Remote Sensing for Agriculture, Ecosystems, and Hydrology XVII, 2015
Saltmarsh is one of the important communities of wetlands, however, due to a range of pressures, ... more Saltmarsh is one of the important communities of wetlands, however, due to a range of pressures, it has been declared as an EEC (Ecological Endangered Community) in Australia. In order to correctly identify different saltmarsh species, development of spectral libraries of saltmarsh species is essential to monitor this EEC. Hyperspectral remote sensing, can explore the area of wetland monitoring and mapping. The benefits of Hyperion data to wetland monitoring have been studied at Hunter Wetland Park, NSW, Australia. After exclusion of bad bands from the original data, an atmospheric correction model was applied to minimize atmospheric effect and to retrieve apparent surface reflectance for different land cover. Large data dimensionality was reduced by Forward Minimum Noise Fraction (MNF) algorithm. It was found that first 32 MNF band contains more than 80% information of the image. Pixel Purity Index (PPI) algorithm worked properly to extract pure pixel for water, builtup area and three vegetation Casuarina sp., Phragmitis sp. and green grass. The result showed it was challenging to extract extreme pure pixel for Sporobolus and Sarcocornia from the data due to coarse resolution (30 m) and small patch size (<3 m) of those vegetation on the ground. Spectral Angle Mapper, classified the image into five classes: Casuarina, Saltmarsh (Phragmitis), Green grass, Water and Builtup area with 43.55 % accuracy. This classification also failed to classify Sporobolus as a distinct group due to the same reason. A high spatial resolution airborne hyperspectral data and a new study site with a bigger patch of Sporobolus and Sarcocornia is proposed to overcome the issue.
Saltmarsh is one of the important communities of wetlands, however, due to a range of pressures, ... more Saltmarsh is one of the important communities of wetlands, however, due to a range of pressures, it has been declared as an EEC (Ecological Endangered Community) in Australia. In order to correctly identify different saltmarsh species, development of spectral libraries of saltmarsh species is essential to monitor this EEC. Hyperspectral remote sensing, can explore the area of wetland monitoring and mapping. The benefits of Hyperion data to wetland monitoring have been studied at Hunter Wetland Park, NSW, Australia. After exclusion of bad bands from the original data, an atmospheric correction model was applied to minimize atmospheric effect and to retrieve apparent surface reflectance for different land cover. Large data dimensionality was reduced by Forward Minimum Noise Fraction (MNF) algorithm. It was found that first 32 MNF band contains more than 80% information of the image. Pixel Purity Index (PPI) algorithm worked properly to extract pure pixel for water, builtup area and three vegetation Casuarina sp., Phragmitis sp. and green grass. The result showed it was challenging to extract extreme pure pixel for Sporobolus and Sarcocornia from the data due to coarse resolution (30 m) and small patch size (<3 m) of those vegetation on the ground. Spectral Angle Mapper, classified the image into five classes: Casuarina, Saltmarsh (Phragmitis), Green grass, Water and Builtup area with 43.55 % accuracy. This classification also failed to classify Sporobolus as a distinct group due to the same reason. A high spatial resolution airborne hyperspectral data and a new study site with a bigger patch of Sporobolus and Sarcocornia is proposed to overcome the issue.
Coastal saltmarsh and their constituent components and processes are of an interest scientificall... more Coastal saltmarsh and their constituent components and processes are of an interest scientifically due to their ecological function and services. However, heterogeneity and seasonal dynamic of the coastal wetland system makes it challenging to map saltmarshes with remotely sensed data. This study selected four important saltmarsh species Pragmitis australis, Sporobolus virginicus, Ficiona nodosa and Schoeloplectus sp. as well as a Mangrove and Pine tree species, Avecinia and Casuarina sp respectively. High Spatial Resolution Worldview-2 data and Coarse Spatial resolution Landsat 8 imagery were selected in this study. Among the selected vegetation types some patches ware fragmented and close to the spatial resolution of Worldview-2 data while and some patch were larger than the 30 meter resolution of Landsat 8 data. This study aims to test the effectiveness of different classifier for the imagery with various spatial and spectral resolutions. Three different classification algorithm, Maximum Likelihood Classifier (MLC), Support Vector Machine (SVM) and Artificial Neural Network (ANN) were tested and compared with their mapping accuracy of the results derived from both satellite imagery. For Worldview-2 data SVM was giving the higher overall accuracy (92.12%, kappa =0.90) followed by ANN (90.82%, Kappa 0.89) and MLC (90.55%, kappa = 0.88). For Landsat 8 data, MLC (82.04%) showed the highest classification accuracy comparing to SVM (77.31%) and ANN (75.23%). The producer accuracy of the classification results were also presented in the paper. .
D ifferent methods are used to measure above ground biomass (AGB) and thus the carbon stock of fo... more D ifferent methods are used to measure above ground biomass (AGB) and thus the carbon stock of forests. Combining very high resolution (VHR) optical satellite data with airborne Lidar data has provided new opportunities to assess and map the carbon stock of forests accurately. The present study was conducted in a subtropical forest in Kayer Khola watershed, Chitwan District, Nepal. The retrieval of canopy height, crown projected area (CPA), and tree species classification was assessed, and their application in carbon stock estimation evaluated. WorldView-2 and Geo-Eye satellite data were co-registered and combined with airborne Lidar data to obtain a canopy height model. The main objective of the study was to map biomass and carbon stock in three community forests in Chitwan District, Nepal. Other specific objectives were to compare community and government forest management, assess the community forest certification process for sustainable forest management, assess forest trees species diversity, and develop a model base to estimate soil organic carbon.
— Now a days the possibility of enhanced carbon storage in soils is of more interest compared to ... more — Now a days the possibility of enhanced carbon storage in soils is of more interest compared to vegetation as it contains more carbon. For this reason, the revised Kyoto protocol includes two new clauses relevant to soil organic carbon sequestration. So, for the countries that have signed the Kyoto protocol, estimation of SOC sequestration is a required strategy. Reliable quantification of carbon held in soil is essential to formulate any kinds of monitoring program. This SOC is dominated by a lot of variables like environmental and soil internal factors as well. This study aims therefore to study the effect of two remotely sensed measured variables on SOC in the subtropical forest of Chitwan, Nepal. Two variables, above ground biomass (AGB) and elevation and other two soil parameters bulk density and soil pH were analysed in context of soil organic carbon. Although soil bulk density and pH cannot be measured through remote sensing technology, they were used to test the relationship. Soil organic carbon was analysed through Walkley-Black and Loss on Ignition (LOI) methods. Canopy Height Model (CHM) was developed from LiDAR data by subtracting the Digital Terrain Model (DTM) from the Digital Surface Model (DSM) to estimate the height of the trees. This CHM image was segmented based on an Object Based Image Analysis (OBIA) technique using e Cognition software. Segmented CPA further analysed to develop a model for DBH prediction. With the information of DBH, tree height and wood specific gravity, AGB was calculated. Elevation height was extracted from LiDAR derived DEM. Results show that there is a positive relationship (r =0.79) between soil organic carbon and above ground biomass (p<0.001). Elevation and soil organic carbon is also positively correlated (r=0.74).
The possibility of carbon storage in soils is of interest because compared to vegetation it conta... more The possibility of carbon storage in soils is of interest because compared to vegetation it contains more carbon. Estimation of soil carbon through remote sensing based techniques can be a cost effective approach, but is limited by available methods. This study aims to develop a model based on remotely sensed variables (elevation, forest type and above ground biomass) to estimate soil carbon stocks. Field observations on soil organic carbon, species composition, and above ground biomass were recorded in the subtropical forest of Chitwan, Nepal. These variables were also estimated using LiDAR data and a WorldView 2 image. Above ground biomass was estimated from the LiDAR image using a novel approach where the image was segmented to identify individual trees, and for these trees estimates of DBH and Height were made. Based on AIC (Akaike Information Criterion) a regression model with above ground biomass derived from LiDAR data, and forest type derived from WorldView 2 imagery was selected to estimate soil organic carbon (SOC) stocks. The selected model had a coefficient of determination (R 2) of 0.69. This shows the scope of estimating SOC with remote sensing derived variables in subtropical forests.
Saltmarsh is one of the important communities of wetlands. Due to a range of pressures, it has be... more Saltmarsh is one of the important communities of wetlands. Due to a range of pressures, it has been declared as an EEC (Ecological Endangered Community) in Australia. In order to correctly identify different saltmarsh species, development of distinct spectral characteristics is essential to monitor this EEC. This research was conducted to classify saltmarsh species based on spectral characteristics in the VNIR wavelength of Hyperion Hyperspectral and Worldview 2 multispectral remote sensing data. Signal Noise Ratio (SNR) and Principal Component Analysis (PCA) were applied in Hyperion data to test data quality and to reduce data dimensionality respectively. FLAASH atmospheric correction was done to get surface reflectance data. Based on spectral and spatial information a supervised classification followed by Mapping Accuracy (%) was used to assess the classification result. SNR of Hyperion data was varied according to season and wavelength and it was higher for all land cover in VNIR wavelength. There was a significant difference between radiance and reflectance spectra. It was found that atmospheric correction improves the spectral information. Based on the PCA of 56 VNIR band of Hyperion, it was possible to segregate 16 bands that contain 99.83 % variability. Based on reference 16 bands were compared with 8 bands of Worldview 2 for classification accuracy. Overall Accuracy (OA) % for Worldview 2 was increased from 72 to 79 while for Hyperion, it was increased from 70.47 to 71.66 when bands were added orderly. Considering the significance test with z values and kappa statistics at 95% confidence level, Worldview 2 classification accuracy was higher than Hyperion data.
Remote Sensing for Agriculture, Ecosystems, and Hydrology XVII, 2015
Saltmarsh is one of the important communities of wetlands, however, due to a range of pressures, ... more Saltmarsh is one of the important communities of wetlands, however, due to a range of pressures, it has been declared as an EEC (Ecological Endangered Community) in Australia. In order to correctly identify different saltmarsh species, development of spectral libraries of saltmarsh species is essential to monitor this EEC. Hyperspectral remote sensing, can explore the area of wetland monitoring and mapping. The benefits of Hyperion data to wetland monitoring have been studied at Hunter Wetland Park, NSW, Australia. After exclusion of bad bands from the original data, an atmospheric correction model was applied to minimize atmospheric effect and to retrieve apparent surface reflectance for different land cover. Large data dimensionality was reduced by Forward Minimum Noise Fraction (MNF) algorithm. It was found that first 32 MNF band contains more than 80% information of the image. Pixel Purity Index (PPI) algorithm worked properly to extract pure pixel for water, builtup area and three vegetation Casuarina sp., Phragmitis sp. and green grass. The result showed it was challenging to extract extreme pure pixel for Sporobolus and Sarcocornia from the data due to coarse resolution (30 m) and small patch size (&lt;3 m) of those vegetation on the ground. Spectral Angle Mapper, classified the image into five classes: Casuarina, Saltmarsh (Phragmitis), Green grass, Water and Builtup area with 43.55 % accuracy. This classification also failed to classify Sporobolus as a distinct group due to the same reason. A high spatial resolution airborne hyperspectral data and a new study site with a bigger patch of Sporobolus and Sarcocornia is proposed to overcome the issue.
Saltmarsh is one of the important communities of wetlands, however, due to a range of pressures, ... more Saltmarsh is one of the important communities of wetlands, however, due to a range of pressures, it has been declared as an EEC (Ecological Endangered Community) in Australia. In order to correctly identify different saltmarsh species, development of spectral libraries of saltmarsh species is essential to monitor this EEC. Hyperspectral remote sensing, can explore the area of wetland monitoring and mapping. The benefits of Hyperion data to wetland monitoring have been studied at Hunter Wetland Park, NSW, Australia. After exclusion of bad bands from the original data, an atmospheric correction model was applied to minimize atmospheric effect and to retrieve apparent surface reflectance for different land cover. Large data dimensionality was reduced by Forward Minimum Noise Fraction (MNF) algorithm. It was found that first 32 MNF band contains more than 80% information of the image. Pixel Purity Index (PPI) algorithm worked properly to extract pure pixel for water, builtup area and three vegetation Casuarina sp., Phragmitis sp. and green grass. The result showed it was challenging to extract extreme pure pixel for Sporobolus and Sarcocornia from the data due to coarse resolution (30 m) and small patch size (<3 m) of those vegetation on the ground. Spectral Angle Mapper, classified the image into five classes: Casuarina, Saltmarsh (Phragmitis), Green grass, Water and Builtup area with 43.55 % accuracy. This classification also failed to classify Sporobolus as a distinct group due to the same reason. A high spatial resolution airborne hyperspectral data and a new study site with a bigger patch of Sporobolus and Sarcocornia is proposed to overcome the issue.
Coastal saltmarsh and their constituent components and processes are of an interest scientificall... more Coastal saltmarsh and their constituent components and processes are of an interest scientifically due to their ecological function and services. However, heterogeneity and seasonal dynamic of the coastal wetland system makes it challenging to map saltmarshes with remotely sensed data. This study selected four important saltmarsh species Pragmitis australis, Sporobolus virginicus, Ficiona nodosa and Schoeloplectus sp. as well as a Mangrove and Pine tree species, Avecinia and Casuarina sp respectively. High Spatial Resolution Worldview-2 data and Coarse Spatial resolution Landsat 8 imagery were selected in this study. Among the selected vegetation types some patches ware fragmented and close to the spatial resolution of Worldview-2 data while and some patch were larger than the 30 meter resolution of Landsat 8 data. This study aims to test the effectiveness of different classifier for the imagery with various spatial and spectral resolutions. Three different classification algorithm, Maximum Likelihood Classifier (MLC), Support Vector Machine (SVM) and Artificial Neural Network (ANN) were tested and compared with their mapping accuracy of the results derived from both satellite imagery. For Worldview-2 data SVM was giving the higher overall accuracy (92.12%, kappa =0.90) followed by ANN (90.82%, Kappa 0.89) and MLC (90.55%, kappa = 0.88). For Landsat 8 data, MLC (82.04%) showed the highest classification accuracy comparing to SVM (77.31%) and ANN (75.23%). The producer accuracy of the classification results were also presented in the paper. .
D ifferent methods are used to measure above ground biomass (AGB) and thus the carbon stock of fo... more D ifferent methods are used to measure above ground biomass (AGB) and thus the carbon stock of forests. Combining very high resolution (VHR) optical satellite data with airborne Lidar data has provided new opportunities to assess and map the carbon stock of forests accurately. The present study was conducted in a subtropical forest in Kayer Khola watershed, Chitwan District, Nepal. The retrieval of canopy height, crown projected area (CPA), and tree species classification was assessed, and their application in carbon stock estimation evaluated. WorldView-2 and Geo-Eye satellite data were co-registered and combined with airborne Lidar data to obtain a canopy height model. The main objective of the study was to map biomass and carbon stock in three community forests in Chitwan District, Nepal. Other specific objectives were to compare community and government forest management, assess the community forest certification process for sustainable forest management, assess forest trees species diversity, and develop a model base to estimate soil organic carbon.
— Now a days the possibility of enhanced carbon storage in soils is of more interest compared to ... more — Now a days the possibility of enhanced carbon storage in soils is of more interest compared to vegetation as it contains more carbon. For this reason, the revised Kyoto protocol includes two new clauses relevant to soil organic carbon sequestration. So, for the countries that have signed the Kyoto protocol, estimation of SOC sequestration is a required strategy. Reliable quantification of carbon held in soil is essential to formulate any kinds of monitoring program. This SOC is dominated by a lot of variables like environmental and soil internal factors as well. This study aims therefore to study the effect of two remotely sensed measured variables on SOC in the subtropical forest of Chitwan, Nepal. Two variables, above ground biomass (AGB) and elevation and other two soil parameters bulk density and soil pH were analysed in context of soil organic carbon. Although soil bulk density and pH cannot be measured through remote sensing technology, they were used to test the relationship. Soil organic carbon was analysed through Walkley-Black and Loss on Ignition (LOI) methods. Canopy Height Model (CHM) was developed from LiDAR data by subtracting the Digital Terrain Model (DTM) from the Digital Surface Model (DSM) to estimate the height of the trees. This CHM image was segmented based on an Object Based Image Analysis (OBIA) technique using e Cognition software. Segmented CPA further analysed to develop a model for DBH prediction. With the information of DBH, tree height and wood specific gravity, AGB was calculated. Elevation height was extracted from LiDAR derived DEM. Results show that there is a positive relationship (r =0.79) between soil organic carbon and above ground biomass (p<0.001). Elevation and soil organic carbon is also positively correlated (r=0.74).
The possibility of carbon storage in soils is of interest because compared to vegetation it conta... more The possibility of carbon storage in soils is of interest because compared to vegetation it contains more carbon. Estimation of soil carbon through remote sensing based techniques can be a cost effective approach, but is limited by available methods. This study aims to develop a model based on remotely sensed variables (elevation, forest type and above ground biomass) to estimate soil carbon stocks. Field observations on soil organic carbon, species composition, and above ground biomass were recorded in the subtropical forest of Chitwan, Nepal. These variables were also estimated using LiDAR data and a WorldView 2 image. Above ground biomass was estimated from the LiDAR image using a novel approach where the image was segmented to identify individual trees, and for these trees estimates of DBH and Height were made. Based on AIC (Akaike Information Criterion) a regression model with above ground biomass derived from LiDAR data, and forest type derived from WorldView 2 imagery was selected to estimate soil organic carbon (SOC) stocks. The selected model had a coefficient of determination (R 2) of 0.69. This shows the scope of estimating SOC with remote sensing derived variables in subtropical forests.
Saltmarsh is one of the important communities of wetlands. Due to a range of pressures, it has be... more Saltmarsh is one of the important communities of wetlands. Due to a range of pressures, it has been declared as an EEC (Ecological Endangered Community) in Australia. In order to correctly identify different saltmarsh species, development of distinct spectral characteristics is essential to monitor this EEC. This research was conducted to classify saltmarsh species based on spectral characteristics in the VNIR wavelength of Hyperion Hyperspectral and Worldview 2 multispectral remote sensing data. Signal Noise Ratio (SNR) and Principal Component Analysis (PCA) were applied in Hyperion data to test data quality and to reduce data dimensionality respectively. FLAASH atmospheric correction was done to get surface reflectance data. Based on spectral and spatial information a supervised classification followed by Mapping Accuracy (%) was used to assess the classification result. SNR of Hyperion data was varied according to season and wavelength and it was higher for all land cover in VNIR wavelength. There was a significant difference between radiance and reflectance spectra. It was found that atmospheric correction improves the spectral information. Based on the PCA of 56 VNIR band of Hyperion, it was possible to segregate 16 bands that contain 99.83 % variability. Based on reference 16 bands were compared with 8 bands of Worldview 2 for classification accuracy. Overall Accuracy (OA) % for Worldview 2 was increased from 72 to 79 while for Hyperion, it was increased from 70.47 to 71.66 when bands were added orderly. Considering the significance test with z values and kappa statistics at 95% confidence level, Worldview 2 classification accuracy was higher than Hyperion data.
Uploads
Papers by Sikdar Rasel