Science of Remote Sensing 3 (2021) 100019

Contents lists available at ScienceDirect

Science of Remote Sensing

journal homepage: www.journals.elsevier.com/science-of-remote-sensing

Review Article

UAV & satellite synergies for optical remote sensing applications: A

literature review
Emilien Alvarez-Vanhard *, Thomas Corpetti, Thomas Houet
CNRS UMR 6554 LETG, Universit
e Rennes, Place Du Recteur Henri le Moal, 35000, Rennes, France

A R T I C L E I N F O A B S T R A C T

Keywords: Unmanned aerial vehicles (UAVs) and satellite constellations are both essential Earth Observation (EO) systems
Drone for monitoring land surface dynamics. The former is frequently used for its acquisition flexibility and its ability to
Spaceborne supply imagery with very high spatial resolution (VHSR); the latter is interesting for supplying time-series data
Sensor synergy
over large areas. However, each of these data sources is generally used separately even though they are com-
Multiscale
Calibration
plementary and have strong and promising potential synergies. Data fusion is a well-known technique to exploit
Fusion this multi-source synergy, but in practice, UAV and satellite synergies are more specific, less well known and need
to be formalized. In this article, we review remote sensing studies that addressed both data sources. Current
approaches were categorized to distinguish four strategies: “data comparison”, “multiscale explanation”, “model
calibration” and “data fusion”. Analysis of the literature revealed emerging trends, the supply of these distinct
strategies for several applications and allowed to identify key contributions of UAV data. Finally, the high po-
tential of this synergy seems currently under-exploited; therefore a discussion is proposed about the related
implications for data interoperability, machine learning and data sharing to reinforce synergies between UAVs
and satellites.

Introduction GM-satellites in high orbit (e.g. MODIS Terra or NOAA AVHRR) or

geostationary satellites have high temporal frequency, which makes it
The synergy between UAV and satellite data (UAV/Satellite synergy) possible to provide daily to infra-hourly data at coarse spatial resolution
is essential for understanding the dynamics of the Earth’s surfaces (>100 m). EM-satellites (e.g. Landsat or Sentinel-2) are balanced in their
(Kuenzer et al., 2015). On the one hand, these two data sources produce resolutions, providing high temporal data (few days) with high spatial
significant volumes of data contributing to the Big Earth Observation (10–100 m) and spectral (ca. ten bands) resolutions. These first two
(EO) Data (Liu et al., 2018). Indeed, this exponential production of EO categories of satellites provide long-term data time-series, for example,
data since the beginning of the 2000s can be partly explained by the up to 50 years for the Landsat legacy. Next, nano-satellites (e.g. Planet),
increase in the number of satellites in orbit (Ghamisi et al., 2019) and the due to the large number of them placed in low orbit, provide daily global
democratization of UAVs due to the decreasing cost of exploitation (Sun coverage at high spatial resolution (1–10 m) but with a lower data quality
and Scanlon, 2019). On the other hand, each EO system has specific that does not satisfy all applications. Finally, civilian-satellites (e.g.
acquisition features that result from a trade-off between resolutions Pleiades or Ikonos) are low-orbit spaceborne satellites with sensors that
(spatial, spectral and temporal), swath (Fig. 1) and signal-to-noise ratio provide data at very high spatial resolution (VHSR; <1 m) but with low
(Alavipanah et al., 2010). Pending a new technology with all required spectral resolution (5 bands), and whose cost limits the achievement of
features, it is necessary to combine data from different sources to global coverage and dense time series. Globaly, optical remote sensing
enhance observations. Satellite constellations and UAV for EO provide (RS) satellites have little flexibility in their acquisition features, for
complementary data interesting for a synergy approach. example they are constrained by cloud cover, view angle or acquisition
Current satellite constellations are numerous and offer different time.
trade-offs that can be grouped into four categories: global monitoring Remote sensing by UAV (RS UAV) has emerged due to the develop-
(GM-), environmental monitoring (EM-), nano- and civilian-satellites. ment of robotics, computer vision and sensor miniaturization (Colomina

* Corresponding author. CNRS UMR 6554 LETG, Universit
e Rennes 2, Place du recteur Henri le Moal, 35000, Rennes, France.
E-mail address: emilien.alvarez-vanhard@univ-rennes2.fr (E. Alvarez-Vanhard).

https://doi.org/10.1016/j.srs.2021.100019
Received 21 October 2020; Received in revised form 24 January 2021; Accepted 25 January 2021
Available online 2 February 2021
2666-0172/© 2021 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
E. Alvarez-Vanhard et al. Science of Remote Sensing 3 (2021) 100019

Fig. 1. Resolution requirements (temporal, spatial, spectral and swath) in the main application fields of remote sensing and data source supply. Information based on
Briottet et al. (2011) (Briottet et al., 2011), Riihim€aki et al. (2019) (Riihim€aki et al., 2019), Transon et al. (2018) (Transon et al., 2018), and Zhu et al. (2018) (Zhu
et al., 2018). EO: Earth Observation; EM: Environmental Monitoring; GM: Global Monitoring; sat.: satellite; LC: Land Cover; LU: Land Use.

and Molina, 2014). The openness of data-acquisition skills (Milas et al., flexibility, it requires a ground operator, management of large volumes of
2018), due to the development of micro-UAVs (i.e. weight less than a few data and pre-processing, while satellite data are easily available on web-
kg), changes the paradigm of RS by giving end-users the ability to control based platforms and are generally ready to analyze. Furthermore, UAV
acquisition features. Ultraspatial resolution (centimetric to millimetric) provides data at a resolution unreachable by satellite but cannot rival the
and acquisition flexibility are the strengths of RS UAV. The ability of latter’s observed extent and remains constrained in particular territories
UAVs to acquire data close to the surface allows such spatial resolution by national and/or international legislation. Between these two sources
but also the ability to be quasi-independent or less-affected by clouds. of RS data, advantages of the former appear to compensate for disad-
UAV flexibility allows acquisition conditions to be chosen: the type of vantages of the latter, and vice versa, revealing a strong potential for
sensor, angles of view, spatial resolution, time and frequency of acqui- synergies. Moreover, it is considered necessary to use this synergy (Zhu
sition. The choice of acquisition dates and times is an essential charac- et al., 2018; Vihervaara et al., 2017) to bridge the gap between the
teristic for certain applications, such as monitoring biological, abilities of EO systems and the data needs of different application fields
hydrological or geomorphological dynamics (Abdullah et al., 2018; (Fig. 1).
Müllerov
a et al., 2017; Fytsilis et al., 2016). However, RS UAV has the To our knowledge, there is no review of the literature that analyzes
disadvantage of having a smaller swath (a few km2), often because of its UAV and satellite synergies. The synergy between multi-resolution RS
low energy reserves and the legislation needed to protect air traffic and data has already focused on the complementarity of satellite and airborne
people safety and privacy (Cracknell, 2017). data (Liu et al., 2018; Zhang, 2010), however the complementarities
Finally, UAVs offer a different and complementary profile to satel- between UAV and satellite are more specific (Kakooei and Baleghi,
lites. There are complementarities in resolution between UAV and sat- 2017). We therefore attempt to fill this gap in the scientific literature by
ellite acquisitions, particularly in spatial and temporal resolution and reviewing these synergies. This review categorizes four types of strategy,
swath (Fig. 1), as well as advantages and disadvantages of using each identify emerging trends and key contributions of UAV data. Then a
vector (Table 1). For example, although RS UAV has high acquisition discussion is proposed to overpass current limits and to fully exploit the
potential of this synergy.

Table 1 Methods
General features of UAV and satellite vectors. Advantageous features are in bold.
Feature/EO sytem UAV Satellite Literature review process
Flexibility High Low
Cloud dependence No Yes This review aimed to collect, in the most exhaustive way, peer-
Direct meteorological constraint Wind and precipitation No reviewed articles that dealt simultaneously with UAV and satellite
Pre-processing High Analysis ready data data. To this end, the Google Scholar, Science Direct and Web of Science
Operator required Yes No
databases were queried for the period 2000–2019. Articles were first
Data management High Low
VHSR cost Low High selected using the following query:
Payload Interchangeable Permanent
Legislation Restrictive None
“UAV syn” AND “Satellite syn"

2
E. Alvarez-Vanhard et al. Science of Remote Sensing 3 (2021) 100019

with UAV syn and Satellite syn corresponding to a list of synonymous The last criterion distinguishes studies that develop models (e.g.
terms or well-known sensor names (Appendix A). A total of 495 articles classification, regression or detection) based on only one data source
were found in this search. By analyzing the entire corpus, articles that (“UAV for Satellite”) from those based on both sources (“UAV and Sat-
mentioned the two data sources but did not deal with them were elimi- ellite”). The former defines “model calibration” strategy, i.e. one data
nated. The remaining articles were then filtered to retain only those that source is used to calibrateingion by inference a model based on the other
dealt with optical data acquired by a passive sensor (panchromatic, RGB, source (Fig. 3C). Two sub-categories are distinguished: qualitative (e.g.
multispectral and hyperspectral) and related to the study of the Earth’s hard classification) and quantitative (e.g. regression or soft classifica-
surface (excluding studies of the atmosphere and oceans, which have tion). The first generally refers to labeling satellite pixels by interpreting
other specific characteristics). Ultimately, the corpus of our study was UAV data. While the second refers mainly to the use of raw (e.g. reflec-
based on 137 articles published in peer-reviewed journals (Appendix A). tance) or derived (e.g. biophysical variables) numerical values from UAV
data to calibrate a satellite-based model. In a few rare cases, the roles of
UAV and satellite data are reversed.
Categorization in strategy types The last type of strategy, “data fusion”, concerns studies in which new
data are produced from a model based on both UAV and satellite data
To guide future studies in the exploitation of UAV/Satellite synergies, (Fig. 3D). These methods make it possible to improve the resolution of
we categorized articles of our corpus. Our categorization approach each data source. There are several ways to fuse data by coupling or
distinguished different strategies according to a three-criteria hierarchi- complementing spatial, spectral and temporal resolutions. In general,
cal decision tree (Fig. 2). fusion techniques in RS can be classified into three categories according
The first criterion distinguishes weak and strong synergies (Fig. 2) to the degree of fusion (Zhang, 2010): pixel-level, feature-level and
based on the degree of data integration. The synergy will be strong if decision-level.
combining the data provides more information than using each data The bibliographical analysis on which this review is based was per-
source separately (“1 þ 1 ¼ 300 ) (Pohl and Genderen, 1998). Conversely, formed with the help of a reading grid (Appendix A). Using this grid, a
weak synergy, called “data comparison”, compares only advantages and variety of information was extracted (e.g. application field, strategy type,
disadvantages to determine which data source is the most suitable (“UAV data type, methods).
or Satellite”) for the application under study (Fig. 3A). Data combination
in this category is for visual or quantitative comparison but does not help Results
in the final interpretation.
Among the strong synergies (Fig. 2), the second criterion distin- Emerging trends
guishes studies whose observation object is the same for both UAV and
satellite acquisitions from those whose object is not the same. The latter The recent emergence and growing interest in UAV/Satellite syn-
refers to the “multiscale explanation” strategy, in which the objects ergies was evident from the bibliographical analysis (Fig. 4A). The first
observed by UAVs and satellites are at different spatio-temporal scales, articles addressing this synergy emerged in 2008, which logically coin-
generally with a different spatial extent. Information extracted at a finer cided with the democratization of UAVs. Interest in this practice
resolution for a small site is used to explain the information for a larger remained modest for several years, accumulating barely 19 articles by
extent that contains the former. Generally, these studies observe an ob- 2013. From 2014-2019, the number of studies increased considerably,
ject precisely by UAV and use satellite data to obtain the more global reaching 37 articles published in 2019.
context in which this object is located. The data are therefore processed Reviewed studies were unevenly distributed among the strategy
separately, and the extracted information is then used together to categories (Fig. 4B). Of the 137 articles, 29% compared only UAV and
improve the scientific interpretation (Fig. 3B).

Fig. 2. Hierarchical decision tree for categorizing UAV/Satellite strategies.

3
E. Alvarez-Vanhard et al. Science of Remote Sensing 3 (2021) 100019

Fig. 3. Diagram of UAV/Satellite strategies.

satellite data and thus exhibited weak synergy. However, when the fusion. Fig. 5 shows that the UAV/Satellite synergy emerged with the
synergy was strong, “model calibration” strategy was most common, at “calibration model” strategy and evolved with it. The “data comparison”
48% of the articles (qualitative and quantitative pooled). “Multiscale appeared later, in 2012, and seemed to evolve slightly. The “multiscale
explanation” and “data fusion” strategies were under-represented (9% explanation” appeared relatively early but remained punctual. And
and 10% of the studies reviewed, respectively). The undefined category lately, the “data fusion” appeared in 2014, and although present every
represented only 0,7%, one methodological article about multiscale data year until 2019 it seemed to evolve poorly.

4
E. Alvarez-Vanhard et al. Science of Remote Sensing 3 (2021) 100019

Fig. 4. Distribution of studies that address UAV/Satellite synergies. A) Number of articles published from 2008-2019; B) Distribution of studies among the strategies
identified; C) Distribution of articles among the main application areas of Earth Observation.

Fig. 5. Distribution of the strategy categories in the period 2008–2019.

Application fields were also unevenly distributed among the strategy strategies through “data comparison” or “multiscale explanation".
categories (Fig. 4C). Ecology and precision agriculture were the main Optical data used for these synergies came mainly from EM- and
fields of application of these synergies, representing respectively 48% civilian-satellites, while nano- and GM-satellites were rarely used
and 26% of the reviewed articles. While ecology was based mainly on (Fig. 7A). This may be due to the recent release of nano-satellites and to
“model calibration” strategy, precision agriculture used all strategies, the coarser spatial resolution of GM-satellites than that of UAVs (i.e. the
with a slight preference for “data fusion”. Although ecology was the field scale factor is too high). Overall, 58% of the studies reviewed used open-
with the most UAV/Satellite synergies, it had no “multiscale explanation” source satellite data such as Sentinel-2, Landsat, Gaofen or MODIS, thus
strategy study for the period studied. Less common fields of applications demonstrating the contribution of open-source data to this synergy. On
like geosciences and disaster were frequent over time, unlike applications the UAV side, multi-rotor UAVs seemed to be used more than fixed-wing
in archaeology, urban and water resources that were rare and punctual. It UAVs, although a large percentage (30,7%) of the articles did not
can be seen that these less common fields of application were mainly mention the characteristics of the UAV used (Fig. 7B). Although, 10% of

5
E. Alvarez-Vanhard et al. Science of Remote Sensing 3 (2021) 100019

articles did not mention the types of UAV optical data collected (RGB, “Multiscale explanation” strategy
multiscpectral or hyperspectral) it seems that RGB data was more used
for strong synergies, and multispectral data for “data comparison” The “multiscale explanation” strategy combines the observation
(Fig. 6). On average, UAVs provided a resolution of 26 cm in a range of scales of each data source to interpret the data better. This complemen-
0.2–500 cm depending on the application, with a median of 8 cm. The tarity was used mainly for natural risk management (33%), geosciences
scale factor between UAV and satellite data therefore varied from 2- (25%) and archaeology (25%). In natural risk management, this
20,000, with a median of 100. approach enables, for example, relations between the presence of gullies
and agro-industrial development to be established (Aït Hssaïne et al.,
Strategy types 2011). The satellite is thus used to obtain information about the dy-
namics of changes in land occupation and use, while the UAV provides
“Data comparison” strategy fine-scale analysis of the gullies via photo-interpretation and analysis of
digital surface models (DSM) generated by photogrammetry (Aït Hssaïne
“Data comparison” is a weak synergy in which data are not combined et al., 2011; Wang et al., 2016). In geosciences, a study of yardangs in a
to improve interpretation. In comparativet studies the specific features of desert environment showed that UAVs can be used to extract morpho-
UAVs and satellites are identified, and thus their complementary nature logical features, while satellites provided information about their spatial
is highlighted (Fig. 1). distribution (Zhao et al., 2018). In archaeology, satellite imagery is used
Satellites provide data with a larger extent that UAVs cannot achieve either for prospection studies (Ding et al., 2016) or to contextualize an
(Jacobsen, 2012; Rau et al., 2014). They cover areas that are difficult for archeological site in its environment (Gruen et al., 2012; Lin et al., 2011),
UAVs to access, such as urban areas (Müllerov
a et al., 2016) or conflict while UAVs reveal details of archeological sites that are invisible even
zones (Matouskov
a et al., 2016). Moreover, standardization of the from the ground (Lin et al., 2011) and provide spatial description in three
pre-processing of satellite data (analysis-ready data) facilitates and ac- dimensions (Ding et al., 2016; Gruen et al., 2012). And, an original use of
celerates the processing of these data (Müllerov
a et al., 2016). In addi- this approach made it possible to study coastal dynamics both physically
tion, UAVs provide low-cost VHSR data that reveal fine patterns such as and socially (Papakonstantinou et al., 2019). Satellite data provided in-
inter-row and intra-plot variability in vineyards (Aleem et al., 2019; formation on the evolution of the coastline and geomorphological
Matese et al., 2015), the fine morphology of glaciers (Fugazza et al., changes of available beaches, while UAVs counted and characterized
2015), and even small water bodies (C^andido et al., 2016). This VHSR tourist infrastructure on these beaches. Finally, through this strategy,
also makes it possible to isolate the study object from areas that could satellites make it possible to analyze spatial or temporal patterns,
interfere, such as adjacent bare soils (Aleem et al., 2019) and shadows contextualize or locate at a regional scale, while UAVs provide additional
(Rupasinghe et al., 2019), which can bias the spectral signatures of information, often in the DSMs produced by photogrammetry, or refine
vegetation. The low cost of the technology makes this vector particularly the characterization of the study objects.
attractive; however, according to Ruwaimana et al. (2018) (Ruwaimana
et al., 2018) UAVs are cost-effective for acquisitions only for long-term
“Model calibration” strategy
monitoring. Finally, the flexibility of UAVs is a major asset when ac-
quisitions must be made at a specific date or time, as may be the case for
“Model calibration” uses one data source to calibrate a model (qual-
precision agriculture to determine the addition of inputs (Brinkhoff et al.,
itative or quantitative) based on the other data source. Among the articles
2018; Jurecka et al., 2018), or to perform rapid assessment during
classified as strong synergies, it was the most frequent strategy, with 70%
climate events and natural disasters, such as landslides (Casagli et al.,
of the articles (“data comparison” excluded) and was widely used by
2017).
applications in ecology and precision agriculture (Fig. 4B).
To conclude the comparison of UAVs and satellites, the choice of data
Typically, the qualitative approach allowed satellite-based classifi-
source depends mainly on the scale (extent, resolution) of analysis and
cation (or post-classification (Nhamo et al., 2018)) to be made using
the objective of the study. The features required for the study (e.g.
labelled samples from UAV data. Labels can be determined by: 1)
ground resolution, temporal frequency, types of acquisition, extent) will
photo-interpretation by an expert, 2) thresholding of metrics (e.g. spec-
guide the choice of data source the most, but it may also depend on the
tral indices or biophysical parameters) or 3) automated classification
study’s organizational and financial conditions (Jacobsen, 2012). Among
(pixel or object oriented) that can be supervised or not. These labels are
these comparative studies, 19% of them considered strong UAV/Satellite
then used to calibrate supervised classification models. Those produce
synergies the most in their perspectives.
land-cover or land-use maps and thus provide information about

Fig. 6. Types of UAV optical data used among strategies.

6
E. Alvarez-Vanhard et al. Science of Remote Sensing 3 (2021) 100019

Fig. 7. Types of satellites (A) and UAVs (B) used for UAV/Satellite synergies.

Fig. 8. Diagram of the “data calibration” strategy.

7
E. Alvarez-Vanhard et al. Science of Remote Sensing 3 (2021) 100019

landscape structure like the spatial distribution of habitats, and can help Hassan-Esfahani et al. (2017) (Hassan-Esfahani et al., 2017)
forto change detection like deforestation in wooded systems (Marx et al., improved the spatial resolution of Landsat imagery by a factor of four
2017). For example, Szantoi et al. (2017) (Szantoi et al., 2017) used this through supervised learning using UAV image patches. Thus,
strategy to map land cover and land use and quantify loss of primary high-frequency details in each band and their derivatives can be
forest to assess their impacts on orangutan habitat in Indonesia. recovered from Landsat imagery.
The quantitative approach uses numerical values directly from UAV  Spatial-temporal (Zhu et al., 2018): Pioneering studies demonstrated
data to calibrate a satellite-based model. Usually, the numerical values the contribution of spatial-temporal fusion, especially the Spatial
extracted are (bio)physical parameters of the surface (e.g. chlorophyll Temporal Adaptive Reflectance Fusion Model algorithm (Gao et al.,
content or aboveground biomass) derived from spectral measurement 2006). Gevaert et al. (2014) (Gevaert et al., 2014) combined UAV
(Zhang et al., 2019) or from photogrammetrically estimated height hyperspectral data with a Formosat-2 time series to derive biophysi-
models (e.g. tree height (St-Onge and Grandin, 2019)). But UAV data cal parameters (leaf area index and chlorophyll content) of potato
may also have other natures, such as land-cover rates for sub-pixel vali- crops and make consistent predictions with fine spatial patterns. In
dation of soft classification models to estimate the fraction of vegetation the same vein, Liu et al. (2019) (Liu et al., 2019a) fused a sparse UAV
cover (Riihim€ aki et al., 2019; Melville et al., 2019), for example to time series with a denser time series from the Planet nano–satellite to
improve estimation of flooded areas (Xia et al., 2017) or detection of estimate grassland forage production finely.
invasive species (Kattenborn et al., 2019). Lastly, raw radiometric data  Spectral-temporal: Gevaert et al. (2015) (Gevaert et al., 2014) used
can be used directly to calibrate spectral unmixing models. For example, Bayesian theory to fill in missing spectral information in multispectral
Alvarez-Vanhard et al. (2020) (Alvarez-Vanhard et al., 2020) used pixels satellite data with hyperspectral data from a UAV. The
considered to be “pure” (not mixed) extracted from UAV data to calibrate Spectral-Temporal Response Surface model designed in their study
an unmixing model of wet grassland habitats based on Sentinel-2 data. provided continuous spectral reflectance at high temporal frequency.
A specific case of the “model calibration” strategy is the data inter- The results generated correlated well with spectral measurements in
calibration where the spectral signature is modeled (Fig. 8). The inter- the field (r ¼ 0.953) and allowed for derivation of biophysical vari-
calibration needs a reference dataset to calibrate the rest of the dataset. ables (leaf area index and chlorophyll content) that were consistent
This reference can be the satellite data to calibrate a finer resolution with the observations.
(Houborg and McCabe, 2018) or the UAV data, itself calibrated by in-situ
spectroradiometer measurements, to calibrate the satellite data (Padr
o The remaining one-third of the studies used a feature-level fusion
et al., 2018). approach with a less classic approach. For example, Kakooei and Baleghi
In most cases, UAV data were used in addition to in-situ data. (2017) (Kakooei and Baleghi, 2017) used different UAV and satellite
Generally, these validation data were integrated in a nested model. For view angles to assess post-disaster damage on buildings. Using oblique
example, Xia et al. (2017) (Xia et al., 2017) used in-situ surveys to images obtained by the UAV, they extracted features on building facades,
validate an object-oriented UAV classification that then facilitated se- while satellite imagery provided features on roofs. These features were
lection of “pure” pixels for a satellite-based spectral unmixing model, as fused to estimate damage levels. Another application in change detection
well as to obtain independent training and validation data. However, used a method for near real-time detection (Fytsilis et al., 2016) that
32% of the studies used UAVs to replace in-situ surveys, and thus compared features of UAV data to those of historical satellite data in
depended only on interpretation of VHSR images. Some studies order to detect potential changes in a territory. The method developed
compared the influence of the use of in-situ data to that of UAV data. For was presented as being robust to differences in spatial and spectral res-
example, Forster et al. (2018) (Forster et al., 2018) demonstrated that olution and misregistration issues. Finally, decision-level fusion,
UAVs can be used with confidence to provide ground truth for producing although having interesting potential (Atkinson, 2013), was not used at
thematic maps. Spence and Mengistu (2016) (Spence and Mengistu, all with UAV and satellite data.
2016) showed that an intermittent-stream detection model calibrated
with in-situ data was more accurate than a UAV-calibrated model, but the Key contributions of UAV data
latter was more robust. Finally, Melville et al. (2019) (Melville et al.,
2019) and Liang et al. (2017) (Liang et al., 2017) were specific cases who In this synergy, UAVs played a “bridging role” by complementing and
calibrated a UAV-based model with satellite data. magnifying the potentials of in-situ and satellite data (Fig. 9). UAVs
provided new data that can be acquired only by this vector and thus
“Data fusion” strategy provided a single intermediate observation scale. This characteristic
allowed UAVs to play three types of roles according to the strategies: (1)
“Data fusion” can be considered as the strongest synergy because it explanation, (2) validation and (3) completion of satellite data.
tries to use the features of each data source fully to create new data. This Explanation was a major use of RS UAV within the UAV/Satellite
strategy, although little used, was nonetheless explored by the precision synergy. It consisted of providing complementary data (e.g. VHSR or
agriculture field (Fig. 4), which aims to extract the land cover and bio- DSM) to reveal inaccessible details or unseen from space or on the
physical features of vegetation cover at a fine resolution. ground. In-situ observations made by experts or a network of sensors are
Two-thirds of “data fusion” strategy articles were pixel-level studies generally precise but punctual (Gamon, 2015), while satellite observa-
with the objective of creating an enhanced-feature dataset by combining tions cover large areas but have a resolution that remains too low to be
one or more resolutions of each source. The most basic method is the interpreted properly without in-situ information. Although these two
densification or completion of time series. For example, using a multi- observation scales complement each other, pairing them remains how-
source time series Nikolakopoulos et al. (2019) (Nikolakopoulos et al., ever uncommon because of the difference in surface of the areas they
2019) mapped evolution of the coastline, and Firla et al. (2019) (Firla observe (Alvarez-Vanhard et al., 2020). Between the two scales, UAVs
et al., 2019) estimated intra-seasonal variations in the penguin popula- provide spatially explicit local data that reveal spatial patterns of prime
tion in Antarctica. More elaborate approaches combine two dimensions: importance to study processes using satellite measurements (Gamon,
2015; Fawcett et al., 2020). This role of UAVs was particularly high-
 Spatial-spectral (also known as super-resolution (Yue et al., 2016)): lighted in the “multiscale explanation” strategy, but it also forms the
Jenerowicz et al. (2017) (Jenerowicz et al., 2017) used the basis of all strong UAV/Satellite synergies.
Gram-Schmidt pansharpening method with Landsat imagery to The validation role consisted of using UAV data as “ground truth” (or
improve the spectral resolution of low-cost UAV sensor imagery and “drone truth”) and relied on UAVs’ ability to explain satellite data. This
thus significantly improve the accuracy of land-cover classification. was the main purpose of UAV data in this synergy in particular via the

8
E. Alvarez-Vanhard et al. Science of Remote Sensing 3 (2021) 100019

“model calibration” strategy. UAV data can be used in synergy with in- toward new applications that require very high resolutions, such as
situ data in a nested model or can replace them. As mentioned, UAV monitoring hydrological regimes, phenological traits of plants or degra-
data can replace in-situ observations for applications in which the object dation states of environments.
to be detected was clearly identifiable, such as water bodies (Tschudi
et al., 2008; G
oraj et al., 2019), the fraction of plant cover (Bian et al., Discussion and perspectives
2016) or certain invasive species (Martínez-S
anchez et al., 2019; Elkind
et al., 2019). Indeed, UAVs provided spatially explicit data fine enough A still under-exploited synergy
for experts to identify elements of the landscape directly. Moreover,
UAVs had the advantage of covering modest to large areas quickly and The analyses made in this paper have highlighted interesting syn-
can easily reach the least accessible environments. In the current era of ergies between UAV and satellite data. However, to our opinion, current
data-driven models developed by non-parametric supervised learning synergies are not fully exploited and one can go a step forward to go
methods (e.g. random forest, support vector machine or neural net- beyond the validation purpose of UAV data and exploit together UAV and
works), UAVs can play a key role in providing “ground truth”. Indeed, satellite data. Today this mutual exploitation is limited to some studies
validation data are critical in these methods, but the heavy logistical mainly for the two following reasons: interoperability is not obvious and
involvement required for large-scale acquisitions reduces the potential of UAV/Satellite synergies often answer to a specific use case without the
such approaches. UAVs provide an affordable solution for acquiring aim of fully exploring its potential. These two points are discussed below.
ground-truth data to analyze or calibrate satellite data. Carbonneau et al.
(2020) (Carbonneau et al., 1002) showed that this approach is valid,
robust and allows for hydrogeomorphological analysis to be extended to Data interoperability remains challenging
the regional scale by applying locally trained models to new satellite
data. The transferability of the models needs to be assessed more widely The quality and interoperability of satellite data are guaranteed by
in terms of seasonal and interannual variability and also between the Committee on Earth Observation Satellites (CEOS), whose original
different geographical areas. Nonetheless, this approach should be tested function was to standardize data formats and ensure the validation, inter-
on other applications. calibration and inter-comparison of satellite products. For UAV data,
Lastly, the completion role of UAVs lied in their ability to fill gaps in however, there is no guarantee of data quality since the data-acquisition
satellite data. Spontaneous UAV acquisitions made it possible to fill skills are transferred to users. Acquisition and pre-processing protocols
temporal gaps in sparse satellite time series (Mengmeng et al., 2017) and can vary among users and sensors, which does not guarantee consistent
to produce change-detection maps in near-real time (Fytsilis et al., 2016). data, thus making multi-source interoperability difficult. Ensuring this
It can also fill spatial gaps, e.g. in which optical satellite data were interoperability is an important challenge because models based on the
degraded by cloud cover. UAVs can provide data under clouds, which is synergy between UAV and satellite data are sensitive to the quality of the
particularly useful in tropical areas where cloud cover is frequent for long input data (Carbonneau et al., 1002; Mengmeng et al., 2017; Belgiu and
periods of the year. Through fusion methods, UAVs can improve spatial, Stein, 2019). Intercalibrating data (geometric and radiometric) and
temporal or spectral resolutions of satellite data by providing a comple- estimating uncertainty in multisource models are therefore essential
mentary dimension. These methods were still rarely used at the scales of steps to ensure the quality of the results from this synergy.
observation provided by UAVs, but the first applications (Gevaert et al., Misregistration between data from one or more sources causes errors
2014, 2015; Liu et al., 2019a) show high potential that can open the field in analysis at different spatial and temporal scales. To ensure the quality
of analyses, sub-pixel geometric intercalibration is necessary (Fytsilis

Fig. 9. Distribution of Earth Observation systems by spatial scale (resolution and spatial extent), based on Riihim€aki et al. (2019) (Riihim€aki et al., 2019). Ultrahigh:
<1 m, High: 1–10 m, Medium: 10–100 m, Coarse: >100 m.

9
E. Alvarez-Vanhard et al. Science of Remote Sensing 3 (2021) 100019

et al., 2016; Pohl and Genderen, 1998). Although pre-processing UAV properties, such as anisotropy (Roosjen et al., 2017; Su et al., 2007;
and satellite data provides sub-pixel inter-band and inter-date co-regis- Martonchik, 1994). Similarly, optical properties of the surface observable
tration, their different resolutions imply calibrations of different orders by multispectral sensors vary and are related to acquisition conditions
(centimetric and metric, respectively). Methods for automatically (e.g. wavelengths, viewing angle, ground resolution); thus, the flexibility
georeferencing multisource data can ensure this consistency. For of UAVs does not yet appear to be fully exploited and new applications
example, the multi-scale SIFT (Scale-Invariant Feature that use both UAVs and satellites that may lead to new types of synergies
Transform)-RANSACT (RANdom SAmple Consensus) methods (Oh et al., remain to be discovered and developed. Lastly, technological advances
2011) automatically generate tie points to locate the data correctly using related to the installation of LiDAR and hyperspectral sensors on satellites
an affine transformation, or optical flow methods (Brigot et al., 2016), or UAVs will also increase the potential of this synergy.
which do not use tie points and are effective for multimodal datasets. Currently, this synergy remains in the “UAV or Satellite” and the
Using the full optical spectrum (multispectral or hyperspectral) “UAV for Satellite” paradigms. The former corresponds to the “data
through different sensors requires ensuring consistency between these comparison” use case, whose purpose is well summarized by the question
radiometric measurements, which remains a challenge in itself. Indeed, of Abdullah et al. (2018) (Abdullah et al., 2018): “Can UAVs replace
reflectance values may vary depending on the type of sensor (wave- Satellites?”. This purpose can finally be seen as opposed to a synergy. The
length, vignetting), acquisition protocol (angle of view, ground resolu- latter paradigm refers to the “model calibration” strategy using UAV data
tion), environmental conditions (atmosphere, topography) and for a validation purpose. Although this strategy is efficient to answer
corrections made to the data (Schaepman-Strub et al., 2006). For scientific questions, we think this way is not optimal and doesn’t keep
example, the bi-directional reflectance distribution function of the sur- information from the multiscale observations. Actually, UAV data is often
face can be ignored and considered as Lambertian for observations at low reduced to a label or up-sampled to satellite spatial resolution losing fine
spatial resolution (satellite), which is not the case for UAV observations, spatial patterns. The strongest synergies “multiscale explanation” and
whose high variability in reflectance values is due to the heterogeneity in “data fusion” are strategies preserving information from each source, and
the optical properties and 3D structure of the surfaces, which leads to should be more explored to move towards a new paradigm: “UAV and
heterogeneous observation geometries (Stow et al., 2019; Stark et al., Satellite”.
2016). Moreover, low-cost UAV sensors may have limitations or defaults
that are not yet well known and that make it difficult to combine them Towards a stronger exploitation of synergies
with satellite data (Fawcett et al., 2020). These differences in reflectance
measurements between UAV and satellite observations must therefore be From the previous section one observes that UAVs have completed
considered because they influence multi-scale models. In the absence of satellite images for validation, completion or explanation purposes but
absolute reflectance measurements, radiometric intercalibration is an these two sources of data have not that much been used together for
effective solution for multi-sensor interoperability as seen in the section fusion issues. This is to our opinion due to the fact that:
“Model calibration” strategy.
Despite data correction and intercalibration, measurement uncer- - the questions of interoperability are tricky;
tainty can persist and must be considered as it propagates into the scales - usually UAV are acquired for a specific application and no more
of fusion or nested inference models. For example, Solazzo et al. (2018) exploited then;
(Solazzo et al., 2018) used Monte-Carlo simulations to estimate uncer- - fusion requires advanced methodologies not easily accessible.
tainty in a dune-volume prediction model based on UAV data and express
it in a satellite-based model. From this latter point, it is important to outline that since several
Up to now, these questions of interoperability and intercalibration years, spectacular advances in machine learning and especially with deep
have strongly limited the nature of exploitation of the synergies between neural networks have enabled a breakthrough in the processing of
UAV and satellite images. massive and heterogeneous spatial data.

An unexplored potential Machine learning for UAV/Satellite fusion

From our review, it seems that the potential of the UAV/Satellite The idea behind deep learning is to construct a neural network
synergy is currently not fully exploited: (1) several scientific fields do not composed of a large number of layers, enabling to model very complex
take enough advantage of this synergy, (2) the capacities of the different relations between inputs (multi-source data) and outputs (estimation of
EO systems are under-exploited, and (3) stronger synergies are less used. parameters, labels, etc). Though the idea is old, the progresses in the
Ecology and agriculture scientific fields have been mainly involved in recent years come from the fact that we have now enough data and
the exploitation of UAV/Satellite complementarities, unlike other earth associated computational resources to train such complex networks (and
observation fields (geosciences, disaster, archaeology, water resource or then to condition the optimization processes associated with the large
city monitoring) that generally require less important resolutions (Fig. 1). number of calibration parameters to estimate). In addition, some theo-
However, a breakthrough can be anticipated in these less common fields. retical progress on the definition and optimization of such networks have
For example, Antoine et al. (2020) (Antoine et al., 2020) highlight the opened a wide range of applications. The reader can find in (Yan et al.,
challenge of combining different observation scales for geosciences and 2015) a general introduction to deep learning.
disaster purposes, and identify combined workflow between space-based As for the processing of spatial data, the state of the art network for
and UAV data as a proper solution. assigning a label to an image is the well-known CNN (Convolutional
EO systems used in this synergy are under-exploited, in particular for Neural Network). Since then, many architectures, either adapted to
UAV capacities. First, EM-satellites are most used, certainly because these assign a label to each pixel (Fully Convolutional Networks and variants
data are open (e.g. Sentinel-2, Landsat and Gaofen) and their observation (Long et al., 2015)), to deal with unstructured data (Qi et al., 2017) or to
scales provide good complementaries with UAV data. Yet, nano- and time series for example (Fazle et al., 2017) have been proposed. More
civilian-satellites are interesting in particular for the “data fusion” recently, the community is also focusing on the fusion of complex data, as
strategy in view of similarities and the low scale factor with UAV data, for example (Zhenfeng and Cai, 2018) in the context of satellite images.
but their cost remains a barrier. Then, the use of UAV data is too limited The great quantitative improvements have encouraged researchers to
to RGB, leaving the high potential of optical RS still unexploited. Specific explore the fusion of heterogeneous sources and some studies using UAV
characteristics of the observation geometry remain little used but have an have already been proposed (Liu et al., 2019b; Barrero and Perdomo,
interesting potential to improve characterization of surface reflectance 2018). Obviously we encourage the community to explore this point and

10
E. Alvarez-Vanhard et al. Science of Remote Sensing 3 (2021) 100019

to fully exploit the synergies between UAV and satellites. Conclusion

Associated with these new methodological insights, the open access
to data is also a trend the community has to follow. The scientific literature was reviewed to guide future studies that
wish to use the strong complementarities between UAVs and satellites.
Facilities for multisource workflow Overall, 137 articles published in peer-reviewed journals were collected.
This corpus of literature made it possible to identify four main strategies:
The UAV/Satellite synergy has the potential to overcome current “data comparison”, “multiscale explanation”, “model calibration” and
limitations of EO systems. However, the complementarities of the data “data fusion”. These four types of synergy helped in filling the gap be-
sources are not sufficient for this potential to be fully exploited. The tween sensor capacities and the data needs in EO applications. However,
combined workflow between UAV and satellite data must be facilitated this categorization was specific to the corpus of this review and will aim
and adapted to the Big EO Data. To achieve this, facilities are needed to to evolve with future applications. In just a few years, this synergy
ensure the Findability, Accessibility, Interoperability, and Reusability emerged with various strategies and applications. Through this biblio-
(FAIR principles (Wilkinson et al., 2016)) of data and methods. In this graphical study highlighted the following trends:
section, we discuss current initiatives that contribute to this vision of an
open and reproducible EO science. 1. Ecology was the main application area that used this synergy and used
There is a trend to the openness of satellite data which are now easier mostly the “model calibration” strategy.
to find due to the multiplication of downloading platforms based on the 2. A significant part (29%) of UAV/Satellite synergies were weaks (“data
model of ESA’s Copernicus program. Conversely, sharing of UAV data is comparison”).
not following this trend to the same degree. The difficulty in sharing UAV 3. “Multiscale explanation” and “data fusion” strong synergies were
data is understandable given the large number of suppliers who acquire under-exploited.
data with specific objectives. The lifespan of UAV data is therefore 4. RS UAV can replace in-situ surveys for basic applications.
currently limited to a single use, although it could be used more widely 5. UAV capacities offered greater potential than what was currently
for global applications if they would become open-access. Nonetheless, used.
initiatives such as Open Aerial Map (OpenAerialMap. OpenAerial, 2020)
or the GEOSS platform (GEOSS portal, 2020) allow UAV data to be We concluded that the UAV/Satellite synergy evolved quickly and
shared freely. At European level, the INSPIRE directive (European provided proper solutions to answer scientific questions in need of
Parliament and t, 2007) requires public authorities to make their envi- multiscale observations. However, this potential was under-exploited
ronmental spatial data open and accessible on the internet. This directive using mainly UAV data for validation purposes. We suggest to the sci-
also concerns public bodies producing ortho-imagery by UAV. For entific community to explore “multiscale explanation” and “data fusion”
example in the West of France, the spatial data infrastructure Indigeo strategies to fully exploit these multiscale data. Advances in multisource
contributes to this effort by sharing UAV data among other geodata interoperability, data sharing and machine learning will help move to-
(Indigeo, 2021). However, the free sharing of data raises the issue of wards these stronger synergies.
personal data and privacy that are accessible through VHSR UAV imag-
ery, i.e. information allowing the direct identification of natural persons Funding
(e.g. identifiable faces), or indirect (e.g. number plates) and private
property. For example, in the European Union, the diffusion of data must This study was supported by public funds (R
egion Bretagne) received
be done in conformity to the General Data Protection Regulation (GDPR) in the framework of the emerging research project (PER) coordinated by
(European Parliament and o, 2016) which prohibits the diffusion and “Groupement Bretagne T
el
ed
etection” (BreTel), the University of Rennes
processing of personal data without the consent of the person, owner or 2, the ANR project MATS (ANR-18-CE23-0006) and from the European
manager of a property. Lastly, data sharing must be accompanied by Regional Development Fund (ERDF) under the umbrella of INTERREG
explicit description of the acquisition conditions based on the model of Atlantic Area (EAPA_261/2016) “Improving the management of Atlantic
the Spatio Temporal Asset Catalog initiative (STAC (SpatioTemporal Landscapes: accounting for biodiversity and ecosystem services (ALICE)”.
asset cata, 2020)), which facilitates queries of spatial data. Indeed, as
mentioned, interoperability must also be achieved for the radiometric
Declaration of competing interest
measures, which depend on the acquisition conditions. These metadata
and their meanings are widely transmitted for satellite products but,
The authors declare that they have no known competing financial
again, it is not yet standard for UAV products.
interests or personal relationships that could have appeared to influence
Finally, infrastructure such as EO Data Cubes (EODCs) facilitate and
the work reported in this paper.
formalize the integration, processing and analysis of Big EO Data, thus
contributing to the reproducibility of EO science (Giuliani et al., 2019).
Acknowledgement
EODCs provide an architecture that brings together different spatial da-
tabases to facilitate data storage and manipulation. Many initiatives have
The authors would like to thank Dr. Romain Tavenard and Dr. Cen-
emerged this decade, such as Open Data Cube (Lewis et al., 2017), Google
drine Mony for providing suggestions throughout the project. The au-
Earth Engine (Gorelick et al., 2017), JEODPP (Soille et al., 2018) or
thors would also like to thank Editor du jour for checking the English
Rasdaman (Baumann et al., 2019). However, each initiative has its own
version.
architecture, which is not compatible with the others. OpenEO (openEO,
2020) addresses this problem by providing an API (Application Pro-
gramming Interface) that generalizes connections between users and Appendix A. Supplementary data
EODCs. Recently, in addition to these initiatives, code sharing on plat-
forms such as GitHub has contributed greatly to strengthening this syn- Supplementary data to this article can be found online at https://doi.
ergy by contributing to the development of new algorithms and making it org/10.1016/j.srs.2021.100019.
possible to test them on new datasets. Although this practice remains
rare, some applicable contributions to this synergy exist (Carbonneau References
et al., 1002; Tan et al., 2019).
Abdullah, M.M., Gholoum, M.M., Abbas, H.A., 2018. Satellite vs. UAVs remote sensing of
arid ecosystems: a review with in an ecological perspective. EAES 2, 1–5. http://cri
msonpublishers.com/eaes/fulltext/EAES.000540.php. (Accessed 16 October 2019).

11
E. Alvarez-Vanhard et al. Science of Remote Sensing 3 (2021) 100019

Aït Hssaïne, A., Ries, J.B., Peter, K.D., Marzolff, I., d’Oleire-Oltmanns, S., 2011. Firla, M., Mustafa, O., Pfeifer, C., Senf, M., Hese, S., 2019. INTRASEASONAL
Monitoring soil erosion in the souss basin, Morocco, with a multiscale object-based VARIABILITY OF GUANO STAINS IN a REMOTELY SENSED PENGUIN COLONY
remote sensing approach using UAV and satellite data. In: Proceedings of the 1st USING UAV AND SATELLITE. In: ISPRS Annals of the Photogrammetry, Remote
World Sustainability Forum. MDPI, Sciforum.net, p. 562. https://doi.org/10.3390/ Sensing and Spatial Information Sciences. IV-2-W5, pp. 111–118. https://doi.org/
wsf-00562. 10.5194/isprs-annals-IV-2-W5-111-2019.
Alavipanah, S.K., Matinfar, H.R., Rafiei Emam, A., Khodaei, K., Hadji Bagheri, R., Yazdan Forster, M., Schmidt, T., Gonzalez, A.R., Cabezas, J., Fassnacht, F.E., 2018. Application of
Panah, A., 2010. Criteria of selecting satellite data for studying land resources. Desert a one-class classifier and a linear spectral unmixing method for detecting invasive
15, 83–102. https://doi.org/10.22059/jdesert.2011.23005. species in central Chile. In: IGARSS 2018 - 2018 IEEE International Geoscience and
Aleem, K., Lorenzo, C., Biglia, A., Ricauda Aimonino, D., Marcello, C., Gay, P., 2019. Remote Sensing Symposium, pp. 2883–2886. https://doi.org/10.1109/
Comparison of satellite and UAV-based multispectral imagery for vineyard variability IGARSS.2018.8518873.
assessment, 11, pp. 436–452. https://doi.org/10.3390/rs11040436. Fugazza, D., Senese, A., Azzoni, R.S., Smiraglia, C., Cernuschi, M., Severi, D.,
Alvarez-Vanhard, E., Houet, T., Mony, C., Lecoq, L., Corpetti, T., 2020. Can UAVs fill the Diolaiuti, G.A., 2015. High-resolution mapping of glacier surface features. UAV Surv.
gap between in situ surveys and satellites for habitat mapping? Rem. Sens. Environ. Forni Glacier (Stelvio Nat. Park, Italy) 38, 25–33. https://doi.org/10.4461/
243, 111780. https://doi.org/10.1016/j.rse.2020.111780. GFDQ.2015.38.03.
Antoine, R., Lopez, T., Tanguy, M., Lissak, C., Gailler, L., Labazuy, P., Fauchard, C., 2020. Fytsilis, A.L., Prokos, A., Koutroumbas, K.D., Michail, D., Kontoes, C.C., 2016.
Geoscientists in the sky: unmanned aerial vehicles responding to geohazards. Surv. A methodology for near real-time change detection between unmanned aerial vehicle
Geophys. 41, 1285–1321. https://doi.org/10.1007/s10712-020-09611-7. and wide area satellite images. ISPRS J. Photogrammetry Remote Sens. 119,
Atkinson, P.M., 2013. Downscaling in remote sensing. Int. J. Appl. Earth Obs. Geoinf. 22, 165–186. https://doi.org/10.1016/j.isprsjprs.2016.06.001.
106–114. https://doi.org/10.1016/j.jag.2012.04.012. Gamon, J.A., 2015. Reviews and syntheses: optical sampling of the flux tower footprint.
Barrero, Oscar, Perdomo, Sammy A., 2018. RGB and multispectral UAV image fusion for Biogeosciences 12, 4509–4523. https://doi.org/10.5194/bg-12-4509-2015.
Gramineae weed detection in rice fields. Precis. Agric. 19, 809–822. Gao, F., Masek, J., Schwaller, M., Hall, F., 2006. On the blending of the landsat and
Baumann, P., Misev, D., Merticariu, V., Huu, B.P., 2019. Datacubes: towards space/time MODIS surface reflectance: predicting daily landsat surface reflectance. IEEE Trans.
analysis-ready data. In: D€ ollner, J., Jobst, M., Schmitz, P. (Eds.), Service-Oriented Geosci. Rem. Sens. 44, 2207–2218. https://doi.org/10.1109/TGRS.2006.872081.
Mapping: Changing Paradigm in Map Production and Geoinformation Management. Geoss portal, (n.d.). https://www.geoportal.org (accessed July 29, 2020).
Springer International Publishing, Cham, pp. 269–299. https://doi.org/10.1007/ Gevaert, C.M., Tang, J., García-Haro, F.J., Suomalainen, J., Kooistra, L., 2014. Combining
978-3-319-72434-8_14. hyperspectral UAV and multispectral formosat-2 imagery for precision agriculture
Belgiu, M., Stein, A., 2019. Spatiotemporal image fusion in remote sensing. Rem. Sens. applications. In: 2014 6th Workshop on Hyperspectral Image and Signal Processing:
11, 818. https://doi.org/10.3390/rs11070818. Evolution in Remote Sensing. WHISPERS), pp. 1–4. https://doi.org/10.1109/
Bian, J., Li, A., Zhang, Z., Zhao, W., Lei, G., Xia, H., Tan, J., 2016. Grassland fractional WHISPERS.2014.8077607.
vegetation cover monitoring using the composited HJ-1A/b time series images and Gevaert, C., Suomalainen, J.M., Tang, J., Kooistra, L., 2015. Generation of
unmanned aerial vehicles: a case study in zoige wetland, China. In: 2016 IEEE spectral–temporal response surfaces by combining multispectral satellite and
International Geoscience and Remote Sensing Symposium. IGARSS), pp. 7192–7195. hyperspectral UAV imagery for precision agriculture applications. IEEE J. Selected
https://doi.org/10.1109/IGARSS.2016.7730876. Top. Appl. Earth Observ. Rem. Sens. 8, 3140–3146. https://doi.org/10.1109/
Brigot, G., Colin-Koeniguer, E., Plyer, A., Janez, F., 2016. Adaptation and evaluation of an JSTARS.2015.2406339.
optical flow method applied to coregistration of forest remote sensing images. IEEE J. Ghamisi, P., Rasti, B., Yokoya, N., Wang, Q., Hofle, B., Bruzzone, L., Bovolo, F., Chi, M.,
Selected Top. Appl. Earth Observ. Rem. Sens. 9, 2923–2939. https://doi.org/ Anders, K., Gloaguen, R., Atkinson, P.M., Benediktsson, J.A., 2019. Multisource and
10.1109/JSTARS.2016.2578362. multitemporal data fusion in remote sensing: a comprehensive review of the state of
Brinkhoff, J., Hornbuckle, J., Barton, J.L., 2018. Assessment of aquatic weed in irrigation the art. IEEE Geosci. Rem. Sens. Mag. 7, 6–39. https://doi.org/10.1109/
channels using UAV and satellite imagery. Water 10, 1497. https://doi.org/10.3390/ MGRS.2018.2890023.
w10111497. Giuliani, G., Camara, G., Killough, B., Minchin, S., 2019. Earth observation open science:
Briottet, X., Marion, R., Carrere, V., Jacquemoud, S., Chevrel, S., Prastault, P., D’oria, M., enhancing reproducible science using data cubes. Data 4, 147. https://doi.org/
Gilouppe, P., Hosford, S., Lubac, B., Bourguignon, A., HYPXIM, 2011. A new 10.3390/data4040147.
hyperspectral sensor combining science/defence applications. In: 2011 3rd Workshop G
oraj, M., Wr
o
zwiak, J., Chorma

oblewski, C., Cię_zkowski, W., J
nski, J., 2019. Free water
on Hyperspectral Image and Signal Processing: Evolution in Remote Sensing table area monitoring on wetlands using satellite and UAV orthophotomaps -
(WHISPERS). IEEE, Lisbon, Portugal, pp. 1–4. https://doi.org/10.1109/ kampinos national park case study, Meteorology Hydrology and Water Management.
WHISPERS.2011.6080957. Resear. Operat. Appl. 7, 1. http://yadda.icm.edu.pl/baztech/element/bwmeta1.ele
C^
andido, A.K.A.A., Filho, A.C.P., Haupenthal, M.R., da Silva, N.M., de Sousa Correa, J., ment.baztech-91ffa1a5-b842-4884-a89a-173929663130. (Accessed 23 May 2019).
Ribeiro, M.L., 2016. Water quality and chlorophyll measurement through vegetation Gorelick, N., Hancher, M., Dixon, M., Ilyushchenko, S., Thau, D., Moore, R., 2017. Google
indices generated from orbital and suborbital images. Water Air Soil Pollut. 227, 224. earth engine: planetary-scale geospatial analysis for everyone. Rem. Sens. Environ.
https://doi.org/10.1007/s11270-016-2919-7. 202, 18–27. https://doi.org/10.1016/j.rse.2017.06.031.
P.E. Carbonneau, B. Belletti, M. Micotti, B. Lastoria, M. Casaioli, S. Mariani, G. Marchetti, Gruen, A., Zhang, Z., Eisenbeiss, H., 2012. Uav photogrammetry IN remote areas – 3D
S. Bizzi, UAV-based training for fully fuzzy classification of sentinel-2 fluvial scenes, modeling OF DRAPHAM dzong Bhutan. Int. Arch. Photogram. Rem. Sens. 375–379.
Earth Surface Processes and Landforms. n/a (n.d.). https://doi.org/10.1002 https://doi.org/10.5194/isprsarchives-XXXIX-B1-375-2012.
/esp.4955. Hassan-Esfahani, L., Ebtehaj, A.M., Torres-Rua, A., McKee, M., 2017. Spatial scale gap
Casagli, N., Frodella, W., Morelli, S., Tofani, V., Ciampalini, A., Intrieri, E., Raspini, F., filling using an unmanned aerial system: a statistical downscaling method for
Rossi, G., Tanteri, L., Lu, P., 2017. Spaceborne, UAV and ground-based remote applications in precision agriculture. Sensors 17, 2106. https://doi.org/10.3390/
sensing techniques for landslide mapping, monitoring and early warning. s17092106.
Geoenviron. Disasters 4, 9. https://doi.org/10.1186/s40677-017-0073-1. Houborg, R., McCabe, M.F., 2018. A cubesat enabled spatio-temporal enhancement
Colomina, I., Molina, P., 2014. Unmanned aerial systems for photogrammetry and remote method (CESTEM) utilizing planet, landsat and MODIS data. Rem. Sens. Environ.
sensing: a review. ISPRS J. Photogrammetry Remote Sens. 92, 79–97. https:// 209, 211–226. https://doi.org/10.1016/j.rse.2018.02.067.
doi.org/10.1016/j.isprsjprs.2014.02.013. Indigeo, (n.d.) https://portail.indigeo.fr/geocms/projects/d2t (accessed 15 January
Cracknell, A.P., 2017. UAVs: regulations and law enforcement. Int. J. Rem. Sens. 38, 2021.
3054–3067. https://doi.org/10.1080/01431161.2017.1302115. Jacobsen, K., 2012. Airborne or Spaceborne Images for Topographic Mapping?, p. 8.
Ding, Huanyu, Cristofalo, E., Wang, J., Casta~ no
n, D., Montijano, E., Saligrama, V., Jenerowicz, A., Siok, K., Woroszkiewicz, M., Orych, A., 2017. The fusion of satellite and
Schwager, M., 2016. A multi-resolution approach for discovery and 3-d modeling of UAV data: simulation of high spatial resolution band. In: Remote Sensing for
archaeological sites using satellite imagery and a UAV-borne camera. In: 2016 Agriculture, Ecosystems, and Hydrology XIX. International Society for Optics,
American Control Conference. ACC), pp. 1359–1365. https://doi.org/10.1109/ Photonics, p. 104211Z. https://doi.org/10.1117/12.2278669.
ACC.2016.7525107. 
Jurecka, F., Lukas, V., Hlavinka, P., Semer
adov
a, D., Zalud, Z., Trnka, M., 2018.
Elkind, K., Sankey, T.T., Munson, S.M., Aslan, C.E., 2019. Invasive buffelgrass detection Estimating crop yields at the field level using landsat and modis products. Acta Univ.
using high-resolution satellite and UAV imagery on google earth engine. Rem. Sens. Agric. Silvic. Mendelianae Brunensis 66, 1141–1150. https://doi.org/10.11118/
Ecol. Conserv. 5, 318331. https://doi.org/10.1002/rse2.116. actaun201866051141.
European Parliament and of the Council of 27 April 2016, Regulation (EU) 2016/679 on Kakooei, M., Baleghi, Y., 2017. Fusion of satellite, aircraft, and UAV data for automatic
the protection of natural persons with regard to the processing of personal data and disaster damage assessment. Int. J. Rem. Sens. 38, 2511–2534. https://doi.org/
on the free movement of such data, and repealing Directive 95/46/EC (General Data 10.1080/01431161.2017.1294780.
Protection Regulation). http://data.europa.eu/eli/reg/2016/679/oj, 2016–. Kattenborn, T., Lopatin, J., F€ orster, M., Braun, A.C., Fassnacht, F.E., 2019. UAV data as
(Accessed 15 January 2021). alternative to field sampling to map woody invasive species based on combined
European parliament and the council of 14 march 2007, directive 2007/2/EC establishing sentinel-1 and sentinel-2 data. Rem. Sens. Environ. 227, 61–73. https://doi.org/
an infrastructure for spatial information in the European community (INSPIRE). htt 10.1016/j.rse.2019.03.025.
p://data.europa.eu/eli/dir/2007/2/oj, 2007–. (Accessed 15 January 2021). Kuenzer, C., Dech, S., Wagner, W., 2015. Remote sensing time series revealing land
Fawcett, D., Panigada, C., Tagliabue, G., Boschetti, M., Celesti, M., Evdokimov, A., surface dynamics: status quo and the pathway ahead. In: Kuenzer, C., Dech, S.,
Biriukova, K., Colombo, R., Miglietta, F., Rascher, U., Anderson, K., 2020. Multi-scale Wagner, W. (Eds.), Remote Sensing Time Series: Revealing Land Surface Dynamics.
evaluation of drone-based multispectral surface reflectance and vegetation indices in Springer International Publishing, Cham, pp. 1–24. https://doi.org/10.1007/978-3-
operational conditions. Rem. Sens. 12, 514. https://doi.org/10.3390/rs12030514. 319-15967-6_1.
Fazle, K., Majumdar, S., Darabi, H., Chen, S., 2017. LSTM fully convolutional networks for Lewis, A., Oliver, S., Lymburner, L., Evans, B., Wyborn, L., Mueller, N., Raevksi, G.,
time series classification. IEEE Access 6, 1662–1669. Hooke, J., Woodcock, R., Sixsmith, J., Wu, W., Tan, P., Li, F., Killough, B.,
Minchin, S., Roberts, D., Ayers, D., Bala, B., Dwyer, J., Dekker, A., Dhu, T., Hicks, A.,

12
E. Alvarez-Vanhard et al. Science of Remote Sensing 3 (2021) 100019

Ip, A., Purss, M., Richards, C., Sagar, S., Trenham, C., Wang, P., Wang, L.-W., 2017. Rau, J.-Y., Jhan, J.-P., Rau, R.-J., 2014. Semiautomatic object-oriented landslide
The Australian geoscience data Cube — foundations and lessons learned. Rem. Sens. recognition scheme from multisensor optical imagery and DEM. IEEE Trans. Geosci.
Environ. 202, 276–292. https://doi.org/10.1016/j.rse.2017.03.015. Rem. Sens. 52, 1336–1349. https://doi.org/10.1109/TGRS.2013.2250293.
Liang, H., Huang, X., Sun, Y., Wang, Y., Liang, T., 2017. Fractional snow-cover mapping Riihim€aki, H., Luoto, M., Heiskanen, J., 2019. Estimating fractional cover of tundra
based on MODIS and UAV data over the Tibetan plateau. Rem. Sens. 9, 1332. https:// vegetation at multiple scales using unmanned aerial systems and optical satellite
doi.org/10.3390/rs9121332. data. Rem. Sens. Environ. 224, 119–132. https://doi.org/10.1016/j.rse.2019.01.030.
Lin, A.Y., Novo, A., Har-Noy, S., Ricklin, N.D., Stamatiou, K., 2011. Combining GeoEye-1 Roosjen, P.P.J., Suomalainen, J.M., Bartholomeus, H.M., Kooistra, L., Clevers, J.G.P.W.,
satellite remote sensing, UAV aerial imaging, and geophysical surveys in anomaly 2017. Mapping reflectance anisotropy of a potato canopy using aerial images
detection applied to archaeology. IEEE J. Selected Top. Appl. Earth Observ. Rem. acquired with an unmanned aerial vehicle. Rem. Sens. 9, 417. https://doi.org/
Sens. 4, 870–876. https://doi.org/10.1109/JSTARS.2011.2143696. 10.3390/rs9050417.
Liu, P., Di, L., Du, Q., Wang, L., 2018. Remote sensing big data: theory, methods and Rupasinghe, P.A., Milas, A.S., Arend, K., Simonson, M.A., Mayer, C., Mackey, S., 2019.
applications. Rem. Sens. 10, 711. https://doi.org/10.3390/rs10050711. Classification of shoreline vegetation in the western basin of lake erie using airborne
Liu, H., Dahlgren, R.A., Larsen, R.E., Devine, S.M., Roche, L.M., O’ Geen, A.T., hyperspectral imager HSI2, pleiades and UAV data. Int. J. Rem. Sens. 40, 3008–3028.
Wong, A.J.Y., Covello, S., Jin, Y., 2019a. Estimating rangeland forage production https://doi.org/10.1080/01431161.2018.1539267.
using remote sensing data from a small unmanned aerial system (sUAS) and Ruwaimana, M., Satyanarayana, B., Otero, V., Muslim, A.M., A, M.S., Ibrahim, S.,
PlanetScope satellite. Rem. Sens. 11, 595. https://doi.org/10.3390/rs11050595. Raymaekers, D., Koedam, N., Dahdouh-Guebas, F., 2018. The advantages of using
Liu, Wei, et al., 2019b. Accurate building extraction from fused DSM and UAV images drones over space-borne imagery in the mapping of mangrove forests. PloS One 13,
using a chain fully convolutional neural network. Rem. Sens. 11, 2912. e0200288. https://doi.org/10.1371/journal.pone.0200288.
Long, J., Shelhamer, E., Darrell, T., 2015. Fully convolutional networks for semantic Schaepman-Strub, G., Schaepman, M.E., Painter, T.H., Dangel, S., Martonchik, J.V., 2006.
segmentation. In: Proceedings of the IEEE Conference on Computer Vision and Reflectance quantities in optical remote sensing—definitions and case studies. Rem.
Pattern Recognition, pp. 3431–3440. Sens. Environ. 103, 27–42. https://doi.org/10.1016/j.rse.2006.03.002.
Martínez-S
anchez, J., Gonz
alez-de Santos, L.M., Novo, A., Gonz
alez-Jorge, H., 2019. Uav Soille, P., Burger, A., De Marchi, D., Kempeneers, P., Rodriguez, D., Syrris, V., Vasilev, V.,
and satellite imagery applied to alien species mapping IN NW Spain. In: ISPRS - 2018. A versatile data-intensive computing platform for information retrieval from
International Archives of the Photogrammetry, Remote Sensing and Spatial big geospatial data. Future Generat. Comput. Syst. 81, 30–40. https://doi.org/
Information Sciences. Copernicus GmbH, pp. 455–459. https://doi.org/10.5194/ 10.1016/j.future.2017.11.007.
isprs-archives-XLII-2-W13-455-2019. Solazzo, D., Sankey, J.B., Sankey, T.T., Munson, S.M., 2018. Mapping and measuring
Martonchik, J.V., 1994. Retrieval of surface directional reflectance properties using aeolian sand dunes with photogrammetry and LiDAR from unmanned aerial vehicles
ground level multiangle measurements. Rem. Sens. Environ. 50, 303–316. https:// (UAV) and multispectral satellite imagery on the paria plateau. AZ, USA,
doi.org/10.1016/0034-4257(94)90080-9. Geomorphol. 319, 174–185. https://doi.org/10.1016/j.geomorph.2018.07.023.
Marx, A., McFarlane, D., Alzahrani, A., 2017. UAV data for multi-temporal landsat SpatioTemporal asset catalog, (n.d.). https://stacspec.org/(accessed July 29, 2020).
analysis of historic reforestation: a case study in Costa Rica. Int. J. Rem. Sens. 38, Spence, C., Mengistu, S., 2016. Deployment of an unmanned aerial system to assist in
2331–2348. https://doi.org/10.1080/01431161.2017.1280637. mapping an intermittent stream. Hydrol. Process. 30, 493–500. https://doi.org/
Matese, A., Toscano, P., Di Gennaro, S.F., Genesio, L., Vaccari, F.P., Primicerio, J., 10.1002/hyp.10597.
Belli, C., Zaldei, A., Bianconi, R., Gioli, B., 2015. Intercomparison of UAV, aircraft and St-Onge, B., Grandin, S., 2019. Estimating the height and basal area at individual tree and
satellite remote sensing platforms for precision viticulture. Rem. Sens. 7, 2971–2990. plot levels in canadian subarctic lichen woodlands using stereo WorldView-3 images.
https://doi.org/10.3390/rs70302971. Rem. Sens. 11, 248. https://doi.org/10.3390/rs11030248.
Matouskov
a, E., Starkov

a, L., Pavelka, K., Nov
acek, K., Sedina, J., Faltýnov
a, M., Stark, B., Zhao, T., Chen, Y., 2016. An analysis of the effect of the bidirectional reflectance
Housarov
a, E., 2016. Using remotely sensed data for documentation OF distribution function on remote sensing imagery accuracy from small unmanned
archaeological sites IN northeastern mesopotamia. In: ISPRS - International Archives aircraft systems. In: 2016 International Conference on Unmanned Aircraft Systems.
of the Photogrammetry, Remote Sensing and Spatial Information Sciences. ICUAS), pp. 1342–1350. https://doi.org/10.1109/ICUAS.2016.7502566.
Copernicus GmbH, pp. 335–342. https://doi.org/10.5194/isprs-archives-XLI-B5-335- Stow, D., Nichol, C.J., Wade, T., Assmann, J.J., Simpson, G., Helfter, C., 2019.
2016. Illumination geometry and flying height influence surface reflectance and NDVI
Melville, B., Fisher, A., Lucieer, A., 2019. Ultra-high spatial resolution fractional derived from multispectral UAS imagery. Drones 3, 55. https://doi.org/10.3390/
vegetation cover from unmanned aerial multispectral imagery. Int. J. Appl. Earth drones3030055.
Obs. Geoinf. 78, 14–24. https://doi.org/10.1016/j.jag.2019.01.013. Su, L., Chopping, M.J., Rango, A., Martonchik, J.V., Peters, D.P.C., 2007. Differentiation
Mengmeng, D., Noboru, N., Atsushi, I., Yukinori, S., 2017. Multi-temporal monitoring of of semi-arid vegetation types based on multi-angular observations from MISR and
wheat growth by using images from satellite and unmanned aerial vehicle. Int. J. MODIS. Int. J. Rem. Sens. 28, 1419–1424. https://doi.org/10.1080/
Agric. Biol. Eng. 10, 1–13. https://doi.org/10.25165/ijabe.v10i5.3180. 01431160601085995.
Milas, A.S., Cracknell, A.P., Warner, T.A., 2018. Drones – the third generation source of Sun, A.Y., Scanlon, B.R., 2019. How can big data and machine learning benefit
remote sensing data. Int. J. Rem. Sens. 39, 7125–7137. https://doi.org/10.1080/ environment and water management: a survey of methods, applications, and future
01431161.2018.1523832. directions. Environ. Res. Lett. 14, 073001 https://doi.org/10.1088/1748-9326/
Müllerov
a, J., Brůna, J., Dvor
ak, P., Bartalos, T., Vítkov
a, M., 2016. Does the data ab1b7d.
resolution/origin matter? SATELLITE, airborne and UAV imagery to tackle plant Szantoi, Z., Smith, S.E., Strona, G., Koh, L.P., Wich, S.A., 2017. Mapping orangutan
invasions. Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci. XLI-B7 903–908. habitat and agricultural areas using landsat OLI imagery augmented with unmanned
https://doi.org/10.5194/isprsarchives-XLI-B7-903-2016. aircraft system aerial photography. Int. J. Rem. Sens. 38, 2231–2245. https://
Müllerov
a, J., Brůna, J., Bartalos, T., Dvor
ak, P., Vítkov
a, M., Pysek, P., 2017. Timing is doi.org/10.1080/01431161.2017.1280638.
important: unmanned aircraft vs. Satellite imagery in plant invasion monitoring. Tan, Z., Di, L., Zhang, M., Guo, L., Gao, M., 2019. An enhanced deep convolutional model
Front. Plant Sci. 8 https://doi.org/10.3389/fpls.2017.00887. for spatiotemporal image fusion. Rem. Sens. 11, 2898. https://doi.org/10.3390/
Nhamo, L., Van Dijk, R., Magidi, J., Wiberg, D., Tshikolomo, K., 2018. Improving the rs11242898.
accuracy of remotely sensed irrigated areas using post-classification enhancement Transon, J., D’Andrimont, R., Maugnard, A., Defourny, P., 2018. Survey of hyperspectral
through UAV capability. Rem. Sens. 10, 712. https://doi.org/10.3390/rs10050712. earth observation applications from space in the sentinel-2 context. Rem. Sens. 10,
Nikolakopoulos, K., Kyriou, A., Koukouvelas, I., Zygouri, V., Apostolopoulos, D., 2019. 157. https://doi.org/10.3390/rs10020157.
Combination of aerial, satellite, and UAV photogrammetry for mapping the Tschudi, M.A., Maslanik, J.A., Perovich, D.K., 2008. Derivation of melt pond coverage on
diachronic coastline evolution: the case of lefkada island. ISPRS Int. J. Geo-Inf. 8, arctic sea ice using MODIS observations. Rem. Sens. Environ. 112, 2605–2614.
489. https://doi.org/10.3390/ijgi8110489. https://doi.org/10.1016/j.rse.2007.12.009.
Oh, J., Toth, C.K., Grejner-Brzezinska, D.A., 2011. Automatic georeferencing of aerial Vihervaara, P., Auvinen, A.-P., Mononen, L., T€ orm€a, M., Ahlroth, P., Anttila, S.,
images using stereo high-resolution satellite images. Photogramm. Eng. Rem. Sens. B€ottcher, K., Forsius, M., Heino, J., Heli€
ol€a, J., Koskelainen, M., Kuussaari, M.,
77, 1157–1168. https://doi.org/10.14358/PERS.77.11.1157. Meissner, K., Ojala, O., Tuominen, S., Viitasalo, M., Virkkala, R., 2017. How essential
OpenAerialMap. OpenAerialMap, (n.d.). http://openaerialmap.org/(accessed July 29, biodiversity variables and remote sensing can help national biodiversity monitoring.
2020). Glob. Ecol. Conserv. 10, 43–59. https://doi.org/10.1016/j.gecco.2017.01.007.
openEO, (n.d.). https://openeo.org/(accessed July 29, 2020). Wang, R., Zhang, S., Pu, L., Yang, J., Yang, C., Chen, J., Guan, C., Wang, Q., Chen, D.,
Padr
o, J.-C., Mu~

noz, F.-J., Avila,
Pesquer, L., Pons, X., 2018. Radiometric correction
L.A., Fu, B., Sang, X., 2016. Gully erosion mapping and monitoring at multiple scales based
of landsat-8 and sentinel-2A scenes using drone imagery in synergy with field on multi-source remote sensing data of the sancha river catchment, northeast China.
spectroradiometry. Rem. Sens. 10, 1687. https://doi.org/10.3390/rs10111687. ISPRS Int. J. Geo-Inf. 5, 200. https://doi.org/10.3390/ijgi5110200.
Papakonstantinou, A., Doukari, M., Stamatis, P., Topouzelis, K., 2019. Coastal Wilkinson, M.D., Dumontier, M., Aalbersberg, I.J., Appleton, G., Axton, M., Baak, A.,
management using UAS and high-resolution satellite images for touristic areas. Int. J. Blomberg, N., Boiten, J.-W., da Silva Santos, L.B., Bourne, P.E., Bouwman, J.,
Appl. Geospatial Res. (IJAGR) 10, 54–72. https://doi.org/10.4018/ Brookes, A.J., Clark, T., Crosas, M., Dillo, I., Dumon, O., Edmunds, S., Evelo, C.T.,
IJAGR.2019010103. Finkers, R., Gonzalez-Beltran, A., Gray, A.J.G., Groth, P., Goble, C., Grethe, J.S.,
Pohl, C., Genderen, J.L.V., 1998. Review article multisensor image fusion in remote Heringa, J., ’t Hoen, P.A.C., Hooft, R., Kuhn, T., Kok, R., Kok, J., Lusher, S.J.,
sensing: concepts, methods and applications. Int. J. Rem. Sens. 19, 823–854. https:// Martone, M.E., Mons, A., Packer, A.L., Persson, B., Rocca-Serra, P., Roos, M., van
doi.org/10.1080/014311698215748. Schaik, R., Sansone, S.-A., Schultes, E., Sengstag, T., Slater, T., Strawn, G.,
Qi, C.R., Su, H., Mo, K., Guibas, L.J., 2017. Pointnet: Deep learning on point sets for 3d Swertz, M.A., Thompson, M., van der Lei, J., van Mulligen, E., Velterop, J.,
classification and segmentation. In: Proceedings of the IEEE Conference on Computer Waagmeester, A., Wittenburg, P., Wolstencroft, K., Zhao, J., Mons, B., 2016. The FAIR
Vision and Pattern Recognition, pp. 652–660. guiding principles for scientific data management and stewardship. Sci. Data 3,
160018. https://doi.org/10.1038/sdata.2016.18.

13
E. Alvarez-Vanhard et al. Science of Remote Sensing 3 (2021) 100019

Xia, H., Zhao, W., Li, A., Bian, J., Zhang, Z., 2017. Subpixel inundation mapping using Zhang, S., Zhao, G., Lang, K., Su, B., Chen, X., Xi, X., Zhang, H., 2019. Integrated satellite,
landsat-8 OLI and UAV data for a wetland region on the zoige plateau, China. Rem. unmanned aerial vehicle (UAV) and ground inversion of the SPAD of winter wheat in
Sens. 9, 31. https://doi.org/10.3390/rs9010031. the reviving stage. Sensors 19, 1485. https://doi.org/10.3390/s19071485.
Yan, Le Cun, Yoshua, B., Geoffrey, H., 2015. Deep learning. Nature 521, 436–444. Zhao, Y., Chen, N., Chen, J., Hu, C., 2018. Automatic extraction of yardangs using landsat
Yue, L., Shen, H., Li, J., Yuan, Q., Zhang, H., Zhang, L., 2016. Image super-resolution: the 8 and UAV images: a case study in the qaidam basin, China. Aeolian Resear. 33,
techniques, applications, and future. Signal Process. 128, 389–408. https://doi.org/ 53–61. https://doi.org/10.1016/j.aeolia.2018.05.002.
10.1016/j.sigpro.2016.05.002. Zhenfeng, S., Cai, J., 2018. Remote sensing image fusion with deep convolutional neural
Zhang, J., 2010. Multi-source remote sensing data fusion: status and trends. Int. J. Image network. IEEE J. Selected Top. Appl. Earth Observ. Rem. Sens. 11, 1656–1669.
Data Fusion 1, 5–24. https://doi.org/10.1080/19479830903561035. Zhu, X., Cai, F., Tian, J., Williams, T.K.-A., 2018. Spatiotemporal fusion of multisource
remote sensing data: literature survey, taxonomy, principles, applications, and future
directions. Rem. Sens. 10, 527. https://doi.org/10.3390/rs10040527.

14