Content uploaded by Hyun-Goo Kim
Author content
All content in this area was uploaded by Hyun-Goo Kim on Aug 29, 2021
Content may be subject to copyright.
Remote Sens. 2021, 13, 3422. https://doi.org/10.3390/rs13173422 www.mdpi.com/journal/remotesensing
Article
Solar Resource Potentials and Annual Capacity Factor Based on
the Korean Solar Irradiance Datasets Derived by the Satellite
Imagery from 1996 to 2019
Chang-Ki Kim, Hyun-Goo Kim *, Yong-Heack Kang, Chang-Yeol Yun, Boyoung Kim and Jin-Young Kim
New and Renewable Energy Map Laboratory, Korea Institute of Energy Research, Daejeon 34129, Korea;
ckkim@kier.re.kr (C.-K.K.); yhkang@kier.re.kr (Y.-H.K.); yuncy@kier.re.kr (C.-Y.Y.); bykim@kier.re.kr (B.K.);
jinyoung.kim@kier.re.kr (J.-Y.K.)
* Correspondence: hyungoo@kier.re.kr
Abstract: The Korea Institute of Energy Research builds Korean solar irradiance datasets, using
gridded solar insolation estimates derived using the University of Arizona solar irradiance based
on Satellite–Korea Institute of Energy Research (UASIBS–KIER) model, with the incorporation of
geostationary satellites over the Korean Peninsula, from 1996 to 2019. During the investigation pe-
riod, the monthly mean of daily total irradiance was in a good agreement with the in situ measure-
ments at 18 ground stations; the mean absolute error is also normalized to 9.4%. It is observed that
the irradiance estimates in the datasets have been gradually increasing at a rate of 0.019 kWh m
−
2
d
−
1
per year. The monthly variation in solar irradiance indicates that the meteorological conditions
in the spring season dominate the annual solar insolation. In addition, the local distribution of solar
irradiance is primarily affected by the geographical environment; higher solar insolation is observed
in the southern part of Korea, but lower solar insolation is observed in the mountainous range in
Korea. The annual capacity factor is the secondary output from the Korean solar irradiance datasets.
The reliability of the estimate of this factor is proven by the high correlation coefficient of 0.912.
Thus, in accordance with the results from the spatial distribution of solar irradiance, the southern
part of Korea is an appropriate region for establishing solar power plants exhibiting a higher annual
capacity factor than the other regions.
Keywords: the Korean solar irradiance dataset; satellite-derived solar irradiance; solar energy re-
source assessment; photovoltaic system; capacity factor
1. Introduction
The long-term variability in solar resources in space and time has been considered as
a crucial factor in the life cycle of solar power systems, including installation, deployment,
and operation (e.g., [1–6]). The analysis of long-term in situ measurements by a pyranom-
eter can help to understand the climatological variability in solar irradiance, due to the
higher accuracy and reliability of the observations compared to remote sensing tech-
niques ([7,8]). Ground observations, however, are still limited in their ability to retrieve
the spatial distribution of solar insolation and further investigate solar energy potentials
at a regional or national scale. Consequently, remote sensing techniques have been con-
sidered as an alternative approach for understanding the climatological characteristics of
solar insolation (e.g., [9–13]). The International Satellite Cloud Climatology Project (IS-
CCP), initiated in July 1983, is an international collaborative program set up to determine
the parameters of cloud microphysics, as well as downwelling surface shortwave radia-
tion ([9,10]). The National Renewable Energy Laboratory (NREL) has developed the na-
tional solar radiation database (NSRDB) that includes meteorological elements, as well as
solar irradiance data, over the United States for the last two decades [12]. The current
Citation: Kim, C.-K.; Kim, H.G.;
Kang, Y.-H.; Yun, C.-Y.; Kim, B.;
Kim, J.-Y. Solar Resource Potentials
and Annual Capacity Factor Based
on the Korean Solar Irradiance
Datasets Derived by the Satellite
Imagery from 1996 to 2019.. Remote
Sens. 2021, 13, 3422.
https://doi.org/10.3390/rs13173422
Academic Editor: Xuan Zhu
Received: 21 July 2021
Accepted: 27 August 2021
Published: 28 August 2021
Publisher’s Note: MDPI stays neu-
tral with regard to jurisdictional
claims in published maps and institu-
tional affiliations.
Copyright: © 2021 by the authors.
Submitted for possible open access
publication under the terms and
conditions of the Creative Commons
Attribution (CC BY) license
(https://creativecommons.org/license
s/by/4.0/).
Remote Sens. 2021, 13, 3422 2 of 22
version of the NSRDB contains gridded datasets at 4 km resolution, which has been de-
veloped using the physical solar model (PSM) with geostationary operational environ-
mental satellite (GOES) data. Thus, there are several ways to understand the long-term
variability in solar irradiance based on satellite imagery.
The empirical parameterization between satellite imagery and ground observation is
the simplest way to derive solar irradiance ([14,15]). However, Noia et al. [16] pointed out
that the relationship between visible reflectance at the top of the atmosphere and down-
welling shortwave radiation at the surface is significantly affected by the meteorological
conditions at the local and regional scales; therefore, the relationship cannot be general-
ized for building a solar resource map. Contrary to the empirical parameterization, the
direct solution of the radiative transfer model is the most reliable method for estimating
solar irradiance. However, the computation cost is too high to derive the solar irradiance
in real time. To resolve the numerical burden of deriving solar irradiance with high relia-
bility, a lookup table technique has been widely employed in remote sensing (e.g., [17–
23]). For example, Pinker et al. [21] derived the solar irradiance over the continental U.S.,
utilizing the University of Maryland shortwave radiation budget (UMD/SRB) model that
implements GOES-8 data. More recently, Kim et al. [24] estimated the downwelling sur-
face shortwave radiation over the southwestern U.S., including Arizona, Western New
Mexico, Southern California, and Southern Nevada, using the visible reflectance and
brightness temperatures from the GOES-15 that can be monitored in the atmospheric state
at 135oW for 24 h each day.
In addition to studies that derived solar irradiance over the United States, several
attempts have been made to derive solar irradiance over Latin America, Africa, Europe,
Australia, and Asia, using geostationary satellites. Deneke et al. [25] estimated the solar
irradiance over the Netherlands, using METEOSAT data. Furthermore, their estimates
were extended towards solar resource assessment in Benelux countries, made by Journée
et al. [26]. In Greece, a series of studies have been conducted to estimate solar resources
based on METEOSAT imagery ([27,28]). Geostationary satellites, operated by the Japanese
Meteorological Agency (JMA), have historically been employed to assess solar irradiance
over Asia and Australia. The geostationary meteorological satellite (GMS) series is a pio-
neer in this region, for the development of meteorological satellites. Tanahashi et al. [29]
presented an improved algorithm for estimating the hourly mean insolation over Japan
for GMS data. The solar resource map of developing countries in Southeast Asia was de-
veloped by Janjai et al. [30], who resolved the physical radiation model using GMS im-
agery. Lu et al. [31] created a look-up table by incorporating a spectral radiative transfer
model, called SBDART, and then retrieved the solar irradiance by comparing the surface
reflectance and atmospheric parameters. The multifunction transport satellite (MTSAT) is
the next-generation satellite on the GMS platform. The spectral bands of the remote sensor
onboard the MTSAT were significantly improved. Consecutive studies have been per-
formed, as new satellites with embedded instruments have become technically progres-
sive (e.g., [32–34]).
Yeom et al. [35] implemented a simple radiative transfer model to generate solar ir-
radiance over the Korean Peninsula, using the communication, ocean, and meteorological
satellite (COMS) geostationary satellite, operated by the Korea Meteorological Admin-
istration (KMA) [35]. Since then, several attempts have been made to raise the accuracy of
satellite estimates (e.g., [36–38]). The Korea Institute of Energy Research (KIER) has devel-
oped the University of Arizona Solar Irradiance Based on Satellite (UASIBS)–KIER model
to assimilate COMS imagery [39]. During the evaluation of the global horizontal irradi-
ance (GHI) at the instantaneous time scale, the UASIBS–KIER model yielded relative root
mean square errors of 7.4% and 16.7% on clear and cloudy skies, respectively. In Decem-
ber 2018, KMA launched the second generation of the COMS, which is called Geo–
KOMPSAT 2A (GK–2A); therefore, the UASIBS–KIER model has also been modified to
employ the GK–2A level-1B data. The modeling performance has increased with im-
proved spectral bands and spatial resolution [40]. However, though innumerable efforts
Remote Sens. 2021, 13, 3422 3 of 22
have been made to derive solar irradiance in Korea, there are no reliable freely available
long-term datasets of solar irradiance. The KIER has completed the Korean solar irradi-
ance datasets at 1 km × 1 km grid cells from January 2012 to July 2020, and now they are
available by the public data portal in Korea [41]. Therefore, the present study aims to in-
troduce the derivation process of solar irradiance from satellite imagery and further build
a solar resource map over the Korean Peninsula for the whole year, from 1996 to 2019. The
remainder of this article is organized as follows: Section 2 explains the outline of the re-
search data, including satellite and research area. Section 3 briefly describes the UASIBS–
KIER model. Subsequently, the evaluation of the data is presented in Section 4. Further
analysis of the solar resource potential and renewable energy system is discussed in Sec-
tion 5. Finally, Section 6 summarizes the major findings of this study.
2. Data and Research Area
2.1. Research Area and Design
The research area is limited to the Korean Peninsula (32oN–40oN, 124oE–130oE) (636
km × 459 km) and a part of the Japanese Islands (Figure. 1). The investigation period was
from 1996 to 2019, except for 2006, in which satellite imagery was not available for public
purposes. Initially, the modeled GHI estimates at the instantaneous time scale are inter-
polated every 5 min. Further, they are all integrated over the interval from 7 Korean Stand-
ard Time (KST = universal time coordinate + 9 hours) to 20 KST, to derive the daily total
irradiance. For inter-comparison between the estimations and observations, the daily total
irradiance was averaged over each month.
Figure 1. Map of the research area and 17 administrative divisions in Korea (a) and the topography and the locations of
the ground observations (b). The number in (b) indicates the station index in Table 1.
Table 1. Summary of the details for the ground observation stations.
Station Index Station ID Station Name Latitude (oN) Longitude (oE) Information
1 100 Daegwallyeong 37.6771 128.7183 Mountain
2 101 Chuncheon 37.9026 127.7357 Rural
3 105 Gangneung 37.7515 128.8910 Coast
4 108 Seoul 37.5714 126.9658 City
Remote Sens. 2021, 13, 3422 4 of 22
5 112 Incheon 37.4777 126.6240 City
6 114 Wonju 37.3376 127.9466 Rural
7 119 Suwon 37.2723 126.9853 City
8 129 Seosan 36.7766 126.4939 Coast
9 131 Cheongju 36.6392 127.4407 City
10 133 Daejeon 36.372 127.3721 City
11 135
Chu-
pungnyeong 36.2202 127.9946 Mountain
12 136 Andong 36.5729 128.7073 Rural
13 143 Daegu 35.8280 128.6522 City
14 146 Jeonju 35.8408 127.1190 City
15 156 Gwangju 35.1729 126.8916 City
16 159 Busan 35.1047 129.0320 City
17 165 Mokpo 34.8169 126.3812 Coas
t
18 184 Jeju 33.5141 126.5297 Island
2.2. Satellite Imagery
Six satellite platforms have been operated to monitor the atmosphere over the Korean
Peninsula since 1996, they are as follows: GMS-5, GOES-9, MTSAT-1R/2, COMS, and GK-
2A. The details of the five satellites with embedded sensors are summarized in Table 2.
GMS-5, operated by JMA, monitored the atmospheric status, implementing a visible and
infrared spin scan radiometer (VISSR) at four different spectral ranges. When compared
with the VISSR onboard GMS-5, the GOES-9 Imager contains five channels, including a
shortwave infrared channel. Unfortunately, the brightness temperature or counts at 3.8
μm were not available in the archive system of Kochi University [42]. The spatial resolu-
tion of the visible and infrared channels of the VISSR and Imager were originally different,
for example, 1.25 km and 5 km for the visible and infrared channels, respectively. The
available datasets were reshaped into 0.05o × 0.05o grid cells for all the spectral bands. The
MTSAT-1R/2 system is equipped with the Japanese advanced meteorological imager
(JAMI), which collects the imagery at five spectral bands. The level-1B data of JAMI are
available at the National Meteorological Satellite Center (NMSC), as part of the KMA, but
the spatial resolution is 4 km for all the spectral bands. The specification of the meteoro-
logical imager onboard the COMS is almost the same as that of MTSAT-1R/2. NMSC pro-
vides the COMS level-1B data at five different channels, as follows: visible (0.67 μm),
shortwave infrared (3.7 μm), water vapor (6.7 μm), and two split-infrared (10.8 μm and
12.0 μm) channels. The horizontal resolutions of the visible and infrared channels were 1
km and 4 km, respectively. The temporal resolution is coarser in GMS-5 and GOES-9,
whereas the COMS facilitates the UASIBS–KIER model to produce GHI estimates every
15 min.
2.3. In situ Observation
The present study uses in situ measurements from 18 automatic synoptic observing
stations (ASOSs), operated by the KMA, to evaluate satellite-derived solar irradiance. The
geographic details are listed in Table 1, which introduces GHI measured by a pyranometer
at all stations. All the KMA stations provide hourly accumulated GHI measurements (MJ
m−2) that are integrated from sunrise to sunset, to finally obtain the daily total irradiance
(kWh m−2 d−1) for comprehensive analysis. The total number of data for the individual
stations is 4860 (= 22.5 years × 12 months × 18 stations). In addition to solar irradiance,
photovoltaic (PV) power generation is employed to determine the application of satellite-
derived solar irradiance to renewable energy resource assessment. Data are available in
the electric power statistics information system (EPSIS), which is operated by the Korea
Power Exchange (KPX). EPSIS data are monthly statistics, such as the monthly total power
or capacity for each administrative division in Korea. Therefore, for the evaluation against
Remote Sens. 2021, 13, 3422 5 of 22
EPSIS data, the modeling output from the UASIBS–KIER model must be integrated for
each month.
Table 2. Summary of geostationary satellite employed in the UASIBS–KIER model from 1996 to
2019.
Satellite GMS—5 GOES—9 MTSAT—1R MTSAT—2 COMS
Data Availa-
bility
1996.01~2003.0
6
2003.07~2005.0
6
2007.01~2009.1
2
2010.01~20111.
12
2012.01~2019.1
2
Instrument VISSR Imager JAMI JAMI MI
Spectral Bands
(μm)
0.5–1.05
10.5–11.5
11.5–12.5
6.5–7.0
0.55–0.75
3.8–4.0
10.5–11.5
11.5–12.5
6.5–7.0
0.55–0.90
3.5–4.0
10.3–11.3
11.5–12.5
6.5–7.0
0.55–0.90
3.5–4.0
10.3–11.3
11.5–12.5
6.5–7.0
0.55–0.90
3.5–4.0
10.3–11.3
11.5–12.5
6.5–7.0
Spatial Reso-
lution 0.05o 0.05o 4 km 4 km 1 km
Center Longi-
tude 140oE 155 oE 140 oE 145 oE 128.2 oE
2.4. ECMWF Reanalysis 5 Land Dataset
This study implements the ECMWF reanalysis 5 (ERA-5) land dataset for the com-
prehensive assessment of the UASIBS–KIER model and in situ observations. ERA-5 is pro-
duced by the Copernicus Climate Change Service as a part of ECMWF, and it provides
hourly estimates of soil parameters, as well as land surface forcing data, such as solar
insolation [43]. Their spatial and temporal resolutions were 9 km and 1 h, respectively.
The data are available from 1980 to present for public use. Out of all the land surface data,
downwelling surface shortwave radiation is employed for this study.
3. UASIBS–KIER Model
The full description of the UASIBS–KIER model is provided in the research series by
Kim et al. [40] [24,39]; therefore, we briefly introduce the model here. This model uses a
look-up table approach to reduce the numerical burden of resolving the radiative transfer
equation at each pixel ([18,21,44]). A look-up table, generated by the Goddard Space Flight
Center radiative transfer model (GSFC RTM) [45], represents the shortwave albedo at the
top of the atmosphere (TOA) as a function of the solar zenith angle, surface albedo, ozone,
water vapor, aerosol optical depth (AOD), and cloud optical depth (COD), for the four
cloud classes, i.e., high-, mid-, and low-level clouds, and cumulus clouds. The monthly
average ozone concentration was obtained from the climatological background datasets
summarized by Tilmes et al. [46]. The vertically integrated AOD measured at Yonsei Uni-
versity, as a part of the AERONET station, is redistributed at each vertical level, to param-
eterize the aerosol extinction coefficient profile in the GSFC RTM using an exponential
distribution [47]. Since the observation of AOD at Yonsei University, initiated in March
2011, the monthly mean AOD at 440, 500, and 870 nm, from 2011 to 2019, was employed
to generate the AOD profiles for the investigation period before 2011. The water vapor
profile was obtained from a rawinsonde that was launched at Suwon/Osan station (Figure
1) at 1200 UTC (2100 KST = UTC + 9) on the previous day.
After creating the look-up table, the cloudy pixels were determined using the bright-
ness temperatures and visible reflectance ([21,48]). As the horizontal resolution and cen-
tral wavelength vary between different remote sensors onboard satellites, cloud detection
procedures are also implemented between satellite imagery. When the UASIBS–KIER
model derives solar irradiance using GMS-5 and GOES-9 satellite imagery, the brightness
Remote Sens. 2021, 13, 3422 6 of 22
temperature at 3.7 μm is missing; therefore, the brightness temperature difference pre-
sented by Jedlovec et al. [48] cannot be used to detect the cloudy pixels. Without the
brightness temperature at 3.7 μm, we simply distinguish the cloudy pixels from the clear
ones through the joint classification table presented by Rossow and Garder [49], who car-
ried out space and time contrast tests. As the spectral coverage in JAMI onboard MTSAT-
1R/2 includes the wavelength range from 3.5 μm to 4.0 μm, the brightness temperature
difference results in the detection of cloudy pixels, in a manner similar to the UASIBS–
KIER model with the COMS satellite [39]. Furthermore, visible reflectance is employed to
detect optically thin clouds, i.e., cirrus or shallow clouds. Clear-sky composite shortwave
albedo was generated by the minimum value of visible reflectance for the previous 15
days. Compared with the clear-sky composite albedo, the observed shortwave TOA al-
bedo classifies pixels into either clear or cloudy pixels.
The horizontal resolution between infrared and visible channels is exactly similar to
all satellite imagery, except the COMS imagery. Cloud detection was first carried out us-
ing brightness temperature for a 4 km resolution; the visible reflectance for a 1 km resolu-
tion is implemented to classify cloudy pixels in the case of COMS imagery. The difference
in spatial resolution in cloud detection raises the accuracy of detecting small cloudy pixels,
i.e., popcorn clouds.
Next, classification of the cloud type was performed using the cloud top pressure
(CTP) and shortwave TOA albedo, based on Rossow and Garder [49], as follows: high-
level cloud (50 ≤ CTP < 440 hPa and shortwave TOA albedo < 0.6), mid-level cloud (440 ≤
CTP < 680 hPa), low-level cloud (680 ≤ CTP < 1000 hPa), and cumulus cloud (50 ≤ CTP <
440 hPa and shortwave TOA albedo ≥ 0.6). The atmospheric transmittance was obtained
by comparing the shortwave TOA albedo between the satellite observations and the look-
up table, under the given conditions of the cloud class, solar zenith angle, and surface
albedo. Finally, the solar irradiance on the ground surface was calculated by multiplying
the cloud fraction, atmospheric transmittance, and time-varying extraterrestrial solar ir-
radiance, corrected by the equation of time and the sun–Earth distance.
4. Results
The monthly average of the daily total irradiance that is derived by the UASIBS–KIER
model, is evaluated against observations conducted at 18 ASOS stations. The error statis-
tics employed in this study were as follows:
()
()
1
1
1
1
N
ii
i
N
i
i
E
O
N
rMBE
O
N
=
=
−
=
,
(1)
()
1
1
1
1
N
ii
i
N
i
i
E
O
N
rMAE
O
N
=
=
−
=
.
(2)
Above Ei, Oi, and N indicate the estimates, observations, and number of samples, respec-
tively. The rMBE and rMAE are the mean bias error and mean absolute error, respectively,
which are normalized to the observed averages. In addition to the aforementioned two
error statistics, Pearson correlation (γ) and determination (γ2) coefficients were used to de-
termine the correlation between the estimations and observations for each station. Figure
2 exhibits the scatter plot of solar irradiance between the UASIBS–KIER model and in situ
observations at 18 ASOS stations, from 1996 to 2019. Most of the estimates are included in
the 95% confidence level, which is consistent with the high correlation coefficient of 0.963
Remote Sens. 2021, 13, 3422 7 of 22
(Table 3). The determination coefficient was also higher than 0.80. The slope of linear re-
gression with intercepts indicates that GHI is overestimated when GHI is lower than 4
kWh m−
2
d−
1
; however, the opposite is true for GHI > 5 kWh m−
2
d−
1
. The contrast between
the estimations and the observed GHI results in a nearly zero value of rMBE (see the rMBE
value in Table 3). However, the dispersion of estimates from the regression line results in
an rMAE of 9.4%.
Figure 2. Scatter plot of the monthly mean of daily total irradiance between the ground observa-
tion (ASOS) and the satellite estimates (UASIBS) with probability density function of frequency for
each dataset. The red dotted line indicates the reference line for perfect correlation.
Table 3. Summary of the relative mean bias error (rMBE, %) and root mean square error (rRMSE,
%) for the hourly mean estimations from the UASIBS/KIER and KMA-INS models for the 18 KMA
ground stations.
Station In-
dex Station ID
ASOS
(kWh m−
2
d−
1
)
UASIBS
(kWh m−
2
d−
1
)
R rMBE
(%)
rMAE
(%)
1 100 3.710 3.471 0.930 –5.85 12.82
2 101 3.664 3.633 0.975 –0.64 7.09
3 105 3.693 3.509 0.959 –4.48 10.50
4 108 3.403 3.719 0.967 9.74 11.38
5 112 3.691 3.807 0.957 4.18 10.55
6 114 3.715 3.652 0.977 –1.46 6.79
7 119 3.537 3.730 0.952 7.03 12.29
Remote Sens. 2021, 13, 3422 8 of 22
8 129 3.672 3.720 0.968 1.93 9.06
9 131 3.704 3.737 0.969 1.12 7.28
10 133 3.989 3.755 0.971 –5.48 9.89
11 135 3.695 3.729 0.940 2.10 11.53
12 136 3.813 3.801 0.965 0.86 11.73
13 143 3.858 3.821 0.971 –0.35 7.72
14 146 3.703 3.764 0.968 2.47 9.98
15 156 3.834 3.749 0.953 –1.70 8.85
16 159 3.871 3.924 0.958 1.92 8.51
17 165 3.840 3.791 0.978 –1.00 6.88
18 184 3.645 3.637 0.976 0.02 9.23
Average 3.724 3.719 0.963 0.58 9.56
The horizontal distribution of solar irradiance is an important parameter in the as-
sessment of solar resources. The scatter plot matrix of solar irradiance between the
UASIBS–KIER model and the measurement for each station is illustrated in Figure 3. The
error statistics in relation to the scatter plots are also listed in Table 3. Almost all the GHI
estimates are in good agreement with the observed correlation coefficients being higher
than 0.92 at all the stations. The rMBE values are between −2% and 2% at 10 stations, which
means that all the estimates are unbiased towards the observed GHI. The rMBE value at
station 108, however, was the largest positive. Figure 4 shows the spatial distribution of
the rMBE values. The UASIBS–KIER model overestimates the observed GHI at the Seoul
station (108), in comparison with satellite cities, Incheon (112) and Suwon (119) stations.
This might be because the current version of the UASIBS–KIER model employs the ob-
served AOD recently, even if the aerosol optical depth gradually decreases in metropoli-
tan cities [50]. In the UASIBS–KIER model, the AOD observed was lower than the actual
values that were depicted in the 1990s and the early 2000s, which resulted in higher solar
irradiance when the sky was clear. This is further discussed at a later stage.
In contrast to the metropolitan cities, negative biases were observed in the mountain-
ous regions, as follows: −5.8% and −4.5% at stations 100 and 105, respectively. These bias
characteristics were also found in a previous study by Kim et al. [39], who concluded that
the cloud optical depth is extremely large when shallow or orographic clouds exist over
the ground stations. The GHI estimations at station 165 appeared to be the most reliable,
because the rMAE was the second lowest and the correlation coefficient was the largest
(Table 3).
Meanwhile, the GHI observations and estimations are distributed in two branches at
stations 135 and 136, as shown in Figure 3. The number of outliers is too large to be treated
as a simple outlier. The UASIBS–KIER model seems to have a symmetrical error while
deriving the solar irradiance. To investigate symmetrical biases, ECMWF reanalysis v5,
ERA-5, was employed in this study. Figure 5 demonstrates the correlation matrix for the
monthly average of the daily total irradiance from the UASIBS–KIER model, ERA-5, and
in situ measurements at stations 135 and 136. In comparison with the ground observa-
tions, the ERA-5 model also estimates the daily total irradiance averaged over each month,
in a similar manner to the UASIBS–KIER model; there are two branches in the correlation
matrix between ERA-5 and ASOS, as observed in Figure 5. Contrary to the relationship
between estimates and observations, the UASIBS – KIER model produces GHI estimates
that are in good agreement with the ERA-5 products. Consequently, the solar irradiance
data from ground stations could be limited to the ground truth. In relation to this limita-
tion, Kim et al. [40] found that the problem in data quality control exists in ground obser-
vation stations that are operated by the KMA.
Remote Sens. 2021, 13, 3422 9 of 22
Figure 3. Same as Figure 2, but distributed for each ground station.
Figure 4. The relative mean bias error for the monthly mean of daily total irradiance that is estimated
by the UASBIS–KIER model with satellite imagery from 1996 to 2019.
Remote Sens. 2021, 13, 3422 10 of 22
Figure 5. Correlation matrix of daily total irradiance averaged over each month from 1996 to 2019 between ground obser-
vation (ASOS), satellite estimates (UASIBS) and reanalysis (ERA-5). The green and orange colors indicate the dataset at
135 and 136 stations, respectively.
From 1996 to 2019, geostationary satellites have changed over the Korean Peninsula.
The UASIBS–KIER model has also been modified to implement various satellite imagery
data and adjust according to the different horizontal resolutions. Thus, henceforth, we
will examine the yearly modeling performance. The scatter plots for the investigation pe-
riod are shown in Figure 6, which indicates that the GHI estimates agree with the observed
values at all the KMA stations. The error statistics for the operation period of each satellite
are summarized in Table 4. The UASIBS–KIER model, with the MTSAT-2 satellite, pro-
duces the largest correlation coefficient between the estimates and observations. Moreo-
ver, the rMAE is the lowest, which means that the GHI estimations are the most reliable
of the four satellite platforms. The UASIBS–KIER model derived larger positive GHI esti-
mates, biased to the observed value, when the GOES-9 satellite was implemented into the
model. Positively biased estimates are also exhibited in Figure 6 (see the scatter plot for
2003 and 2004). Moreover, the rMBE averages appear to have a discrepancy between the
old- and new-generation satellites. As mentioned in Section 2.2, the GMS-5 and GOES-9
satellites are equipped with instruments that can produce the image at spectral bands (3.0
μm−4.0 μm). Without a shortwave infrared channel, the UASIBS–KIER model has a limi-
tation in finding low-level clouds in the cloud classification. On a few occasions, the model
could not detect cloudy pixels, instead defining cloudy pixels as clear. Consequently, the
average GHI estimates were positively larger than the observed values.
Remote Sens. 2021, 13, 3422 11 of 22
Table 4. Error statistics of satellite estimates for each satellite.
Satellite R rMBE (%) rMAE (%)
GMS–5 0.958 5.9 11.0
GOES–9 0.922 7.3 10.7
MTSAT–1R/2 0.976 −3.8 7.1
COMS 0.974 −3.9 9.4
Figure 6. Same as Figure 2, but distributed for every year.
The monthly variation in error statistics is substantially related to the Korean mon-
soon, i.e., the East Asian monsoon. In spring and fall seasons, the sky is usually clear;
however, heavy precipitation events occur in the summer, due to the strong front system
([51–53]). The rMAE values were relatively low in the spring and fall seasons. However,
the UASIBS–KIER model failed to estimate the GHI in the right direction in June and July.
Furthermore, the behavior of rMBE is strictly distinguished by season; it is almost nega-
tive from January to June, but positive from July to December. Therefore, the derivation
model overestimated the COD in the warming season, whereas the opposite was true in
the cooling season.
Remote Sens. 2021, 13, 3422 12 of 22
5. Discussion
5.1. Solar Resource Potentials
The UASIBS–KIER model is capable of estimating solar irradiance for at least 5 km ×
5 km pixels for more than 20 years, from 1996 to 2019. Long-term solar irradiance datasets
are essential for producing typical yearly meteorological data for the energy system de-
sign, or building a solar resource map itself. This section examines the solar resource po-
tentials, based on the Korean solar irradiance datasets, as a result of the UASIBS–KIER
model. The annual mean daily total irradiance over the Korean Peninsula, from 1996 to
2019, is illustrated in Figure 7a, with an average value of 22 years. Even if there are fluc-
tuations by year, the solar irradiance over the Korean Peninsula increases gradually, at a
rate of 0.019 kWh m−2 d−1 per year. The annual mean of the daily total irradiance averaged
over 22 years is found to be 3.602 kWh m−2 d−1, which is equivalent to 3.6 hours at the peak
sun hour. Based on this average value, the present study determines the average year
during the investigation period as 2007, with an annual mean of 3.604 kWh m−2 d−1. In this
study, the dark year was defined as the year when the daily total irradiance averaged over
the year was the lowest. Conversely, the bright year was defined as the year when the
daily total irradiance averaged over the year was the highest. As shown in Figure 7a, the
dark and bright years are determined as 1998 and 2019, respectively.
The monthly variation in solar irradiance for average, dark, and bright years is shown
in Figure 7b. The meteorological conditions in the spring season play a dominant role in
differentiating between dark and bright years. As mentioned in the previous section and
literature ([51–53]), the monthly solar insolation from March to May is recognized as being
higher in both average and bright years, whereas the opposite is true in the dark year.
This might be attributed to the high precipitation amount in 1998. The KMA annual re-
ports from 1996 to 2019 demonstrated that the annual precipitation in Korea was the larg-
est in 1998, for the last three decades [54]. The difference in the daily total irradiance be-
tween dark and bright years, accumulated for three months from March to May, is 112.78
kWh m−2, which is equivalent to 1.227 kWh m−2 d−1 in daily total irradiance. When it is
converted into the peak sun hour in the standard test condition, the operation hour of the
PV system can be increased more in 2017 than in 1998, by 1 hour. Therefore, solar power
generation would be increased to more than 10 GWh when the capacity of the photovol-
taic system is assumed to be 10 GW in Korea.
Figure 8 shows the horizontal distribution of the daily total solar irradiance, averaged
over four satellite operation periods. The solar irradiance derived from each satellite im-
agery was consistent with the annual mean of the daily total irradiance. It is interesting to
investigate solar irradiance by region or administrative divisions. Korea is composed of
17 administrative divisions. The behavior of solar irradiance in 2007 is very similar to that
of the climatological mean, i.e., averages over 22 years from 1996 to 2019. The three largest
values in the average year were the GN, JN, and BS divisions. These results, along with
the climatological mean, are the same for the bright year. In the dark year, the JJ division,
instead of the BS division, is included as the region with the third largest solar irradiance,
out of seventeen divisions. The lowest solar irradiance is observed in the GW division for
all the representative years, because this division is located in the mountainous range and
forest (Figures 1b and 8).
Remote Sens. 2021, 13, 3422 13 of 22
Figure 7. Time series of daily total irradiance that is estimated by the UASIBS–KIER model from 1996 to 2019 in Korea;
annual mean (a) and monthly mean (b). The red and blue line in (a) indicate the average value for 22 years and trend line,
respectively. The black, green and red line mean the monthly variations at dark, bright and average year, respectively.
Figure 8. Horizontal distribution of annual mean of daily total irradiance (kWh m−2 d−1) that is
estimated by the UASIBS–KIER model with GMS–5 (a), GOES–9 (b), MTSAT–1R/2 (c) and COMS
(d).
5.2. Applications for the Photovoltaic System
Solar resource assessments were carried out for photovoltaic system installation and
operation in the feasibility study. We now analyze the capacity factor of PV power plants
Remote Sens. 2021, 13, 3422 14 of 22
installed over seventeen administrative divisions, and then compare it with the solar re-
source map built by satellite-derived solar irradiance. The capacity factor is a measure that
indicates the amount of the power plant operated for a given time period. The annual
capacity factor was formulated as follows:
( )
(%) 100
( ) 8760 ( )
AEP Wh
CF Capacity W h
=×
×, (3)
where AEP indicates the annual energy production, i.e., solar power generation for a year.
Eq. (3) is modified for the monthly capacity factor to which the monthly energy produc-
tion is employed. Figure 9a demonstrates the annual capacity factor of PV power plants
over the Korean Peninsula, from 2005 to 2019. The actual capacity factors are calculated
by solar power generation and capacity, which are extracted from the EPSIS data operated
by KPX. The annual capacity factor gradually increases with an average capacity factor of
13.8% over 15 years. The length of the bar every year implies the magnitude of monthly
variation for each year; the monthly variation has decreased in the late 2010s. The rank
correlation between the standard deviation of the monthly capacity factor and year is dis-
played in Figure 10a, which provides good evidence to prove the inter-annual trend of
monthly variation. A negative value of −0.714 was recognized in the correlation. The
monthly variation in the capacity factor is illustrated in Figure 9b, with inter-annual vari-
ability (= length of bar in Figure 9b). The capacity factors were relatively higher in the
spring season; the opposite was observed in the monsoon season. The behavior of the
capacity factor is similar to that of solar irradiance in Figure 7b (see the solar irradiance in
the average year in Figure 7b). This seems logical because the capacity factor is dependent
on the actual power generation by PV power plants that are pre-dominated by incoming
solar irradiance. Figure 10b demonstrates that the monthly capacity is highly correlated
with the daily total irradiance averaged for each month, with a determination coefficient
of 0.894.
Figure 9. Capacity factor of photovoltaic systems in Korea with their standard deviation (bar) as a function of year (a) and
month (b). Capacity factor in (a,b) is calculated by using EPSIS data. The red bar in (a) indicates the average value from
1996 to 2019.
Remote Sens. 2021, 13, 3422 15 of 22
Figure 10. Scatter plots: (a) rank between year and
σ
CF
(standard deviation) that is EPSIS data, (b) capacity factor and
monthly mean of daily total irradiance that is output from the UASIBS–KIER model. .
The technological sophistication of PV power systems over the years is the reason
why monthly variation has reduced with time. For example, the champion PV module
efficiency chart, examined by NREL [55], shows that, on average, PV module efficiency
has increased by 20% from 2005 to 2020. In Korea, installed PV capacity has drastically
increased during the last decade (Figure 11). Consequently, a PV module with higher ef-
ficiency can generate electricity, even at lower solar insolation; the annual capacity factor
could be increased throughout the year.
Figure 11. Annual capacity of photovoltaic system installed in Korea, based on the EPSIS data.
The UASIBS–KIER model output was employed to estimate the capacity factor in
Korea. Due to a lack of information on PV plants, the annual capacity factor from the
UASIBS–KIER model was derived, with the following assumptions [56]:
Remote Sens. 2021, 13, 3422 16 of 22
1. The plane of array irradiance is not considered.
2. The effect of cell temperature on the nameplate efficiency is ignored.
3. The peak power of PV module is the same for all grid cells in the model.
4. With the aforementioned assumptions, the capacity factor is formulated by Equation
(4), as follows:
21
365
2
1
( ) ( )
1000 ( )
(%) 100
( ) 8760 ( )
i
DTI kWh m d Capacity W
Wm
CF Capacity W h
−−
−
=
×
=×
×
, (4)
where DTI is the total daily irradiance. Eq. (4) is calculated for 365 days, as it is the annual
capacity factor. When applied to the monthly capacity factor, the number of days for each
month was employed, instead of days in the year. Statistical analysis was performed to
identify the reliability of the modeling output against the measurements. Figure 12 shows
a scatter plot of the monthly capacity factor between EPSIS and the UASIBS–KIER model.
The Pearson correlation and determination coefficients are 0.912 and 0.832, respectively,
inferring that the UASIBS–KIER model is sufficiently reliable to represent the PV output
and solar irradiance. The slope of the linear regression, however, was 1.301, i.e., the
UASIBS–KIER model overestimated the capacity factor when the measurements exceeded
15%. Vilanova et al. [57] stated that solar irradiance is not directly proportional to PV out-
put, because higher solar insolation plays a role in heating the solar cell itself. This reduces
the efficiency of the PV module. For simplicity, however, this study ignored the effect of
cell temperature on the PV module. Consequently, the PV output is commensurate with
the solar irradiance, regardless of the solar cell influence.
Figure 12. Scatter plot of the annual capacity factors between the EPSIS data and the UASIBS–
KIER model estimates with probability density function of frequency for each dataset. The red
dotted line indicates the reference line for perfect correlation.
Remote Sens. 2021, 13, 3422 17 of 22
Meanwhile, in Figure 13, the horizontal distributions of the annual capacity factor,
derived from the UASIBS–KIER model from 2016 to 2019, are demonstrated. The differ-
ence in the annual capacity factors between the 17 administrative divisions is not large
enough to justify the local characteristics. Nevertheless, GN, BS, and JN are determined
as administrative divisions in which the capacity factors are larger than the other divi-
sions. This is similar to the results from the solar irradiance distribution shown in Figure
8. Moreover, the divisions GW and JJ are administrative divisions demonstrating rela-
tively small capacity factors. Lastly, the monthly capacity factor was compared for each
division from 2016 to 2019. Similarly to the annual capacity factor in Figure 12, an overes-
timation is observed for all the divisions with high correlation coefficients. Positive biases
at larger EPSIS capacity factors might be related to the impact of cell temperature on the
PV module efficiency. In general, the capacity factor is higher on a clear day in summer,
because solar irradiance is not diffused by the cloud droplets. However, the electricity
generation is not proportional to the incoming solar irradiance, since the high temperature
on the PV module, raised by the solar insolation, makes the conversion rate into electricity
worse. In Eq. (4), we do not assume the cell-temperature impacts that play a role in reduc-
ing the capacity factor when solar irradiance is high. Therefore, the effect of cell tempera-
ture on the PV module must be considered, to reduce the positive bias of the capacity
factor (Figure 14).
Figure 13. Horizontal distribution of annual capacity factor that is estimated by the UASIBS–KIER
model in 2016 (a), 2017 (b), 2018 (c), and 2019 (d).
Remote Sens. 2021, 13, 3422 18 of 22
Figure 14. Same as Figure 12, but distributed for each administrative division in Korea.
6. Conclusions
Korean solar irradiance datasets for the long-term period were built using the
UASIBS–KIER model, with geostationary satellite imagery that is available in Korea. The
monthly estimations appear to be unbiased, because the rMBE is only 0.2% for all the sta-
tions. These results are also consistent with the higher correlation coefficient of 0.92. The
dispersion of estimates results in an rMAE of 9.4%. The local characteristics of rMBE are
interesting; they are positive in metropolitan regions, but negative in mountainous re-
gions. AOD profiles, used in generating the look-up table, ensued from the in situ obser-
Remote Sens. 2021, 13, 3422 19 of 22
vation at Yonsei University, as part of the AERONET program. As the AERONET pro-
gram was initiated in March 2011, the UASIBS–KIER model estimated the solar irradiance
that was measured before 2011 using the climatological mean of the AOD data from the
Yonsei University, from 2011 to 2019. This results in larger positive GHI estimates than
that observed in the metropolitan region, as the AOD that is lower than the actual obser-
vation is parameterized into the UASIBS–KIER model. The UASIBS–KIER model with
MTSAT-1R/2 satellites performed the best in deriving the solar irradiance, even if the spa-
tial and temporal resolutions were better in the COMS than in the MTSAT series. The
horizontal resolution of the MTSAT series was 4 km, but the COMS produced the imagery
at 1 km. Therefore, by upscaling into coarse resolution, the biases can be eliminated from
each other, as a result of smoothing.
Solar resources have been gradually increasing in Korea, at a rate of 0.019 kWh m−2
d−1 per year. The daily total irradiance, averaged over 22 years, was 3.602 kWh m−2 d−1,
which is equivalent to 3.6 hours at the peak sun hour. From 1996 to 2019, the average year
was determined to be 2007, which was when the average value of daily total irradiance
was the closest to 3.602 kWh m−2 d−1. The dark and bright years are set as 1998 and 2019,
respectively, based on the annual mean of the daily total irradiance. The distinct difference
between the dark and bright years is the solar irradiance in the spring season. In Korea,
clear skies prevail from March to May. During this period, on average, the peak sun hours
every day were an hour longer in the bright year than in the dark year. Out of the 17
administrative divisions, the GN, JN, and BS divisions that belong to the southern part of
the Korean Peninsula were divisions with comparatively more solar irradiance than the
other divisions. In contrast to these divisions, GW located in a mountainous region was
recognized as a region with a relatively small amount of solar resources.
As an application of the PV power system, the annual capacity factor averaged over
15 years, from 2005 to 2019, is computed to be 13.8%, based on EPSIS data. The capacity
factor is derived from the GHI estimates using the UASIBS–KIER model. The correlation
coefficient between the model estimates and EPSIS data is 0.912, which implies that the
UASIBS–KIER model is capable of extending the viability of this study in renewable en-
ergy. However, positive biases appear at larger capacity factors, as recorded by EPSIS.
Therefore, this study did not consider the temperature dependence of the efficiency of the
PV module. When solar insolation is high, the surface temperature of the PV module also
increases, thereby lowering the efficiency of the module. Consistent with the solar re-
source map, the southern part of the Korean Peninsula is determined as a region where
the annual capacity factor is larger than the other parts. This study is the first to estimate
solar irradiance values as a gridded dataset in Korea for a long period. The reliable solar
irradiance datasets, built by the UASBIS–KIER model, are expected to contribute to the
typical meteorological year data for energy system modeling.
Author Contributions: C.-K.K. and H.-G.K. conceptualized and designed the study. B.K., C.-Y.Y.
and J.-Y.K. provided the data for system. Y.-H.K. gave insight into the research and then H.-G.K.
supervised the research. C.-K.K. wrote sections of the manuscript. All authors contributed to the
manuscript revision, read, and approved the submitted version.
Funding: This work was conducted under framework of the research and development program of
the Korea Institute of Energy Research (C1-2410).
Data Availability Statement: The original contributions presented in the study are included in the
article/supplementary material; further inquiries can be directed to the corresponding author.
Acknowledgments: The authors express deep thanks to the Korea Power Exchange for providing
the electricity generation data at the individual solar power plants.
Conflicts of Interest: The authors declare no conflict of interest.
Remote Sens. 2021, 13, 3422 20 of 22
References
1. Zelenka, A.; Perez, R.; Seals, R.; Renné, D. Effective Accuracy of Satellite-Derived Hourly Irradiances. Theor. Appl. Climatol. 1999,
62, 199–207, doi:10.1007/s007040050084.
2. Vignola, F.; Harlan, P.; Perez, R.; Kmiecik, M. Analysis of satellite derived beam and global solar radiation data. Sol. Energy
2007, 81, 768–772, http://dx.doi.org/10.1016/j.solener.2006.10.003.
3. Chow, C.W.; Urquhart, B.; Lave, M.; Dominguez, A.; Kleissl, J.; Shields, J.; Washom, B. Intra-hour forecasting with a total sky
imager at the UC San Diego solar energy testbed. Sol. Energy 2011, 85, 2881–2893, http://dx.doi.org/10.1016/j.solener.2011.08.025.
4. Kleissl, J.P. Solar Energy Forecasting and Resource Assessment, 1st ed.; Academic Press: Cambridge, MA, USA, 2013;
http://dx.doi.org/10.1016/S0074-6142(10)09912-2pp. 416.
5. He, G.; Kammen, D.M. Where, when and how much solar is available? A provincial-scale solar resource assessment for China.
Renew. Energy 2016, 85, 74–82, https://doi.org/10.1016/j.renene.2015.06.027.
6. Wegertseder, P.; Lund, P.; Mikkola, J.; Alvarado, R.G. Combining solar resource mapping and energy system integration meth-
ods for realistic valuation of urban solar energy potential. Sol. Energy 2016, 135, 325–336.
7. Gilgen, H.; Wild, M.; Ohmura, A. Means and Trends of Shortwave Irradiance at the Surface Estimated from Global Energy
Balance Archive Data. J. Clim. 1998, 11, 2042–2061, doi:10.1175/1520-0442-11.8.2042.
8. Stanhill, G.; Cohen, S. Solar Radiation Changes in the United States during the Twentieth Century: Evidence from Sunshine
Duration Measurements. J. Clim. 2005, 18, 1503–1512, doi:10.1175/jcli3354.1.
9. Bishop, J.K.B.; Rossow, W.B. Spatial and temporal variability of global surface solar irradiance. J. Geophys. Res. Ocean. 1991, 96,
16839–16858, doi:10.1029/91jc01754.
10. Bishop, J.K.B.; Rossow, W.B.; Dutton, E.G. Surface solar irradiance from the International Satellite Cloud Climatology Project
1983–1991. J. Geophy. Res. 1997, 102, 6883–6910, https://doi.org/10.1029/96JD03865.
11. Janjai, S.; Pankaew, P.; Laksanaboonsong, J.; Kitichantaropas, P. Estimation of solar radiation over Cambodia from long-term
satellite data. Renew. Energy 2011, 36, 1214–1220, https://doi.org/10.1016/j.renene.2010.09.023.
12. Sengupta, M.; Xie, Y.; Lopez, A.; Habte, A.; Maclaurin, G.; Shelby, J. The National Solar Radiation Data Base (NSRDB). Renew.
Sustain. Energy Rev. 2018, 89, 51–60, https://doi.org/10.1016/j.rser.2018.03.003.
13. Haupt, S.E.; Kosović, B. Variable Generation Power Forecasting as a Big Data Problem. IEEE Trans. Sustain. Energy 2017, 8, 725–
732, doi:10.1109/TSTE.2016.2604679.
14. Gautier, C.; Diak, G.; Masse, S. A Simple Physical Model to Estimate Incident Solar Radiation at the Surface from GOES Satellite
Data. J. Appl. Meteorol. Climatol. 1980, 19, 1005–1012, doi:10.1175/1520-0450(1980)019<1005:aspmte>2.0.co;2.
15. Moser, W.; Raschke, E. Incident Solar Radiation over Europe Estimated from METEOSAT Data. J. Clim. Appl. Meteorol. 1984, 23,
166–170, doi:10.1175/1520-0450(1984)023<0166:isroee>2.0.co;2.
16. Noia, M.; Ratto, C.F.; Festa, R. Solar irradiance estimation from geostationary satellite data: I. Statistical models. Sol. Energy
1993, 51, 449–456, https://doi.org/10.1016/0038-092X(93)90130-G.
17. Pinker, R.T.; Ewing, J.A. Modeling Surface Solar Radiation: Model Formulation and Validation. J. Clim. Appl. Meteorol. 1985, 24,
389–401, doi:10.1175/1520-0450(1985)024<0389:mssrmf>2.0.co;2.
18. Pinker, R.T.; Laszlo, I. Modeling Surface Solar Irradiance for Satellite Applications on a Global Scale. J. Appl. Meteorol. 1992, 31,
194–211, doi:10.1175/1520-0450(1992)031<0194:mssifs>2.0.co;2.
19. Li, Z.; Leighton, H.G. Global climatologies of solar radiation budgets at the surface and in the atmosphere from 5 years of ERBE
data. J. Geophys. Res. 1993, 98, 4919–4930, doi:10.1029/93jd00003.
20. Gupta, S.K.; Ritchey, N.A.; Wilber, A.C.; Whitlock, C.H.; Gibson, G.G.; Stackhouse, P.W. A Climatology of Surface Radiation
Budget Derived from Satellite Data. J. Clim. 1999, 12, 2691–2710, doi:10.1175/1520-0442(1999)012<2691:acosrb>2.0.co;2.
21. Pinker, R.T.; Tarpley, J.D.; Laszlo, I.; Mitchell, K.E.; Houser, P.R.; Wood, E.F.; Schaake, J.C.; Robock, A.; Lohmann, D.; Cosgrove,
B.A.; et al. Surface radiation budgets in support of the GEWEX Continental-Scale International Project (GCIP) and the GEWEX
Americas Prediction Project (GAPP), including the North American Land Data Assimilation System (NLDAS) project. J. Ge-
ophys. Res. 2003, 108, 8844, doi:10.1029/2002jd003301.
22. Wang, H.; Pinker, R.T. Shortwave radiative fluxes from MODIS: Model development and implementation. J. Geophys. Res. 2009,
114, D20201, doi:10.1029/2008jd010442.
23. Ma, Y.; Pinker, R.T. Modeling shortwave radiative fluxes from satellites. J. Geophys. Res. 2012, 117, D23202,
doi:10.1029/2012jd018332.
24. Kim, C.K.; Holmgren, W.F.; Stovern, M.; Betterton, E.A. Toward Improved Solar Irradiance Forecasts: Derivation of Down-
welling Surface Shortwave Radiation in Arizona from Satellite. Pure Appl. Geophys. 2016, 173, 2535–2553, doi:10.1007/s00024-
016-1302-3.
25. Deneke, H.M.; Feijt, A.J.; Roebeling, R.A. Estimating surface solar irradiance from METEOSAT SEVIRI-derived cloud proper-
ties. Remote Sens. Environ. 2008, 112, 3131–3141, doi:https://doi.org/10.1016/j.rse.2008.03.012.
26. Journée, M.; Müller, R.; Bertrand, C. Solar resource assessment in the Benelux by merging Meteosat-derived climate data and
ground measurements. Sol. Energy 2012, 86, 3561–3574, doi:https://doi.org/10.1016/j.solener.2012.06.023.
27. Nikitidou, E.; Kazantzidis, A.; Tzoumanikas, P.; Salamalikis, V.; Bais, A.F. Retrieval of surface solar irradiance, based on satel-
lite-derived cloud information, in Greece. Energy 2015, 90, 776–783, doi:https://doi.org/10.1016/j.energy.2015.07.103.
28. Kazantzidis, A.; Nikitidou, E.; Salamalikis, V.; Tzoumanikas, P.; Zagouras, A. New challenges in solar energy resource and
forecasting in Greece. Int. J. Sustain. Energy 2018, 37, 428–435, doi:10.1080/14786451.2017.1280495.
Remote Sens. 2021, 13, 3422 21 of 22
29. Tanahashi, S.; Kawamura, H.; Matsuura, T.; Takahashi, T.; Yusa, H. Improved Estimates of Hourly Insolation from GMS S-
VISSR Data. Remote Sens. Environ. 2000, 74, 409–413, doi:https://doi.org/10.1016/S0034-4257(00)00133-4.
30. Janjai, S.; Laksanaboonsong, J.; Nunez, M.; Thongsathitya, A. Development of a method for generating operational solar radia-
tion maps from satellite data for a tropical environment. Sol. Energy 2005, 78, 739–751,
doi:https://doi.org/10.1016/j.solener.2004.09.009.
31. Lu, N.; Liu, R.; Liu, J.; Liang, S. An algorithm for estimating downward shortwave radiation from GMS 5 visible imagery and
its evaluation over China. J. Geophys. Res. 2010, 115, doi:https://doi.org/10.1029/2009JD013457.
32. Blanksby, C.; Bennett, D.; Langford, S. Improvement to an existing satellite data set in support of an Australia solar atlas. Sol.
Energy 2013, 98, 111–124, doi:https://doi.org/10.1016/j.solener.2012.10.026.
33. Zhang, X.; Liang, S.; Zhou, G. Estimating downward surface shortwave radiation using MTSAT-1R and ground measurements
data by Bayesian maximum entropy method. In Proceedings of the 2013 IEEE International Geoscience and Remote Sensing
Symposium—IGARSS, Melbourne, VIC, Australia, 21–26 July 2013; pp. 1541–1543.
34. Dehghan, A.; Prasad, A.A.; Sherwood, S.C.; Kay, M. Evaluation and improvement of TAPM in estimating solar irradiance in
Eastern Australia. Sol. Energy 2014, 107, 668–680, doi:https://doi.org/10.1016/j.solener.2014.06.018.
35. Yeom, J.-M.; Han, K.-S.; Lee, C.-S.; Kim, D.-Y. An Improved Validation Technique for the Temporal Discrepancy when Esti-
mated Solar Surface Insolation Compare with Ground-based Pyranometer: MTSAT-1R Data use. Korean J. Remote Sens. 2008, 24,
605–612.
36. Lee, K.; Yoo, H.; Levermore, G.J. Quality control and estimation hourly solar irradiation on inclined surfaces in South Korea.
Renew. Energy 2013, 57, 190–199, doi:https://doi.org/10.1016/j.renene.2013.01.028.
37. Jee, J.-B.; Zo, I.-S.; Lee, K.-T. A Study on the Retrievals of Downward Solar Radiation at the Surface based on the Observations
from Multiple Geostationary Satellites. Korean J. Remote Sens. 2013, 29, 123–135.
38. Zo, I.-S.; Jee, J.-B.; Lee, K.-T. Development of GWNU (Gangneung-Wonju National University) one-layer transfer model for
calculation of solar radiation distribution of the Korean peninsula. Asia-Pac. J. Atmos. Sci. 2014, 50, 575–584, doi:10.1007/s13143-
014-0047-0.
39. Kim, C.K.; Kim, H.-G.; Kang, Y.-H.; Yun, C.-Y. Toward Improved Solar Irradiance Forecasts: Comparison of the Global Hori-
zontal Irradiances Derived from the COMS Satellite Imagery Over the Korean Peninsula. Pure Appl. Geophys. 2017, 174, 2773–
2792, doi:10.1007/s00024-017-1578-y.
40. Kim, C.K.; Kim, H.-G.; Kang, Y.-H.; Yun, C.-Y.; Lee, Y.G. Intercomparison of Satellite-Derived Solar Irradiance from the GEO-
KOMSAT-2A and HIMAWARI-8/9 Satellites by the Evaluation with Ground Observations. Remote Sens. 2020, 12, 2149.
41. OGD. Public Data Portal. Availabe online: https://www.data.go.kr/en/index.do (accessed on 1 March 2021).
42. Kikuchi, T.; Kitsuregawa, M. GMS-5 Meteorological Satellite Image Database and Integrated Visualization System; 2001. (in Japanese
with English abstract). IPSJ Rep. SIG8 (TOD10)
43. Muñoz Sabater, J. ERA5-Land Hourly Data from 1981 to Present. Copernicus Climate Change Service (C3S) Climate Data Store
(CDS). Availabe online: https://cds.climate.copernicus.eu/cdsapp#!/dataset/reanalysis-era5-land?tab=overview (accessed on 1
March 2021).
44. Ineichen, P.; Barroso, C.S.; Geiger, B.; Hollmann, R.; Marsouin, A.; Mueller, R. Satellite Application Facilities irradiance prod-
ucts: Hourly time step comparison and validation over Europe. Int. J. Remote Sens. 2009, 30, 5549–5571,
doi:10.1080/01431160802680560.
45. Chou, M.-D.; Suarez, M.J. A Solar Radiation Parameterization for Atmospheric Studies; 1999–104606; Goddard Space Flight Center:
Greenbelt, MD, USA, 1999; p. 65.
46. Tilmes, S.; Lamarque, J.F.; Emmons, L.K.; Conley, A.; Schultz, M.G.; Saunois, M.; Thouret, V.; Thompson, A.M.; Oltmans, S.J.;
Johnson, B.; et al. Technical Note: Ozonesonde climatology between 1995 and 2011: Description, evaluation and applications.
Atmos. Chem. Phys. 2012, 12, 7475–7497, doi:10.5194/acp-12-7475-2012.
47. Ruiz-Arias, J.A.; Dudhia, J.; Gueymard, C.A.; Pozo-Vázquez, D. Assessment of the Level-3 MODIS daily aerosol optical depth
in the context of surface solar radiation and numerical weather modeling. Atmos. Chem. Phys. 2013, 13, 675–692, doi:10.5194/acp-
13-675-2013.
48. Jedlovec, G.J.; Haines, S.L.; LaFontaine, F.J. Spatial and Temporal Varying Thresholds for Cloud Detection in GOES Imagery.
Geosci. Remote Sens. IEEE Trans. 2008, 46, 1705–1717, doi:10.1109/tgrs.2008.916208.
49. Rossow, W.B.; Walker, A.W.; Garder, L.C. Comparison of ISCCP and Other Cloud Amounts. J. Clim. 1993, 6, 2394–2418,
doi:10.1175/1520-0442(1993)006<2394:coiaoc>2.0.co;2.
50. Nam, J.; Kim, S.-W.; Park, R.J.; Park, J.-S.; Park, S.S. Changes in column aerosol optical depth and ground-level particulate
matter concentration over East Asia. Air Qual. Atmos. Health 2018, 11, 49–60, doi:10.1007/s11869-017-0517-5.
51. Chung, Y.-S.; Yoon, M.-B.; Kim, H.-S. On Climate Variations and Changes Observed in South Korea. Clim. Chang. 2004, 66, 151–
161, doi:10.1023/B:CLIM.0000043141.54763.f8.
52. Ho, C.-H.; Lee, E.-J.; Lee, I.; Jeong, S.-J. Earlier spring in Seoul, Korea. Int. J. Climatol. 2006, 26, 2117–2127,
doi:https://doi.org/10.1002/joc.1356.
53. Ha, K.-J.; Heo, K.-Y.; Lee, S.-S.; Yun, K.-S.; Jhun, J.-G. Variability in the East Asian Monsoon: A review. Meteorol. Appl. 2012, 19,
200–215, doi:https://doi.org/10.1002/met.1320.
54. KMA. Climate of Korea. Availabe online: https://web.kma.go.kr/eng/biz/climate_01.jsp (accessed on 1 March 2021).
Remote Sens. 2021, 13, 3422 22 of 22
55. NREL. Best Research-Cell Efficieny Chart. Availabe online: https://www.nrel.gov/pv/cell-efficiency.html (accessed on 1 March
2021).
56. Dobos, A.P. PVWatts Version 5 Manual Technical Report NREL/TP-6A20-62641; 2014, doi:10.2172/1158421.
57. Vilanova, A.; Kim, B.-Y.; Kim, C.K.; Kim, H.-G. Linear-Gompertz Model-Based Regression of Photovoltaic Power Generation
by Satellite Imagery-Based Solar Irradiance. Energies 2020, 13, 781.
