In Indonesia, extensive tropical peatland has been facing land degradation due to deforestation and drainage canal construction. Such activities lead to the decrease of groundwater level (GWL), accelerating the peat decomposition, followed by peatland subsidence. This study addresses to estimate a three-year (from Jan. 2018 to Jan. 2020) tropical peatland subsidence over Kalimantan, Indonesia, using the time-series interferometric synthetic aperture radar (TlnSAR) technique by Sentinel-IA. TlnSAR analysis revealed the apparent large displacement found in 2019 due to a significant decline of GWL caused by positive Indian-Ocean dipole mode (IOD) event compared to other years. Furthermore, we employed GWL derived from satellite-based remote sensing data to investigate the relationship between subsidence and GWL and showed a mutual relationship.

Immediate recognition of landslides after a disaster is essential in assessing the consequent hazards and risks present in the affected areas. Landslide mapping has been done accurately through manual interpretation of high-resolution aerial imagery. However, to speed up the process of landslide delineation, several studies developed semi-automated techniques using object-based image analysis (OBIA) for landslide recognition. Unlike pixel-based classification, OBIA considers the textural, morphological, and contextual characteristics of image objects generated when clustering homogenous pixels. In this study, a semi-automated method of landslide detection using OBIA and a machine learning algorithm, the support vector machine (SVM) is proposed to delineate the landslides that occurred in Eastern Hiroshima, Japan after the 2018 Typhoon Prapiroon devastated the area. The proposed method generated 494 landslide polygons with a total area of 298,557m 2 . Comparison with a manual inventory reveals that the method was able to conservatively delineate landslides, thus making it a practical substitute for manual landslide mapping for preliminary landslide hazard assessment.

The capacity of installed solar photovoltaic (PV) panels has been increased dramatically in recent years due to the strong demand for renewable energy resources. Locations and spatial extent of those facilities are important for grid-managers, environmental monitoring purposes, and disaster mitigation works. Here, a simple machine-learning based solar PV detection in medium-resolution satellite images is proposed. The approach is cost-effective and scalable to a larger area compared to the exiting studies which usually rely on high-resolution but expensive satellite or aerial data. The proposed method is applied on a case-study site in Vietnam, and a good performance of the PV detection is achieved for the study site. The scale of the study area is currently limited to a part of Vietnam due to a lack of computation resources, but it could be possible to process and create a solar PV dataset for the whole country if enough computation power is available.

SAR image simulation is a radar imaging simulation technology developed in recent years, which helps the analysis design, and verification of SAR system, assists SAR image interpretation, SAR image processing algorithm verification, and SAR image geometry correction, and thus has important theoretical significance and practical value. A 3-dimensional canonical target known as SLICY (Sandia Laboratory Implementation of Cylinders) and a collapse model are used for synthetic aperture radar imaging simulations. Reflectivity maps corresponding to signals with different types of radar backscatter are plotted. Combined with the geometrical physical information of the target, the correspondence between the point and line features, different signals of scattering, and object-specific structures in the simulated SAR images are analyzed. The results show that reflection levels of the radar signal are closely related to the object structure details. The analysis of the reflective intensity characteristics of this paper provides a good basis for later inverse estimation of damage to earthquake-damaged buildings.

Electronic health record (EHR) data are increasingly used to develop prediction models to support clinical care, including the care of patients with common chronic conditions. A key challenge for individual healthcare systems in developing such models is that they may not be able to achieve the desired degree of robustness using only their own data. A potential solution-combining data from multiple sources-faces barriers such as the need for data normalization and concerns about sharing patient information across institutions. To address these challenges, we evaluated three alternative approaches to using EHR data from multiple healthcare systems in predicting the outcome of pharmacotherapy for type 2 diabetes mellitus(T2DM). Two of the three approaches, named Selecting Better (SB) and Weighted Average(WA), allowed the data to remain within institutional boundaries by using pre-built prediction models; the third, named Combining Data (CD), aggregated raw patient data into a single dataset. The prediction performance and prediction coverage of the resulting models were compared to single-institution models to help judge the relative value of adding external data and to determine the best method to generate optimal models for clinical decision support. The results showed that models using WA and CD achieved higher prediction performance than single-institution models for common treatment patterns. CD outperformed the other two approaches in prediction coverage, which we defined as the number of treatment patterns predicted with an Area Under Curve of 0.70 or more. We concluded that 1) WA is an effective option for improving prediction performance for common treatment patterns when data cannot be shared across institutional boundaries and 2) CD is the most effective approach when such sharing is possible, especially for increasing the range of treatment patterns that can be predicted to support clinical decision making.

Tropical peatlands have been experienced human-induced disturbances including drainage constructions, oil palm plantations, and wildfires. Since peat soils consist of large organic matter, it is important for the carbon cycle in the tropical area. This study aimed at the delineating of the distribution of drainage canals using microwave images. As the result of validation, overall accuracy was 56.3 % and about 10 % of the peatland areas detected as the drainage canals. For further study, a Keetch Byram Drought Index-based groundwater table will be revised applying the decreasing effect by drainage canals.

In this study, we used the multi-temporal Landsat images of 1985, 2000 and 2017 to evaluate the land use/cover change of Horqin sandy land in the past 33 years by using the classification method of Support Vector Machine (SVM). The study shows that grassland, cropland, woodland, water area and sandy land changed significantly during this period. The analysis of Land cover/use in the Horqin sandy land reveals two separate specific trends between the first 16 years (1985-2000) and the subsequent 17 years (2001- 2017). Generally, grassland and woodland decreased gradually, replaced by a significant increase in cropland area and a gradual expansion of sandy land prior to 2000. This could be related to the household responsibility system. While after 2000 the grassland increased. This could be due to the policy of returning grazing to grassland. From 1985 to 2017, however, the water area has continued to decrease.

A multi-dimensional approach is an upcoming technology in disaster risk to understand a real risk. In the case of flooding, dimensions of time and elevation (height) should be considered to produce rapid and accurate risk information. Dynamic spatio-temporal flood prediction is an urgent issue for the risk management of short- and long-duration flooding. The main objective of this paper is to propose a step-wise process for dynamic flood detection using the synchronized floodwater index (SfWI2) coupled with 5-day lead-time forecasting data and in-situ water-level data with the primary focus on daily change in flood extent. We mainly employed the 2015 time-series Moderate Resolution Imaging Spectrometer (MODIS) data acquired over the study area along the Brahmaputra River where flood events occur annually. Based on scenario-based forecasting water-levels, flood extent maps were created and show the possibility of the proposed multi-dimensional approach as a useful tool for short-duration flood prediction.

Combination of satellite images, Land-Use/Land-Cover (LULC) and census data has potential to identify and characterize urban growth. In this study, we used NTL to compare urban growth pattern in Adelaide, Australia and Tokyo, Japan, and then investigated the relationship among NTL, LULC and population census data for analyzing urban growth. The subsequent spatial correlation analysis show that there was a strong positive correlation between urban/built-up area and population density in both Adelaide (r = 0.90) and Tokyo (r = 0.81). Furthermore, Multiple linear regression model quantified population density in Tokyo using the combination of NTL imagery and urban/built-up data obtaining a correlation value of R2= 0.80. The study demonstrated that combination of NTL and LULC data is able to predict urban population growth.

Peatland is a natural carbon reservoir in terrestrial ecosystem. The ground water table in peatland is a key factor for carbon exchange through soil decomposition or carbon sink. Especially, Indonesia has the biggest peatland area in Asia. This study focused on the carbon dioxide budget between emissions by ecosystem respiration including fire event and the absorption by photosynthesis. As the result, the annual average of net biome ecosystem carbon dioxide exchange during 12 years were reached to 195.03 MtC/yr. Over the three times of CO2 from fire emissions were emitted by ecosystem respiration from whole peatlands in Indonesia.

For high accuracy flood mapping, an algorithm that integrates multiple satellite data sources is essential to maximize the sensor ability and compensate the limitations of optical and SAR data. The main objective of this study is to propose an algorithm of dynamic flood detection using optical and Synthetic Aperture Radar (SAR) images that compares and combines two different statistical thresholding approaches. To improve the flood detection accuracy, image fusion technique was investigated to maximize the utilization of calibrated and optimized flood maps as the integrated flood detection approach. To showcase the advantages of the proposed methodology, we employ MODIS, Landsat-8 and Sentinel-IA images acquired over a challenging area along the Brahmaputra River where flood events often occur.

In cloudy area of the Earth, MODIS limits the sensor's ability to quantify biophysical processes in heterogeneous landscape. A passive microwave sensor AMSR2 is not subject to cloud contamination although its spatial resolution is relatively coarse. In this paper, a new spatial and temporal adaptive data fusion model algorithm is presented and demonstrated to blend MODIS and AMSR2 to predict daily land surface water coverage. The MODIS 8 day composite 1km normalized difference water index (NDWI) and AMSR2 daily 16km normalized difference frequency index (NDFI) are used to map land surface water coverage (LSWC) which is effective to monitor agriculture and flood monitoring issues. It was found that the algorithm accurately predicts daily LSWC of AMSR2 at an effective fractional coverage close to that of MODIS.

In this paper, we built LSWC database in time series from 2002 to 2015 derived from AMSR-E, Windsat and AMSR2 to detect flood events. Then, by conducting spatio-temporal analysis we made clear the flooding pattern and dug up some regularity, hidden information of flooding in different land use land cover situation based on viewpoint of retrieval of historical global record. Last, PALSAR data was used to conduct more precise calibration of AMSR-E by establishing a calibration function of index NDFI, NDPI with PALSAR LSWC. It shows a good relationship between these indexes. Thus, the availability and importance of LSWC database for flooding detection on global scale was indicated.

Methane is one of the main souse of global warming and its emission from rice paddy field is a big issue. To fully investigate the special and temporal distribution of methane emission from rice paddy, we utilize the rice crop calendar mapping and ground water coverage data from satellite observation for IPCC 2006 guidelines. Then, global methane emission inventory from rice paddy is estimated. Comparison between atmospheric methane concentration from GOSAT and estimated methane emission inventory in regional scale is done and positive relationship between them is found.

Carbon dioxide (CO 2 ) budget from tropical peat lands in Indonesia was estimated by using satellite data. Indonesia has many CO 2 sources that related with peat lands. Our previous results have been continued to estimate CO 2 budget from ground water decreasing. The new motivation of this study is to estimate human impact on the peat lands hydrologic environment. The drainage canal is supposed to be artificial in this region. ALOS PALSAR mosaic data was used for detection of drainage canal in Indonesia. Canny edge detection method was used. As the result, daily soil respirations of Jambi and Palangkaraya became 12.63% and 16.84% larger than before consideration of human effect.

The peatlands are known as the carbon sink in natural. However some disturbances such as fire and drainage are occurred in Indonesian peatlands. The declining of ground water table is the most influential reason of carbon emission from peatlands. For detecting CO2 emission from peatlands in Indonesia, ground water table was estimated by satellite based precipitation and land surface temperature. The CO2 emission is represented by NEE(Net Ecosystem CO2 Exchange) which can be calculated with this equation; NEE = ER - GPP; where ER is ecosystem respiration, and GPP is gross primary productivity. As the result, the ecosystem respiration was larger than GPP in this peat forest. The annual average of ER was about twice of GPP in this region. The GPP of MOD17A2 is underestimated from in-situ observed GPP with 37.2%. The CO2 emission through fire and respiration is increased when GWT was declined.

Nowadays estimation methods by using remote sensing to approaching actual reality are come out one after the other even it is not easy to consider each of impact factors. In this paper using glob-cover land-cover map of Eastern Europe for the essential data, masking out the wetland distribution according classification. Then, in line with create 73 points interested in, through calculating snow coverage and LSWC (Land Surface Water Coverage) offset and onset timing and extension duration from AMSR-E database to find out the regular pattern of climate seasonal changes phenomenon. Thirdly, during comparing batch processing within pixel value in time series from 2003 to 2011, to find out the changes of snow and water coverage. Finally, combine with NDVI, precipitation and LST (land surface temperature) data to estimate the CH4 flux.

As a cause of global warming, CO2 is most effective green house gas and many countries are trying to reduce emission of that. However, the report of quantitative description is few still because the amount of CO2 emission is difficult to estimate for their complex process. This research is to devise a method to estimate the carbon dioxide (CO2) emissions from biomass burning such as forest fires and field burning in croplands in Global scale. The methodology mainly composed of two parts including emissions from wild fires of above ground biomass (AGB) and soil organic matter (SOM). The estimated CO2 emissions from 2002 to 2012 were demonstrated and compared with Global Fire Emission Database version 3.1 (GFED 3.1) which is widely used as a global database. A comparison of biomass burning carbon emissions derived from this study and GFED3.1 database in global scale shows that overall the map derived from those two models have very similar spatial patterns. High carbon emission is found at forested area in South America, Central Africa, Far East Russia, North America and Southeast Asia as well as croplands in Central Eurasia and Southeast Asia. A comparison of biomass burning carbon emissions derived from this study and GFED3.1 database in global scale shows that our study estimated biomass burning carbon emissions as 2.62-0.22 PgC/yr whereas GFED is 2.29-0.26 PgC/yr.

The MODerate resolution Imaging Spectrometer (MODIS) aboard Terra and Aqua platforms is performing well overall, except for Aqua MODIS band 6. Fifteen of the 20 detectors in Aqua MODIS band 6 are nonfunctional or noisy. Without correction, it will cause problems in the higher MODIS products. This paper develops a restoration algorithm to restore the missing data of Aqua MODIS band 6 by combining a histogram-matching algorithm with local least squares fitting. Histogram matching corrects detector-to-detector striping of the functional detectors. Local least squares fitting restores the missing data of the nonfunctional detector based on a cubic polynomial derived from the relationship between Aqua MODIS bands 6 and 7. The Aqua MODIS image data used in this research are in digital number format and are not georectified. The proposed restoring algorithm can be used on both 1000- and 500-m pixel resolutions. The algorithm was tested on both Terra and Aqua MODIS images. For Terra MODIS images, results of restoring the synthetic nonfunctional detectors of band 6 demonstrate that local least squares fitting can fill in the missing data with little distortion. For Aqua MODIS images, the results of the restoring algorithm with and without applying histogram matching were compared to evaluate the capability in removing detector-to-detector stripe noise. To evaluate the performance of the proposed method, quantitative and qualitative analyses were carried out by visual inspection and quality index. For all the scenes used in this research, the correlation coefficients were near 0.99 and root mean square error between the original Terra band 6 and its simulated one was 2times10-5. The proposed algorithm can thus be used satisfactorily for restoring Aqua MODIS band 6.

This study proposes an extended subspace method (ESM) in feature extraction and dimension-reduction problems for land cover classification of hyperspectral and multi-spectral remote sensing images. The main idea of our method is to use a multiple similarity method (MSM) onto an averaged learning subspace method (ALSM) and makes use of fidelity value criteria in the selection of the optimal subspace dimensions. This method is compared with the support vector machine (SVM) method using Compact Airborne Spectrographic Imager-2 (CASI-2) hyperspectral remote sensing data. Experimental results show that ESM is a valid and effective alternative to other pattern recognition approaches for the classification of remote sensing data.

The Moderate Resolution Imaging Spectrometer (MODIS) aboard Terra and Aqua platforms are contaminated by stripe noises. There are three types of stripe noises in MODIS data: detector-to-detector stripes, mirror side stripes, and noisy stripes. Without correction, stripe noises will cause processing errors to the other MODIS products. In this paper, a noise-reduction algorithm is developed to reduce the stripe noise effects in both Terra MODIS and Aqua MODIS data by combining a histogram-matching algorithm with an iterated weighted least-squares (WLS) facet filter. Histogram matching corrects for detector-to-detector stripes and mirror side stripes. The iterated WLS facet filter corrects for noisy stripes. The method was tested on heavily striped Terra MODIS and Aqua MODIS images. Results of Terra MODIS and Aqua MODIS data show that the proposed algorithm reduced stripes noises without degrading image quality. To evaluate performance of the proposed method, quantitative and qualitative analyses were carried out by visual inspection and quality indexes of destriped images

In Asian countries, paddy field is indispensable for our lives not only as a source of food but also ecosystem, hydrology, landscape and culture. It is presumed to be one of the most likely source of atmospheric methane and affects atmospheric phenomena in the local scale. In this sense it is necessary to get the detailed spatial distribution of paddy field in Asian region. The improved understanding of paddy field distribution at large spatial scales has increased the interest in deriving crop yield and methane emission estimations. Nevertheless, the collection of such data through field surveys is time-consuming and expensive in South-East Asia regions. Remotely sensing data from satellite images provide an alternative means of obtaining paddy field distribution. In this study, land cover characterization technique for paddy fields is investigated using time series of NOAA/AVHRR data. The method proposed in this study includes six processing steps: 1) precise geometric and radiometric correction, 2) automatic cloud screening, 3) atmospheric correction by 6S code, 4) BRDF correction by kernel model, 5) automatic endmember selection for spectral mixture analysis, 6) generation of vegetation, soil and water map and their time series analysis.

The West Siberian Lowland (WSL) is the world's largest high-latitude wetland covering nearly 2/3 of western Siberia. At least half of this area consists of peatlands, which sequester atmospheric carbon in the form of undecomposed plant matter and it is presumed to be a source of methane gas. In this paper, firstly, an ASTER image near Noyabrsk mire was used to map six wetland ecosystems (red pine, white birch, bog, palsa, open water and bare soil) supplemented by field observation. Then spectral linear mixture analysis was performed between MODIS and ASTER data acquired on the same day. Secondly, field observations were scaled up with these different spatial resolution satellite data. Each of the wetland ecosystem coverage ratio in sub-pixel level was provided by the spectral linear mixture analysis. Field observation shows that the mean rate of CH/sub 4/ emission from bog and open water averaged 5.246 and 1.081 (mg CH/sub 4/ m/sup -2/ h/sup -1/) respectively. The methane emission from the area was estimated by multiplying these average methane emission rate and the area percentage of bog and open water in each pixel. Finally, the mean methane emission over MODIS coverage was estimated to be 1.864/spl times/10/sup 9/ g CH/sub 4/ day/sup -1/.