensors offer a wide variety of altimeter data
with different characteristics in terms of spatial and temporal
coverage, and the data are freely available worldwide.
Thus far, the altimetry technique has been widely used to
derive water levels in large lakes (e.g., [21] and [112]-[114]),
reservoirs (e.g., [28] and [115]-[117]), rivers (e.g., [111] and
[118]-[120]), and floodplain wetlands (e.g., [5] and [121]-
[123]). It has been demonstrated to be highly efficient and
promising [6], [108].
However, it is still very challenging to obtain accurate
estimates of water level variations, especially for small
water bodies, such as rivers [6], [9], [124]. Two difficulties
may be encountered when using altimeter data to
derive water levels over small water bodies. First, due to
the smaller target size, altimeter signals are mostly contaminated
by noise from adjacent land features, resulting
in degraded accuracy and difficulties [21], [125]. To effectively
address this issue, specific data processing, including
the retracking of waveforms, for altimetry data
remains challenging [111], [126]. Second, the capability
of each satellite altimeter is still limited by long revisit
periods and coarser intertrack spatial resolution. This
limitation will surely cause spatiotemporal gaps during
each altimeter overpass [21], [118]. This problem is anticipated
to be alleviated in the next few years through
missions such as SWOT, Sentinel-3C, and Sentinel-6B
(Table 3), which will provide water level measurements
for small water bodies with higher accuracy [127]-[130].
Furthermore, the combination of multimission altimeter
data would be another better way to overcome this problem
[15], [17], [118], [119], [131].
15
IEEE Geoscience and Remote Sensing Magazine - September 2022

Table of Contents for the Digital Edition of IEEE Geoscience and Remote Sensing Magazine - September 2022
