Himawari-8 Advanced Himawari Imager Research Articles

Time series retrieval of Multi-wavelength Aerosol optical depth by adapting Transformer (TMAT) using Himawari-8 AHI data

Accurate aerosol optical depth (AOD) with high temporal resolution is required for dynamic monitoring changes of aerosol properties. The Himawari-8 Advanced Himawari imager (AHI) data with 10-min temporal resolution has been widely used for AOD retrieval but usually involves radiative transfer models and assumptions on surface reflection and aerosol properties that are hard to satisfy. This paper introduces a data-driven algorithm without any priori assumptions on surface or aerosol properties called Time series retrieval of Multi-wavelength AOD by adapting Transformer (TMAT). TMAT directly retrieves AOD at eight-wavelengths every 10-min throughout the day, using AHI top of atmosphere (TOA) reflectances in 6 wavebands and ancillary data. TMAT explores diurnal temporal change patterns of AOD and surface reflection using time series data. TMAT was trained and evaluated using samples derived by matching a year (2019) of AHI full-disk data and AERONET/SONET observations with 12,768 diurnal time series, each of which includes up to 81 10-min 6-band AHI TOA observations and 81 (time) × 8 (wavelength) AOD values per day. There are up to 273,413 (<12,768 × 81) TOA observations in 6 wavebands and up to 1,420,716 (<12,768 × 81 × 8) AOD values in all the time series samples, with filled values (−9999) to represent no data due to cloud contamination and incomplete wavelength configurations of AOD stations. The masking mechanism in Transformer is leveraged to exclude the contribution of the AHI filled values in time series AOD retrieval and exclude the contribution of AOD filled values in training. TMAT was trained with a physical constraint that the retrieved AOD decreases toward longer wavelengths. TMAT AOD was evaluated by separating training and validation samples using site-specific leave-one-station-out validation. TMAT AOD shows extremely high accuracy at all wavelengths, with RMSE = 0.101, R = 0.949 and 81.69% of retrievals within ±0.05+15%∗AODsites for the AOD at 550 nm. We also validated the accuracy of the NASA multiangle implementation of atmospheric correction (MAIAC) AHI AOD over these sites with RMSE = 0.186, R = 0.832 and 52.10% of retrievals within ±0.05+15%∗AODsites. The Ångström exponent (AE) derived from the multi-wavelength TMAT AOD also shows good agreement with AERONET/SONET AE, with a correlation coefficient of 0.620 and an RMSE of 0.291. The TMAT effectiveness to capture diurnal temporal change patterns of surface reflection and AOD was evidenced by the significant reduction of AOD accuracy when using a non-time series deep learning method. The effectiveness of multi-wavelength joint retrieval was evidenced by the significant reduction of AE accuracy when using single wavelength retrieval models. Evidence shows the training and evaluation split methods have a large impact on evaluation accuracy and this study further used ground network independent training and evaluation split to confirm the leave-one-station-out validation accuracy. Despite this, future study will focus on more spatially independent validation and more comprehensive comparison between the MAIAC and TMAT results.

Probabilistic Convective Initiation Nowcasting Using Himawari-8 AHI with Explainable Deep Learning Models

Abstract Convective initiation (CI) nowcasting is crucial for reducing loss of human life and property caused by severe convective weather. A novel deep learning method based on the U-Net model (named as CIUnet) was developed for forecasting CI during the warm season with eight interest fields of Himawari-8 Advanced Himawari Imager (AHI) and terrain height. The results showed that the CIUnet model produced probability forecasts of CI occurrence location and time with probability of detection (POD) at 93.3% ± 0.3% and false alarm ratio (FAR) at 18.3% ± 0.4% at a lead time of 30 min. Sensitivity and permutation importance experiments on the input fields of the CIUnet model revealed that the differences in brightness temperature for spectral channels were more critical for CI nowcasts than the original infrared channel brightness temperatures. The brightness temperature difference between band 10 (7.3 μm) and band 13 (10.4 μm), which represents the cloud-top height relative to the lower troposphere, is identified as the most important input fields for CI nowcasting. The tri-spectral brightness temperature difference (TTD), which represents cloud-top glaciation, is ranked the second and it significantly reduced the FAR of the CI forecast. Using terrain heights as an extra input feature improved the POD, but slightly overestimated CI over complex terrain. In addition, a layer-wise relevance propagation (LRP) analyses was performed, and confirmed that the CIUnet model can effectively identify the crucial regions and features of the input fields for accurate CI prediction. Therefore, both permutation importance experiments and LPR analyses are useful for improving the CIUnet model and advancing the understanding of CI mechanisms.

Combined assimilation of hourly rainfall data and every 10-min AHI radiance with WRF 4DVar for the short-range heavy rainfall forecast in Eastern China

Rainfall observations and geostationary satellite water vapor radiances account for complementary humidity information over rainy areas and large-scale environments, respectively. This study investigates the simultaneous assimilation of China Hourly Merged Precipitation Analysis (CHMPA) data and clear-sky water vapor radiances provided by the Himawari-8 Advanced Himawari Imager (AHI) with Four-Dimensional Variational (4DVar) data assimilation method using Weather Research and Forecasting (WRF) model with an hourly rapid-update cycling data assimilation configuration. The rapid-update 4DVar assimilation is configured as 10-min interval for AHI radiances and 1-h interval for CHMPA rainfall data in a same 1-h assimilation time window. Generally positive impacts from assimilating CHMPA rainfall and AHI water vapor radiance observations are achieved for the 12-h short-term precipitation forecast. Error reductions are found after assimilating CHMPA rainfall and AHI water vapor radiance observations for both analyses and forecasts when compared to ERA5 reanalysis. It is found that combined assimilation of CHMPA rainfall and AHI water vapor radiance data has clearly positive impacts on heavy rainfall forecasts. A heavy rainfall case that occurred in Jiangsu Province on 28 June 2020 is examined in details. The combined assimilation of infrared radiances from AHI water vapor channels together with CHMPA rainfall data improves the vertical and horizontal transport of humidity conditions, which better adjusts the humidity distribution. The over-estimation of rainfall forecast caused by the assimilation of CHMPA is alleviated. This result can provide reference for quantitative precipitation forecasts of regional numerical weather model.

The China Coastal Front from Himawari-8 AHI SST Data—Part 1: East China Sea

High-resolution (2 km) high-frequency (hourly) SST data from 2015–2020 provided by the Advanced Himawari Imager (AHI) onboard the Japanese Himawari-8 geostationary satellite positioned over 140.7°E were used to study spatial and temporal variability of the China Coastal Front (CCF) in the East China Sea. This dataset was processed with the Belkin and O’Reilly algorithm to generate long-term mean monthly maps of the SST gradient magnitude (GM) and frontal frequency (FF). The horizontal structure of the SST field in the vicinity of the CCF was also investigated from the cross-frontal distributions of SST along eight parallels between 31°N and 24°N. The monthly mean distributions of SST along these 8 parallels were used to determine inshore and offshore boundaries of the CCF and to calculate the CCF strength defined as the total cross-frontal step (range) dSST calculated as the difference between offshore and inshore SST. The CCF emerges in November, fully develops in December, and peaks in strength in January–February. The front’s fragmentation and shrinking/weakening begins in February and March, respectively. In winter (December–February), the front’s strength dSST exceeds 5 °C offshore the Zhejiang-Fujian coast and could be as high as 7.5 °C when nearshore waters cool down to 7 °C. In winter, the front’s strength decreases downstream between 31°N and 24°N. The CCF changes its physical nature as the seasons progress. In winter, the CCF is a water mass front between the cold and fresh water coming from the north and the warm and salty water coming from the south. In summer, the CCF becomes a coastal upwelling front maintained largely by southerly winds. In winter, the CCF’s cross-frontal structure in the SST field is ramp-shaped, with SST increasing monotonously in the offshore direction. In summer, the CCF’s cross-frontal structure in the SST field is V-shaped or U-shaped, featuring a minimum SST formed by cold upwelled water at some distance from the shore. Thus, the summer SST structure effectively consists of two parallel fronts, an inshore one and an offshore one, with a minimum SST in between. Across the inshore/offshore front, the SST decreases/increases in the offshore direction.

How useful are the lake surface temperature estimates from a geostationary satellite (Himawari-8) to detect seasonal and diurnal changes?

ABSTRACT The current geostationary satellite images have a time resolution of minutes. It is not clear how useful they are for detecting diurnal and seasonal changes. We tested this by comparing lake surface temperatures (LSTs) from Himawari-8 Advanced Himawari Imager (AHI) images with in situ measurements in Lake Kasumigaura, Japan, for one year. LST products of the Sentinel-3 satellites were also compared. We found that instantaneous LSTs can be estimated with an root mean square difference (RMSE) of 0.6°C. A large amount of data (N = 102–103) is available for each month, but N differs greatly in the range of 216–2771. A good agreement (RMSE of 0.9°C) was found between monthly mean LSTs from Himawari-8 and in situ measurements, and it is better than the RMSE of 1.2–1.7°C found for the Sentinel-3 LST products. The diurnal change detection was more difficult due to insufficient data numbers around noon. Practical remedies for this issue are suggested.

A Case Study of a West Sumatra Squall Line Using Satellite Observations

Abstract A West Sumatra squall line occurred on 10 January 2016, with a clear offshore propagation of convection. Satellite-derived products from Himawari-8 Advanced Himawari Imager and the Geostationary Cloud Algorithm Testbed Geocat are used to investigate the westward propagation of cloudiness from Sumatra to the Indian Ocean with a lifetime of 1.5 days. A convective mask based on deep convective cell detection and a cell-tracking algorithm are used to estimate the propagation speed of the cloud system. Two distinct mesoscale convective responses are identified: 1) a rapid development in South Sumatra is influenced by the convective environment over the Indian Ocean. The propagation speed is estimated to be ∼5 m s−1 within the first 200 km from the coast. This speed is consistent with density currents. In contrast, 2) the coupling to the inertia–gravity wave is only evident for the northwest of Sumatra with speeds of ∼12 m s−1. The analysis of brightness temperature from the 10.4-μm spectral band and cloud-top temperature showed that the lifetime of the squall line is approximately 30 h with a propagating distance of ∼1000 km. Retrieved cloud properties and tracking of the offshore propagation indicated that the cloud structure consisted of multiple types of cells, propagating as envelopes of convection, and revealed the influence of large-scale variability of the Indian Ocean. Filtered OLR anomalies, satellite-derived rainfall, moisture flux convergence, and background winds flow around Sumatra are used to explore the effects of Kelvin wave activity that likely influenced the lifetime of the squall line.

Internal Wave Signature Extraction From SAR and Optical Satellite Imagery Based on Deep Learning

Internal waves (IWs) are a common characteristic of oceans and serve a crucial role in transmitting energies between large-scale tides and small-scale mixing. This study developed a deep-learning-based method for extracting IW signatures on multiple satellite imagery from synthetic aperture radar (SAR) and optical sensors in sun-synchronous and geostationary orbits with varying spatial resolution. We collected 1115 satellite images, including 116 ENVISAT ASAR (Advanced SAR), 839 MODIS (MODerate-resolution Imaging Spectroradiometer), and 160 Himawari-8 AHI (Advanced Himawari Imager) images with clear IW signatures in the South China Sea (SCS), Sulu Sea, and Celebes Sea for model training. Considering the distinct IW characteristics under different imaging mechanisms, the specially tailored IW Extraction network (IWE-Net) leverages three modifications to improve the accuracy and robustness: online data augmentation, squeeze and excitation blocks, and Matthews correlation coefficient loss. The overall mean Precision, Recall, and F1-score of the IWE-Net model are 85.75%, 85.67%, and 85.71%, demonstrating the model is accurate for IW signature extraction.We also proved the transferability of our method to sea areas worldwide, long-term periods, and Sentinel satellite sensors completely independent of the model training. Globally, the number of IW images and extracted pixels show an obvious tidal-related double-peak distribution. Furthermore, we processed 15461 MODIS images in the northeastern SCS to present a holistic IW distribution map over the past 22 years. An unreported IW silent zone caused by drastic topography changes has been discovered, indicating the great potential of deep learning in information retrieval from remote sensing imagery.

Early Stage Forest Fire Detection from Himawari-8 AHI Images Using a Modified MOD14 Algorithm Combined with Machine Learning

The early detection and rapid extinguishing of forest fires are effective in reducing their spread. Based on the MODIS Thermal Anomaly (MOD14) algorithm, we propose an early stage fire detection method from low-spatial-resolution but high-temporal-resolution images, observed by the Advanced Himawari Imager (AHI) onboard the geostationary meteorological satellite Himawari-8. In order to not miss early stage forest fire pixels with low temperature, we omit the potential fire pixel detection from the MOD14 algorithm and parameterize four contextual conditions included in the MOD14 algorithm as features. The proposed method detects fire pixels from forest areas using a random forest classifier taking these contextual parameters, nine AHI band values, solar zenith angle, and five meteorological values as inputs. To evaluate the proposed method, we trained the random forest classifier using an early stage forest fire data set generated by a time-reversal approach with MOD14 products and time-series AHI images in Australia. The results demonstrate that the proposed method with all parameters can detect fire pixels with about 90% precision and recall, and that the contribution of contextual parameters is particularly significant in the random forest classifier. The proposed method is applicable to other geostationary and polar-orbiting satellite sensors, and it is expected to be used as an effective method for forest fire detection.

A deep learning model using geostationary satellite data for forest fire detection with reduced detection latency

ABSTRACT Although remote sensing of active fires is well-researched, their early detection has received less attention. Additionally, simple threshold approaches based on contextual statistical analysis suffer from generalization problems. Therefore, this study proposes a deep learning-based forest fire detection algorithm, with a focus on reducing detection latency, utilizing 10-min interval high temporal resolution Himawari-8 Advanced Himawari Imager. Random forest (RF) and convolutional neural network (CNN) were utilized for model development. The CNN model accurately reflected the contextual approach adopted in previous studies by learning information between adjacent matrices from an image. This study also investigates the contribution of temporal and spatial information to the two machine learning techniques by combining input features. Temporal and spatial factors contributed to the reduction in detection latency and false alarms, respectively, and forest fires could be most effectively detected using both types of information. The overall accuracy, precision, recall, and F1-score were 0.97, 0.89, 0.41, and 0.54, respectively, in the best scheme among the RF-based schemes and 0.98, 0.91, 0.63, and 0.74, respectively, in that among the CNN-based schemes. This indicated better performance of the CNN model for forest fire detection that is attributed to its spatial pattern training and data augmentation. The CNN model detected all test forest fires within an average of 12 min, and one case was detected 9 min earlier than the recording time. Moreover, the proposed model outperformed the recent operational satellite-based active fire detection algorithms. Further spatial generality test results showed that the CNN model had reliable generality and was robust under varied environmental conditions. Overall, our results demonstrated the benefits of geostationary satellite-based remote sensing for forest fire monitoring.

A Data Fusion Method for Generating Hourly Seamless Land Surface Temperature from Himawari-8 AHI Data

High temporal resolution and spatially complete (seamless) land surface temperature (LST) play a crucial role in numerous geoscientific aspects. This paper proposes a data fusion method for producing hourly seamless LST from Himawari-8 Advanced Himawari Imager (AHI) data. First, the high-quality hourly clear-sky LST was retrieved from AHI data by an improved temperature and emissivity separation algorithm; then, the hourly spatially complete China Land Data Assimilation System (CLDAS) LST was calibrated by a bias correction method. Finally, the strengths of the retrieved AHI LST and bias-corrected CLDAS LST were combined by the multiresolution Kalman filter (MKF) algorithm to generate hourly seamless LST at different spatial scales. Validation results showed the bias and root mean square error (RMSE) of the fused LST at a finer scale (0.02°) were −0.65 K and 3.38 K under cloudy sky conditions, the values were −0.55 K and 3.03 K for all sky conditions, respectively. The bias and RMSE of the fused LST at the coarse scale (0.06°) are −0.46 K and 3.11 K, respectively. This accuracy is comparable to the accuracy of all-weather LST derived by various methods reported in the published literature. In addition, we obtained the consistent LST images across different scales. The seamless finer LST data over East Asia can not only reflect the spatial distribution characteristics of LST during different seasons, but also exactly present the diurnal variation of the LST. With the proposed method, we have produced a 0.02° seamless LST dataset from 2016 through 2021 that is freely available at the National Tibetan Plateau Data Center. It is the first time that we can obtain the hourly seamless LST data from AHI.

Improved estimation of fire particulate emissions using a combination of VIIRS and AHI data for Indonesia during 2015–2020

Smoke aerosol emissions from Indonesian fires frequently cause adverse environmental consequences across southeast Asia. Satellite observations provide us with a great opportunity to monitor such emissions at large scales. However, existing satellite-based estimates of Indonesian fire emissions vary considerably in magnitude, differing by a factor of four. Here, we aim to improve Indonesian fire emissions estimates through improved calculations of fire radiative energy (FRE: time-integrated fire radiative power (FRP)) and smoke aerosol emission coefficients (Ce) using multiple new-generation satellite observations. Specifically, peatland and non-peatland Ce values were derived from FRP and emission rates of smoke aerosols based on Visible Infrared Imaging Radiometer Suite (VIIRS) active fire and aerosol products. FRE was calculated from the diurnal FRP cycle that was reconstructed by fusing cloud-corrected FRP retrievals from the high temporal-resolution Himawari-8 Advanced Himawari Imager (AHI) with those from high spatial-resolution VIIRS. Then, fuel type-specific Ce values and fused AHI-VIIRS FRE were used to produce hourly and daily fire emission data from 2015 to 2020 across Indonesia. To evaluate AHI-VIIRS estimates, we generated a reference dataset of total particulate matter (TPM) by computing Moderate Resolution Imaging Spectroradiometer (MODIS) aerosol observation differences from successive Terra and Aqua satellites overpasses over a set of manually-selected smoke plumes. AHI-VIIRS-based fire emissions correlated significantly with the MODIS reference data (r = 0.84; p < 0.001). The interannual time series analysis of AHI-VIIRS emissions showed extreme variability (~26 fold), with the greatest amount in 2015 (6.09 Tg) and the least in 2020 (0.23 Tg), during the study period. Mean annual fire emissions distributed across Indonesia's islands were primarily from Kalimantan (43.8%) and Sumatra (41.7%), followed by Sulawesi (5.6%), Java (4.5%), and Papua (4.4%). Additionally, contributions of Indonesian fire emissions from peatland, forest, cropland, and savanna/grassland fuel types were 51.4%, 37.8%, 7.0%, and 3.8%, respectively, during the fire season of the strong 2015 El Niño event. The analysis suggests that the majority of Indonesian fire emissions are very likely associated with large-scale land use conversion from peatland to agriculture, as well as prolonged droughts induced by El Niño events.

Evaluating the potential of H8/AHI geostationary observations for monitoring vegetation phenology over different ecosystem types in northern China

• Different strategies for compositing daily EVI2 from 10-min AHI data were assessed. • Mean value compositing using data around solar noon minimized anisotropic effects. • Composited AHI daily EVI2 time series have 5.3 times more observations than MODIS. • Phenology dates were successfully retrieved from composited AHI EVI2 time series. • AHI can reduce uncertainty in phenology monitoring with more valid data and less gaps. Sensors onboard the new generation of geostationary (GEO) satellites launched over past few years have much improved radiometric and temporal resolutions. As one of the latest GEO sensors, the Japanese Himawari-8 Advanced Himawari Imager (hereafter H8/AHI) has temporal resolution of 10 minutes and spectral bands similar to MODIS. Observations from H8/AHI therefore have the potential to improve the precision of vegetation phenology monitoring by significantly shortening the compositing period. In this study, we evaluated methods for compositing daily 2-band Enhanced Vegetation Index (EVI2) from 10-minutes H8/AHI observations at six study sites in northern China encompassing forest, cropland, and grassland. After screening the cloudy observations by a physics-based cloud detection algorithm, key phenological metrics were retrieved from the composited AHI EVI2. H8/AHI 10-min raw EVI2 showed significant diurnal variations that are mainly associated with sun angle variations. By calculating the daily average EVI2 values in different time intervals, we concluded that the mean value using a fixed time window (10:30–13:30) centered around local solar noon effectively reduced the anisotropy effect from the diurnal sun-angle variations, leading to stable EVI2 time series suitable for extracting phenological metrics. On average, the amount of the composited H8/AHI EVI2 time-series data was 5.3 times more than the Terra/Aqua combined MODIS EVI2 product per year and the average size (width) of the gaps in H8/AHI EVI2 time series was 75% smaller than that of the MODIS. Our results demonstrated the promising capability of the new generation GEO satellite for generating time series vegetation index with fewer cloud-induced gaps and higher temporal resolution, hence improving the monitoring of global vegetation phenology and ecosystem responses to climate change.

FENGYUN-4A Advanced Geosynchronous Radiation Imager Layered Precipitable Water Vapor Products’ Comprehensive Evaluation Based on Quality Control System

A physical retrieval algorithm has been developed for deriving the layered precipitable water vapor (LPWs) product from infrared radiances of the Advanced Geosynchronous Radiation Imager (AGRI) onboard FengYun-4A (FY-4A), the first of the new generation of Chinese geostationary weather satellites (FengYun-4, or FY-4 Series). The FY-4A AGRI LPWs are evaluated with different types of reference datasets based on Quality Control System (QCS), including those from Himawari-8 AHI (Advanced Himawari Imager), MODIS (Moderate Resolution Imaging Spectroradiometer), Radiosonde, ERA5 (European Centre for Medium-Range Weather Forecasts Reanalysis v5), NCEP (National Centers for Environmental Prediction) reanalysis and CMA (China Meteorological Administration) forecast product from global medium range numerical weather prediction (NWP) system. QCS is one of the important components of FY-4A ground segment, which mainly focuses on the satellite products’ evaluation and validation. It is found that the AGRI LPW product has a good agreement with different evaluating sources and the quality is favorable and stable. With the capability of frequent (5-min interval) observations over the East Asia and Western Pacific regions, the AGRI LPW products can be used to investigate the atmospheric temporal and spatial variations in the pre-landfall environment for typhoons. The QCS is a useful tool to monitor, evaluate, and validate the AGRI LPW products.

Estimating Surface BRDF/Albedo Over Rugged Terrain Using an Extended Multisensor Combined BRDF Inversion (EMCBI) Model

Land surface albedo is a crucial variable of earth energy budget and global climate change. Rugged terrain significantly impacts surface bidirectional reflectance distribution function (BRDF) and the subsequent albedo retrieval using satellite remote sensing. Existing studies of estimating surface BRDF/albedo from satellite observations are limited to neglecting topographic impacts, resulting in large uncertainty in satellite albedo product, especially for low spatial resolution satellite sensors that are primarily regulated by subpixel-scale topographic effects. To fill this knowledge gap, we proposed an extended multisensor combined BRDF inversion (EMCBI) model to characterize subpixel-scale topographic effects, and applied this model to estimate BRDF/albedo from the Himawari-8 Advanced Himawari Imager (AHI) and Terra/Aqua moderate resolution imaging spectroradiometer (MODIS) data and finally validated the satellite-derived albedo with ground measurements of two stations located in Tibet plateau. Our results show that: 1) EMCBI can generate a daily BRDF/albedo dataset with more than 90% spatial coverage and 2) EMCBI-derived albedo agrees well with the referenced albedo corrected from ground measurement, with a root-mean-square-error (RMSE) of 0.0537 and 0.0608 for black-sky albedo (BSA) and white-sky albedo (WSA), and a mean absolute percentage error (MAPE) of 21.93% and 25.13% for BSA and WSA, respectively. These results demonstrate EMCBI has great potential for mapping large-scale high temporal resolution BRDF/albedo product over rugged terrain.

Characterization and monitoring of GOES-16 ABI stray light and comparison with Himawari-8 AHI and GOES-17 ABI

The Advanced Baseline Imager (ABI) aboard Geostationary Operational Environmental Satellite (GOES)-16 and -17 satellites represent the next-generation geostationary multispectral imaging instrument. Since GOES-16 ABI imagery data became available, stray light was observed in ABI visible, near-infrared (VNIR), and 3.9 μm, i.e., CH07, channels. A stray-light characterization scheme was developed to quantitatively monitor stray-light variation in ABI imagery. The stray-light analysis is focused on ABI CH07, whose nighttime radiometric performance being impacted by stray light is of main concern. It is found that the stray light in the ABI imagery occurs over ∼3 months around spring and fall equinox each year. The maximum stray light of GOES-16 ABI CH07 is ∼0.65 K at 300 K scene in zone of normal performance (ZONP), i.e., region with relative solar angle >7.5 deg, which is within the radiometric requirement of 1 K. The analysis of Himawari-8 Advanced Himawari Imager (AHI) CH07 data indicates that its maximum stray light is ∼3.35 K at 300 K scene in ZONP, much higher than that of GOES-16. This confirms the effectiveness of reducing the major stray-light leaking path in ABI as a result of lessons learned from Himawari-8 AHI. The magnitude of GOES-17 ABI CH07 stray light is shown to be ∼0.45 K in ZONP, slightly lower than GOES-16. The analysis of AHI 3.9 and 6.2 μm channel stray-light radiance ratio is shown to match the ratio of solar irradiance spectrum, which suggests that the stray-light issue is due to direct leakage of solar radiation through the instrument. Further characterization and monitoring of GOES-16 ABI VNIR channel stray light also help understand the solar origin of ABI stray light. This paper also investigated cases of strong atmospheric refraction-induced stray-light contamination onto ABI detectors during solar eclipse seasons and presents scheme to reduce such potentially harmful contamination.

Himawari-8 Aerosol Optical Depth (AOD) Retrieval Using a Deep Neural Network Trained Using AERONET Observations

Spectral aerosol optical depth (AOD) estimation from satellite-measured top of atmosphere (TOA) reflectances is challenging because of the complicated TOA-AOD relationship and a nexus of land surface and atmospheric state variations. This task is usually undertaken using a physical model to provide a first estimate of the TOA reflectances which are then optimized by comparison with the satellite data. Recently developed deep neural network (DNN) models provide a powerful tool to represent the complicated relationship statistically. This study presents a methodology based on DNN to estimate AOD using Himawari-8 Advanced Himawari Imager (AHI) TOA observations. A year (2017) of AHI TOA observations over the Himawari-8 full disk collocated in space and time with Aerosol Robotic Network (AERONET) AOD data were used to derive a total of 14,154 training and validation samples. The TOA reflectance in all six AHI solar bands, three TOA reflectance ratios derived based on the dark-target assumptions, sun-sensor geometry, and auxiliary data are used as predictors to estimate AOD at 500 nm. The DNN AOD is validated by separating training and validation samples using random k-fold cross-validation and using AERONET site-specific leave-one-station-out validation, and is compared with a random forest regression estimator and Japan Meteorological Agency (JMA) AOD. The DNN AOD shows high accuracy: (1) RMSE = 0.094, R2 = 0.915 for k-fold cross-validation, and (2) RMSE = 0.172, R2 = 0.730 for leave-one-station-out validation. The k-fold cross-validation overestimates the DNN accuracy as the training and validation samples may come from the same AHI pixel location. The leave-one-station-out validation reflects the accuracy for large-area applications where there are no training samples for the pixel location to be estimated. The DNN AOD has better accuracy than the random forest AOD and JMA AOD. In addition, the contribution of the dark-target derived TOA ratio predictors is examined and confirmed, and the sensitivity to the DNN structure is discussed.

Evaluation and possible uncertainty source analysis of JAXA Himawari-8 aerosol optical depth product over China

The latest version of JAXA aerosol optical depth (AOD) product over China in 2018 from Himawari-8 Advanced Himawari Imager (AHI) was evaluated, including validation with global AErosol RObotic NETwork (AERONET) and China Aerosol Research Network (CARSNET) ground-based observations, and comparison against the Moderate Resolution Imaging Spectroradiometer (MODIS) AOD product. Considering both the quality and quantity of the AHI AOD product, the Quality Assurance Flags (QA_Flags) with “good” and “very good” are chosen for evaluation and analysis. AHI and MODIS AOD products are well correlated and the spatial distributions are similar, even though underestimation from AHI AOD in the morning and overestimation in the afternoon are found when compared with MODIS AOD. Validation with ground-based observations also indicates a good correlation (R > 0.75) when sparsely vegetated surfaces are excluded. It is found that 52.74% fall below the expected error (EE) at 10:30 BJT when compared with CARSNET which has more vegetation sites and 56.91% fall above the EE at 13:30 BJT when compared with AERONET which has more urban sites. The slope of the fit line is around 1.0 for all matched sites, while the scatter plots have divided into two parts as the performance of AHI AOD depends on the observation times and surface types. The temporal variations of AHI and AERONET AODs show that the overestimation from AHI AOD occurs in the afternoon for barren surface while the underestimation occurs in the morning for well-vegetation surface. For monthly variation, the overestimation of AHI AOD is found over Northern China in winter and those seasons when the plants are growing or disappearing rapidly.The assumption that the surface reflectance remains unchanged and the solar geometries are nearly identical within a month causes the underestimation of surface reflectance and thus contributes to the AHI AOD overestimation for low vegetation and bare surfaces in autumn and winter. While over the vegetated surfaces, neglecting the background aerosols causes overestimation of the surface reflectance, and thus leads to underestimation of AHI AOD. The difference between backward and forward scattering contributes to the larger JAXA AHI AOD error in the afternoon, which aggravates positive deviation for barren land sites and alleviates the underestimation for vegetation surface types in the afternoon. In addition, aerosol model selection and cloud contamination also affect the accuracy of aerosol products, especially when AOD is large.Overall, the JAXA AHI AOD provides a high temporal resolution and relative reliable aerosol product for environmental research. The analysis on characteristics and uncertainty sources in this study are to further enhance the understanding of the geostationary satellite aerosol products for applications, and provide useful information for improving the aerosol retrieval algorithm in future.

An Operational Split-Window Algorithm for Retrieving Land Surface Temperature from Geostationary Satellite Data: A Case Study on Himawari-8 AHI Data

An operational split-window (SW) algorithm was developed to retrieve high-temporal-resolution land surface temperature (LST) from global geostationary (GEO) satellite data. First, the MODTRAN 5.2 and SeeBor V5.0 atmospheric profiles were used to establish a simulation database to derive the SW algorithm coefficients for GEO satellites. Then, the dynamic land surface emissivities (LSEs) in the two SW bands were estimated using the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) Global Emissivity Dataset (GED), fractional vegetation cover (FVC), and snow cover products. Here, the proposed SW algorithm was applied to Himawari-8 Advanced Himawari Imager (AHI) observations. LST estimates were retrieved in January, April, July, and October 2016, and three validation methods were used to evaluate the LST retrievals, including the temperature-based (T-based) method, radiance-based (R-based) method, and intercomparison method. The in situ night-time observations from two Heihe Watershed Allied Telemetry Experimental Research (HiWATER) sites and four Terrestrial Ecosystem Research Network (TERN) OzFlux sites were used in the T-based validation, where a mean bias of −0.70 K and a mean root-mean-square error (RMSE) of 2.29 K were achieved. In the R-based validation, the biases were 0.14 and −0.13 K and RMSEs were 0.83 and 0.86 K for the daytime and nighttime, respectively, over four forest sites, four desert sites, and two inland water sites. Additionally, the AHI LST estimates were compared with the Collection 6 MYD11_L2 and MYD21_L2 LST products over southeastern China and the Australian continent, and the results indicated that the AHI LST was more consistent with the MYD21 LST and was generally higher than the MYD11 LST. The pronounced discrepancy between the AHI and MYD11 LST could be mainly caused by the differences in the emissivities used. We conclude that the developed SW algorithm is of high accuracy and shows promise in producing LST data with global coverage using observations from a constellation of GEO satellites.

Seasonal Comparisons of Himawari-8 AHI and MODIS Vegetation Indices over Latitudinal Australian Grassland Sites

The Advanced Himawari Imager (AHI) on board the Himawari-8 geostationary (GEO) satellite offers comparable spectral and spatial resolutions as low earth orbiting (LEO) sensors such as the Moderate Resolution Imaging Spectroradiometer (MODIS) and Visible Infrared Imaging Radiometer Suite (VIIRS) sensors, but with hypertemporal image acquisition capability. This raises the possibility of improved monitoring of highly dynamic ecosystems, such as grasslands, including fine-scale phenology retrievals from vegetation index (VI) time series. However, identifying and understanding how GEO VI temporal profiles would be different from traditional LEO VIs need to be evaluated, especially with the new generation of geostationary satellites, with unfamiliar observation geometries not experienced with MODIS, VIIRS, or Advanced Very High Resolution Radiometer (AVHRR) VI time series data. The objectives of this study were to investigate the variations in AHI reflectances and normalized difference vegetation index (NDVI), enhanced vegetation index (EVI), and two-band EVI (EVI2) in relation to diurnal phase angle variations, and to compare AHI VI seasonal datasets with MODIS VIs (standard and sun and view angle-adjusted VIs) over a functional range of dry grassland sites in eastern Australia. Strong NDVI diurnal variations and negative NDVI hotspot effects were found due to differential red and NIR band sensitivities to diurnal phase angle changes. In contrast, EVI and EVI2 were nearly insensitive to diurnal phase angle variations and displayed nearly flat diurnal profiles without noticeable hotspot influences. At seasonal time scales, AHI NDVI values were consistently lower than MODIS NDVI values, while AHI EVI and EVI2 values were significantly higher than MODIS EVI and EVI2 values, respectively. We attributed the cross-sensor differences in VI patterns to the year-round smaller phase angles and backscatter observations from AHI, in which the sunlit canopies induced a positive EVI/ EVI2 response and negative NDVI response. BRDF adjustments of MODIS VIs to solar noon and to the oblique view zenith angle of AHI resulted in strong cross-sensor convergence of VI values (R2 &gt; 0.94, mean absolute difference &lt;0.02). These results highlight the importance of accounting for cross-sensor observation geometries for generating compatible AHI and MODIS annual VI time series. The strong agreement found in this study shows promise in cross-sensor applications and suggests that a denser time series can be formed through combined GEO and LEO measurement synergies.

Landslide Detection with Himawari-8 Geostationary Satellite Data: A Case Study of a Torrential Rain Event in Kyushu, Japan

In this study, we investigated the utility of Himawari-8 Advanced Himawari Imager (AHI), one of third-generation geostationary satellite sensors, for mapping landslides caused by torrential rain that hit the northern Kyushu area in Japan in the summer of 2017. AHI normalized difference vegetation index (NDVI) time series data had distinctive temporal signatures over landslide areas where the NDVI abruptly decreased after the rain event. The observed changes in the NDVI were linearly correlated with the percent landslide area, the percentage of landslide areas within the AHI pixel footprint, obtained with aerial survey (r = 0.78). AHI 10 min resolution data obtained near cloud-free coverage of the landslide region by the 8th day after the disaster event. This was comparable to the amount of time it took to obtain near cloud-free image coverage with aerial survey, and better than those with the polar-orbiting satellite sensors of Suomi National Polar-orbiting Partnership Visible Infrared Imaging Radiometer Suite, Landsat-8 Operational Land Imager, and Sentinel-2A/B MultiSpectral Instrument. These results suggest that third-generation geostationary satellite data can serve as another useful resource for post-event, region-wide initial assessment of landslide areas after a heavy rain event.