Toward More Integrated Utilizations of Geostationary Satellite Data for Disaster Management and Risk Mitigation

Third-generation geostationary meteorological satellites (GEOs), such as Himawari-8/9 Advanced Himawari Imager (AHI), Geostationary Operational Environmental Satellites (GOES)-R Series Advanced Baseline Imager (ABI), and Meteosat Third Generation (MTG) Flexible Combined Imager (FCI), provide advanced imagery and atmospheric measurements of the Earth’s weather, oceans, and terrestrial environments at high-frequency intervals. Third-generation GEOs also significantly improve capabilities by increasing the number of observation bands suitable for environmental change detection. This review focuses on the significantly enhanced contribution of third-generation GEOs for disaster monitoring and risk mitigation, focusing on atmospheric and terrestrial environment monitoring. In addition, to demonstrate the collaboration between GEOs and Low Earth orbit satellites (LEOs) as supporting information for fine-spatial-resolution observations required in the event of a disaster, the landfall of Typhoon No. 19 Hagibis in 2019, which caused tremendous damage to Japan, is used as a case study.


Introduction
For weather analysis and forecasting, geostationary satellites (GEOs) have an advantage over polar orbiters; images are captured frequently rather than once or twice per day. Thus, the motion and rate of change in weather systems can be observed. Meteorological agencies such as the US National Oceanic and Atmospheric Administration (NOAA), Japan Meteorological Agency (JMA), China Meteorological Administration (CMA), and European Organisation for the Exploitation of Meteorological Satellites (EUMETSAT) have developed and launched geostationary meteorological satellites [1]. Typically, there are five or six satellites that can cover the globe near the equator. The era of imaging the Earth from a geostationary perspective began on 6 December 1966, with the launch of an experimental sensor (Spin-Scan Cloudcover Camera) onboard Application Technology Satellite-1 (ATS-1, [2,3]). The JMA launched its first geostationary meteorological satellite (GMS) in 1977 [4,5].
First-generation GEOs (for example, GMS 1 to 3, and Geostationary Operational Environmental Satellites' (GOES) first-generation Visible and Infrared Spin Scan Radiometer (VISSR) [6]) had only two channels that were visible, and the other was a thermal infrared (TIR) channel of approximately 10 µm, which could capture images of the Earth once every three hours. In second-generation GEOs (for example, GMS4/5 VISSR, GOES second generations [6], Multi-functional Transport Satellite (MTSAT)-1R/-2 Japanese Advanced Meteorological Imager (JAMI) [7]), enhancements were made, such as increasing the frequency of observations (hourly full-disk (FD) scan), adding bands in the atmospheric window region by utilizing the split-window technique [8], and adding water vapor channels [9,10] of approximately 7 µm. The METEOSAT Second Generation (MSG) Spinning Enhanced Visible and InfraRed Imager (SEVIRI) [11] is considered to be a second-anda-half-generation GEO. With enhancements such as an increase in the number of bands Table 1. Band center wavelength specifications for major third-generation geostationary meteorological satellite optical imagers (in µm). AHI, Advanced Himawari Imager; GOES, Geostationary Operational Environmental Satellites; ABI, Advanced Baseline Imager; MTG, Meteosat Third Generation; FCI, Flexible Combined Imager; FY-4A, Fengyun-4A; AGRI, Advanced Geostationary Radiation Imager; GK-2A, GEO The amount of information obtained due to image discrimination by the human eye is dramatically different between monochrome and color images. An increase in the number of bands of imagers onboard satellites, represented by the change from NOAA/ Advanced Very High Resolution Radiometer (AVHRR) to Terra/Aqua Moderate Resolution Imaging Spectroradiometer (MODIS), has made global True Color RGB composition possible. The large increase in the number of observation bands in third-generation GEOs (Table 1) has had the same impact on high-frequency observations as the change from AVHRR to Remote Sens. 2021, 13, 1553 3 of 16 MODIS had. In other words, third-generation GEOs have resulted in a breakthrough from "monochrome images" to "full-color movies". To achieve a True Color RGB composite, it is necessary to separate the visible wavelength into three bands (red, green, and blue (RGB)); however, Table 1 shows that not all the third-generation GEO imagers have three bands in the visible wavelength. In addition, even if they do achieve RGB separation, their central wavelengths are often biased. Therefore, the development of technology for RGB composites that match each imager's characteristics is in progress [14][15][16][17][18].
Multi-band GEOs are used for RGB composite during the daytime, whereas pseudo-RGB mainly uses thermal infrared bands from the viewpoint of monitoring atmospheric phenomena. The EUMETSAT, JMA, and other agencies share the generation and interpretation of multi-band RGB composite images by publishing the recipes of RGB composites and their interpretations on the Internet [19][20][21].

Visualization through the Web-Interface
From a weather disaster mitigation point of view, the quick dissemination of weather information to all countries in a risk region is of great importance. GEOs play a vital role in providing continuous atmospheric observations for weather forecasting and monitoring a wide range of environmental phenomena. GEOs provide valuable weather information, but providing extensive information on a website in real time is a challenge [22][23][24][25][26][27]. The Himawari-8 real-time website [25] is a highly sophisticated website for monitoring Himawari-8 images in multiple languages, including English, Japanese, Korean, Chinese, Indonesian, Burmese, Thai, Russian, French, and Tetum. An interactive, well-designed website can help in visualizing weather events, so the Himawari-8 real-time website is a good example viewer to capture target events by users. The Himawari-8 real-time website takes international collaboration into account and shows that the latency of real-time image acquisition can be improved by introducing proxy servers in the target countries [23]. However, the Japan Aerospace Exploration Agency (JAXA) Himawari monitor can superimpose a group of products applied to Himawari, such as algorithms developed mainly for the Second-generation Global Imager (SGLI) boarded on the Global Change Observation Mission-Climate (GCOM-C) by JAXA [25].

Baseline Dataset and Dataset Infrastructure
Because the primary purpose of GEO data is weather monitoring, the quasi-real-time use of captured images is fundamental. GEOs are operated by meteorological operational agencies or meteorological satellite organizations, and GEO data are highly public; thus, GEO data are provided free of charge [28][29][30]. In the Himawari series operated by JMA, the data are provided for a fee to share the operational cost for stable data provision for commercial use [31]. Third-generation GEOs release a large amount of data by enhancing functions. In addition to data latency being essential for access, private clouds provide the GOES-R series dataset [32,33].
As GEOs observe the entire globe from approximately 36,000 km above the Earth, they have the advantage of capturing the Earth in a spherical shape. However, spherical datasets are difficult to handle from a data analysis point of view. Therefore, it is desirable to convert spherical datasets to a latitude-longitude coordinate system (so-called gridded data) or to use a converted gridded dataset. Sample programs for geo-correction based on satellites' navigation information are provided by meteorological agencies or research communities [34,35]. The navigation control of third-generation GEOs is now more precise than ever before [36,37], and even gridded data obtained by applying the aforementioned geo-correction program can maintain sufficient geometric correction accuracy if the primary purpose is weather monitoring. Takenaka et al. [38] developed a fast and accurate geometric correction technique using the phase-only correlation (POC) method [39] and applied it to the Himawari-8 and GOES-R series [40,41]. Yamamoto et al. [42] validated the geocorrection accuracy of the Himawari-8 gridded dataset and found that the geometric correction was more accurate than that based on satellite navigation information. Such a Remote Sens. 2021, 13, 1553 4 of 16 highly accurate geometric correction dataset is valuable as a baseline dataset for terrestrialenvironment monitoring, especially for environmental monitoring during disaster events.
However, it is essential to develop long-term observational GEO datasets in climate data records (CDRs). A global 30-min high-frequency GEO synthetic infrared (IR) dataset was generated as baseline data for multi-satellite data blended precipitation products [43,44]. This globally merged IR dataset is useful for convection monitoring. As part of the CDR program, a globally synthesized product (GridSat-B1, [45]) with basic 3-band (10 µm IR, 7 µm WV, and VIS) data onboard GEOs and the GOES dataset have also been released [46], which are useful for long-term variability analysis. The Center for Environmental Remote Sensing (CEReS), Chiba University, Japan, has also been archiving GEO data. The second-generation GEO generates and publishes gridded data of all onboard band data based on satellite navigation information as one of the GEO's active archive centers.

Clouds and Precipitation
GEOs' primary function is to monitor the weather by observing exact locations with a high observation frequency. Thus, GEOs are suitable for monitoring cloud and precipitation processes, such as the diurnal cycle of convective activity using first-and second-generation GEO data [9,[47][48][49], and life-stage identification for convective clouds with ground-based or satellite-based radars [50][51][52][53]. By further increasing the observation frequencies (with a rapid-scan observation experiment in MTSAT-1R and a rapid-scan mode in MSG), it has become possible to capture more of the life stages for single convection [54,55], demonstrating the usefulness of high-frequency observations by GEOs.
One of the features of third-generation GEOs is that they are multichannel, similar to MODIS. Such similarities mean that the optical characteristics of clouds (the optical thickness of clouds, and the effective radiative radius of cloud particles) estimated by optical sensors such as MODIS [56][57][58][59] can be applied to third-generation GEOs [60][61][62][63]. In addition, the synergistic observation of the same target by the multi-principle sensors called A-Train [64,65] can deepen the understanding of the cloud growth process [66][67][68] system by taking advantage of the characteristics of high-frequency observations. As a good example, the comprehensive observation of cloud and precipitation systems [69] in the Boso area, Chiba Prefecture, Japan, will be introduced. In the same area, X-band Phased Array Weather Radar (PAWR) was installed and operated to monitor the three-dimensional precipitation cell structure with a 30-s ultra-fine time resolution [70]. Figure 1 demonstrates the time evolution of the optical properties of a cloud system estimated by the cloud microphysical properties algorithm (CAPCOM) [71] (Figure 1, top), and the same time changes in the vertical profile of radar reflectivity captured by the PAWR (Figure 1, bottom), with the time of the first-echo detection by PAWR set to zero. Himawari-8 captured an apparent increase in cloud optical thickness (red line) starting 30 min before the first echo detection, and followed by a peak in visible reflectance (black line). The dramatic enhancement of the third-generation GEO provides a breakthrough in the better understanding cloud and precipitation processes by identifying the detail of cloud systems lifetime. In addition, attempts are made to refine the classification of cloud types proposed by the International Satellite Cloud Climatology Project (ISCCP) [72], by using the split windows method with multiband data and the brightness temperature difference parameter [73]. enhancement of the third-generation GEO provides a breakthrough in the better understanding cloud and precipitation processes by identifying the detail of cloud systems lifetime. In addition, attempts are made to refine the classification of cloud types proposed by the International Satellite Cloud Climatology Project (ISCCP) [72], by using the split windows method with multiband data and the brightness temperature difference parameter [73]. The enhanced functionality of third-generation GEOs has a substantial impact on weather forecasting with computer resource advancement and data assimilation [74][75][76]. One of the original purposes of GEOs was to calculate atmospheric moving vector winds, and they have contributed to the accuracy of weather forecasting by improving the atmospheric wind field. In addition, the Himawari-8 has a target area observation mode, especially for typhoons. It can make a high-frequency observation every 2.5 min [5]. Typhoons are a great source of material for machine learning because they cause significant damage by landing [77,78]. An excellent example of the combination of the development of machine learning and the enhancement of third-generation GEOs is the estimation of rain rate by the Himawari-8 using random-forest machine learning [79]. Hirose et al. [79] applied the methodology used initially for MSG to learn from ground-based radar data [80], and they succeeded in better reproducing precipitation with low storm height but heavy The enhanced functionality of third-generation GEOs has a substantial impact on weather forecasting with computer resource advancement and data assimilation [74][75][76]. One of the original purposes of GEOs was to calculate atmospheric moving vector winds, and they have contributed to the accuracy of weather forecasting by improving the atmospheric wind field. In addition, the Himawari-8 has a target area observation mode, especially for typhoons. It can make a high-frequency observation every 2.5 min [5]. Typhoons are a great source of material for machine learning because they cause significant damage by landing [77,78]. An excellent example of the combination of the development of machine learning and the enhancement of third-generation GEOs is the estimation of rain rate by the Himawari-8 using random-forest machine learning [79]. Hirose et al. [79] applied the methodology used initially for MSG to learn from ground-based radar data [80], and they succeeded in better reproducing precipitation with low storm height but heavy rain rate, which is often seen in wet areas [81], by utilizing the three-splitting bands in water vapor (WV) absorption bands around 7 µm. In addition, the improvement of precipitation estimation accuracy by nowcasting global precipitation products [82] using high-frequency observations may improve disaster prediction accuracy, mainly for flood event detection.

Dust Events and Aerosols
Precise monitoring of large-scale dust events in the temporal direction provides essential information for reducing the impact of dust on human activities [83]. Therefore, Remote Sens. 2021, 13, 1553 6 of 16 dust monitoring by satellite observations was conducted during the early satellite operation era [84]. Ackerman [85] described the utilization of tree bands in TIR (8.5, 11 (or 10), and 12 µm) for the detection of volcanic and soil-derived aerosols. The satellites have been used to generate RGB composite images (Dust RGB: ∆Tbb 12.4-10.4 µm in red, ∆Tbb 10.4-8.6 µm in green, and Tbb 10.4 µm in blue) by applying thermal infrared 3-band information. Because MSG and third-generation GEOs have achieved multiple bands in the thermal infrared region and increased observation frequencies, many studies on dust monitoring now use the brightness temperature difference in the thermal infrared region [86][87][88][89][90][91][92].
Aerosol optical parameters, such as aerosol optical thickness, can also be retrieved from optical sensors onboard satellites, although only during the daytime and only in areas without cloud cover [93][94][95][96][97][98][99]. The multi-band capability of third-generation GEOs has improved aerosol optical parameter estimation accuracy, and its use is expanding [100][101][102][103][104][105]. The aging of optical sensors' sensitivity, which is essential for aerosol estimation, has been reported [106], and aerosol optical properties are steadily moving into the monitoring phase. Although aerosol optical properties and vertical structure of cloud-covered areas cannot be determined in principle, more detailed aerosol information in the spatio-temporal direction has been successfully derived by using the data assimilation technique as described in Section 2.3.1 [107][108][109]. It is noteworthy that the aerosol observation impact of thirdgeneration GEOs can be derived to a greater extent through more advanced coordination with numerical forecast information.

Volcanic Plumes and Lightning Activity
It is essential to use satellite observations for disaster monitoring and immediate response in monitoring volcanic plumes, which are extremely difficult to predict in advance. Similar to the dust event monitoring described in Section 2.3.2, multi-band monitoring is effective by taking advantage of the aerosol ejection fraction (mostly SO 2 ) emitted in an eruption, depending on the thermal infrared wavelength region [85,110]. It is helpful to summarize what approaches, including satellite observations, are effective for eruption monitoring, using the example of Eyjaföll volcano in Iceland, which erupted from 23 March to 14 April 2010 [111]. There are examples of utilizing the multi-band TIR channels in second-generation GEOs [112,113]. Monitoring examples by third-generation GEOs with increased observation frequency and number of bands, include the eruption of Aso Caldera in Japan on 8 October 2016, by Himawari-8 [114], and the eruption of Raung, Indonesia [115], focusing on the behavior of shorter wavelengths (1.6, 2.3, and 3.9 µm) [116].
From a natural disaster monitoring point of view, such as the effects of lightning strikes on electronic equipment and the causes of spontaneous ignition, lightning activity monitoring using satellites is also effective [117]. Because it can also be used as an indicator of atmospheric upwelling motion [118], a lightning imaging sensor (LIS) was installed on the Tropical Rainfall Measuring Mission (TRMM) [119,120]. The geostationary lightning mapper (GLM) was installed in the GOES-R series [121]. Preliminary observations of lightning captured by GLM have been obtained [122], and there are reports on lightningrelated Fuego eruptions [123].

Vegetation Activity and Forest Fires
The Landsat series was the first satellite to monitor terrestrial environments. The NOAA/AVHRR series of meteorological satellites separated the visible and near-infrared bands in the AVHRR/2 series. Thus, the normalized difference vegetation index (NDVI), which is a widely utilized vegetation index, can be applied. The NOAA/AVHRR could monitor vegetation dynamics globally in the early era of environmental remote sensing [124,125]. The generation of global datasets derived from the NOAA/AVHRR has begun to yield important insights into the response of vegetation to climate change [126,127], the application of satellite data for phenological timings [128], and fundamental studies of vegetation response to temperature and precipitation [129]. Our understanding of global environmental research has been enhanced by using highly accurate products produced by a well-organized MODIS science team for analysis [130]. However, the interpretation of tropical rainforests with frequent cloud cover is, for example, in contrast to the vegetation response of the Amazon rainforest to the 2005 drought [131,132], and increasing the frequency of cloud-free observations is considered to be the key to further understanding of vegetation response, especially in the tropics.
With an increase in the number of observation bands in MSG (Table 1) and an increase in the observation frequency (once every 15 min), the possibility of vegetation monitoring by GEO has greatly expanded. A series of investigations by Fensholt et al. [133][134][135] and others [136] have shown its usefulness. As evidence, the EUMETSAT provides vegetation indices such as leaf area index (LAI) from MSG [137], and expectations for vegetation monitoring by GEO with enhanced functions are high.
In the mid-latitudes, Miura et al. [138] monitored deciduous broadleaf trees in Japan by Himawari-8, and Wheeler and Dietze [139] by the GEOS-R series; this points to a dramatic increase in third-generation GEOs' frequency of observations compared to LEOs (approximately 50 times more). Hashimoto et al. [140] conducted an analysis using GOES-16 on the Amazon rainforest. They demonstrated seasonal changes in NDVI in the Amazon rainforest, which could not be confirmed by conventional LEO, such as MODIS, due to the dramatic increase in the frequency of observations by GOES-R. Key to this result is the geometric correction [38,40,42] and atmospheric correction accuracies including in aerosols. For the latter, the standard atmospheric correction developed in MODIS (multiangle implementation of atmospheric correction (MAIAC) [141]) was adapted to GOES-R, and its usefulness was demonstrated by comparison with the local observation network "AERONET". In terms of aerosol correction, there is room for further improvement in accuracy by utilizing the results obtained in atmospheric research [99,109], but the high computational cost is prohibitive in terms of providing data in quasi-real-time. To overcome this problem, further advances using AI, such as neural networks, are required. In addition, there is a more integrated relationship between GEOs and LEOs, and it is necessary for future applications to make a more sophisticated link between the high temporal frequency and low spatial resolution vegetation information obtained by GEOs, with research that looks at vegetation in a more detailed spatial direction (using LEO satellites such as Landsat, Satellite Pour l'Observation de la Terre (SPOT), and Sentinel [142]). For example, improving the spatial resolution of GEO by coupling the use of LEO optics is the one of potential utilization. Otherwise, based on LEO optics, phenological cycles interpolation with the assist of GEO data is another way.
One of the most effective applications that benefits from GEOs' high temporal resolution and excellent latency to provide data, is forest fire monitoring [117]. There is a long history of forest fire monitoring by satellites [143,144], and the detection algorithms developed in MODIS [145,146] can be easily applied, as well as the algorithms developed in the second-generation GEOs [147]. It has been visualized in the JAXA Himawari Monitor [25] and the GOES-R standard product.

Land Surface Temperature (LST), Heat Islands, and Heatwaves
Estimating land surface temperature (LST) using the split windows method with multi-band NOAA/AVHRR thermal infrared has taken place since the early days when satellite observations began [148][149][150][151]. LST estimation using second-generation GEOs has also been conducted [152,153], and with third-generation GEOs, LST estimation methods that take advantage of the further multi-banding of TIR have been proposed [154][155][156][157]. As the frequency of observations increases dramatically, the removal of cloud pixels [158,159] becomes key to accurate LST estimation. Because GEOs can observe the diurnal cycle of LST at a higher frequency, it effectively analyzes the phenomena that cause daily changes, especially the factors that cause heat islands [160][161][162]. In addition, there is a possibility that third-generation GEOs can provide more detailed information in the time direction of the heatwave [163], which has been occurring increasingly in recent years. With LEOs, only a snapshot can be obtained, but with GEOs, for example, in principle it is possible to predict high LST duration times that can be harmful to the human body.

Landslides and Flooded Area Monitoring
The most acceptable resolution for third-generation GEOs is 500 m at the Nadir, which is equivalent to MODIS. By taking advantage of high-frequency observation characteristics, it is possible to monitor the general state of a large-scale disaster more quickly than any other satellite observation when the weather is clear. Miura and Nagai [164] demonstrated that landslides occurring after record-breaking heavy rainfall in Japan could be predicted by focusing on a rapid decrease in NDVI. The prediction of landslides based on changes in NDVI requires the original NDVI values to be high, and it is difficult to apply this method during the low NDVI periods of spring and fall in deciduous broadleaf forests, but it remains a highly effective approach.
As all third-generation GEOs are equipped with a 1.6 µm band, water-related monitoring, for example, flood area monitoring by calculating the normalized difference water index (NDWI) [165], is also possible in principle. Even without inter-band computation, it is possible to visualize a wide area of the flooded area using a natural color RGB composite by assigning a 1.6 µm band. In MSG, the resolution at 1.6 µm was 4 km, which was rough, but in third-generation GEOs, the resolution is 2 km, even in the 1.6 µm band, so more detailed information can be obtained.

Possible Further Collaboration between GEOs and LEOs-Case Study
Using the example of Typhoon Hagibis in 2019, which caused extensive flooding in Japan, we posed the question, "If we had been able to issue a high-resolution LEO observation request based on GEO information, how quickly would we have been able to acquire high-resolution observation images over the damaged area?" A simple simulation was conducted to answer this question. Please note that this is only a hypothetical check, and the assumptions presented here have not been explored in detail.
Typhoon Hagibis made landfall in Japan on 12 October 2019, reaching a minimum pressure of 915 hPa and causing extensive damage mainly in eastern Japan. This typhoon brought the heaviest daily precipitation recorded since 1982 at 613 comparable JMA observation sites, with a 24-h accumulated precipitation amount reaching 942.5 mm at Hakone. The death toll was 105, and according to the Ministry of Land, Infrastructure, Transport, and Tourism (MLIT), the area inundated by Hagibis reached approximately 25,000 ha in extent, exceeding the 18,500 ha by the extreme heavy rainfall event of July 2018 [166].
We first tried to extract the inundated area under a typhoon situation such as that of the case study, by hypothetically using Himawari-8 data. We used the Himawari-8 gridded FD data [41] provided by CEReS, Chiba University, Japan. Cloud removal used in the LST estimation algorithm [154] was performed, and the NDWI [165] was computed.
To determine the threshold value of the flooded area, we used the map information of the flooded area visually identified by high-resolution optical sensors, and tentatively identified the pixels below −0.2 in the NDWI as the flooded area. Figure 2 shows the geographical map of the flooded area with an NDWI below −0.2 on the day after the Hagibis passage (09:00 Japan Standard Time (JST), 13 October 2019) captured in Himawari-8 true-color composite image from the Himawari real-time website at the same time. It can be seen that there were many misidentifications of inundated areas in the metropolitan urban areas, but the suburban areas appropriately reflected the inundated areas of rivers. The NDWI calculations were available from 06:00 JST on the same day, but due to the low solar elevation, the NDWI did not correctly represent the flooded area (figures are not shown), and although qualitatively, it showed an overestimation from 07:30 JST (figures are not shown). color composite image from the Himawari real-time website at the same time. It can be seen that there were many misidentifications of inundated areas in the metropolitan urban areas, but the suburban areas appropriately reflected the inundated areas of rivers. The NDWI calculations were available from 06:00 JST on the same day, but due to the low solar elevation, the NDWI did not correctly represent the flooded area (figures are not shown), and although qualitatively, it showed an overestimation from 07:30 JST (figures are not shown). Based on the observation information from Himawari-8, we hypothetically assumed the observation requirements for a commercial-based high-resolution optical satellite and a synthetic-aperture radar (SAR) satellite. On the one hand, for optical satellites, requests are accepted up to 03:00 JST on the same day, so in this case, it was difficult to determine the possible flooded area and issue an observation request on the same day (12 October). On the other hand, SAR requests can be accepted up to one hour before the command uplink, so it was possible to request observations by making a quick decision based on the NDWI flooded area information on the morning after typhoon Hagibis hit. This showed that disaster monitoring could be organically linked to LEO observations by using GEO observation information more appropriately than has been done to date.

Closing Remarks and Future Perspectives
A review was conducted focusing on third-generation GEOs, with visualization (RGB full-color composite, visualization via the web interface), baseline dataset, and phenomena to be covered limited to the atmospheric and terrestrial environments. For both atmospheric and terrestrial environments, enhanced capabilities of third-generation GEO were found to provide critical information for disaster management and risk mitigation. In particular, it is essential to reiterate that GEO data make a better contribution to bridging the gap between observation data and virtual data such as numerical forecasts, through the latest technologies such as data assimilation and Artificial Intelligence (AI). 〇：flooded 〇：flooded Based on the observation information from Himawari-8, we hypothetically assumed the observation requirements for a commercial-based high-resolution optical satellite and a synthetic-aperture radar (SAR) satellite. On the one hand, for optical satellites, requests are accepted up to 03:00 JST on the same day, so in this case, it was difficult to determine the possible flooded area and issue an observation request on the same day (12 October). On the other hand, SAR requests can be accepted up to one hour before the command uplink, so it was possible to request observations by making a quick decision based on the NDWI flooded area information on the morning after typhoon Hagibis hit. This showed that disaster monitoring could be organically linked to LEO observations by using GEO observation information more appropriately than has been done to date.

Closing Remarks and Future Perspectives
A review was conducted focusing on third-generation GEOs, with visualization (RGB full-color composite, visualization via the web interface), baseline dataset, and phenomena to be covered limited to the atmospheric and terrestrial environments. For both atmospheric and terrestrial environments, enhanced capabilities of third-generation GEO were found to provide critical information for disaster management and risk mitigation. In particular, it is essential to reiterate that GEO data make a better contribution to bridging the gap between observation data and virtual data such as numerical forecasts, through the latest technologies such as data assimilation and Artificial Intelligence (AI). The improvement of near-future forecast with the assist of GEO data implicitly contributes to disaster risk mitigations. Numerical simulations of hydrological processes using global precipitation datasets as inputs [167,168], which also utilize GEO data such as GSMaP, are also highly effective for risk mitigation of flood damage, and further integration of satellite observation and numerical prediction, not limited to GEO, is an excellent way to mitigate damage caused by disasters.
According to the Vision for the World Meteorological Organization (WMO) Integrated Global Observing System in 2040 [169], there are four types of sensors that should be onboard geostationary meteorological satellites by the year 2040: (1) high-frequency, multi-wavelength imagers or radiometers; (2) hyperspectral infrared sounders; (3) lightning imagers; and (4) ultraviolet, visible, and near-infrared sounders. By achieving this, hyperspectral infrared sounder observations will, for example, be able to provide more frequent and accurate information on the vertical profiles of temperature and water vapor in clear-sky pixels. Using this information, the accuracy of the atmospheric correction of optical sensors can be drastically improved.
The enhancement of functions by third-generation GEOs has resulted in a dramatic expansion in data volume, and some GEOs have started to provide data through cloud services, but even in the context of citizen science, the general use of third-generation GEOs is still low compared to the high level of interest shown in it. Putting aside the issue of who is responsible for associated costs, it is essential to establish a system where the data can be used in a broader range, considering the amount of publicity it has achieved and the need for its immediate use. Data Availability Statement: The data analyzed in Figures 1 and 2 are partly available. Himawari-8 base gridded data can be accessed via the Internet [41]. The Himawari-8 analyzed dataset, the socalled AMATERASS, can be accessed via the NPO solar radiation consortium (http://www.amaterass. org/, accessed on 19 March 2021). XRAIN data can also be accessed via the Data Integration and Analysis System Program (DIAS) (https://diasjp.net/en/, accessed on 19 March 2021). PAWR data were obtained from Japan Radio Co. (JRC) Ltd., with data policies decided by the JRC. Validation data for determination of the NDWI threshold were obtained from PASCO with data policies decided by PASCO.