Tomohiro Shiraishi

Comparative Assessment of Supervised Classifiers for Land Use–Land Cover Classification in a Tropical Region Using Time-Series PALSAR Mosaic Data

Numerous classification algorithms have been proposed to create accurate classification maps using optical remote sensing data. However, few comparative studies evaluate the performance of classification algorithms with focus on tropical forests due to cloud effects. Advances in synthetic aperture radar (SAR) techniques and spatial resolution, mapping, and comparison of classification algorithms are possible. This research investigated the accuracy and processing speeds of five supervised classifiers, including Naïve Bayes, AdaBoost, multi-layer perceptron, random forest (RF), and support vector machine, for land use-land cover (LULC) classification in a tropical region using time-series Advanced Land Observing Satellite-phased array type L-band SAR (ALOS-PALSAR) 25-m mosaic data. The study area is located in central Sumatra, Indonesia, where abundant forest-related carbon stocks exist. This investigation was intended to aid the implementation of a classification algorithm for the automatic creation of LULC classification maps. We perform object-based and pixel-based analyses to investigate the ability of the classifiers and their accuracies, respectively. RF had the best classification accuracy and processing speed in which the accuracies for 10 classes and 2 classes were 64.07% and 90.22% for pixel-based and 82.94% and 86.23% for object-based evaluations, respectively. These results indicate that RF is a useful classifier for the analysis of PALSAR mosaic data and that the automatic creation of highly accurate classification maps is possible by using time-series data. The outcome of this research will be valuable resources for biodiversity and global-warming mitigation efforts in the region.

Calibration of Aboveground Forest Carbon Stock Models for Major Tropical Forests in Central Sumatra Using Airborne LiDAR and Field Measurement Data

Despite substantial policy attention, tropical forests in Southeast Asian region are releasing large amount of carbon to the atmosphere due to accelerating deforestation. Accurately determining forest statistics and characterizing aboveground forest carbon stocks (AFCSs) are always challenging in the region. In order to develop more accurate estimates of AFCS, the present study collected airborne LiDAR and field measurements data and calibrated AFCS models to estimate carbon stock in the tropical forests in central Sumatra. The study region consists of natural forests, including peat swamp, dry moist, regrowth, and mangrove, and plantation forests, including rubber, acacia, oil palm, and coconut. To cover the different forest types, 60 field plots of 1 ha in size were inventoried. Eight transects crossing these field plots were acquired to calibrate the LiDAR to AFCS models. The AFCS values for the field plots ranged from 4 to 161 Mg ha-1. General models were fitted without considering forest types, whereas a specific model was fitted for each specific forest type. Five alternative general models with different LiDAR metrics were calibrated with model performance expressed as R2 ranging from 0.73 to 0.87 and root-meansquare error (RMSE) values ranging from 17.4 to 25.0 Mg ha-1 . Seven forest-specific AFCS models were calibrated for different forest types, with R2 values ranging from 0.72 to 0.97 and RMSE values ranging from 1.4 to 10.7 Mg ha-1. The performance of each model was cross-validated by iteratively removing one data point. While forest-specific models provide better AFCS estimates, the general models are still useful when forest types are ambiguous.

Generation of 10m resolution PALSAR and JERS-SAR mosaic and forest/non-forest maps for forest carbon tracking

The Japan Aerospace Exploration Agency (JAXA) has produced the worlds first 10m resolution L-band SAR global mosaic datasets. These data sets were generated to monitor forest changes from the 1990s to present. SRTM-3 (90m resolution) DEM was used to correct the terrain-induced SAR intensity variations and the ortho-rectification. Both corrections were applied for geometric and radiometric calibration purposes. The data sets are useful to monitor the temporal forest cover and forest change, and were used to derive forest/non-forest information.

Multitemporal Fluctuations in L-Band Backscatter From a Japanese Forest

The temporal variations (diurnal and annual) in arboreal (ε_Tree) and bare soil (ε_Soil) dielectric constants and their correlation with precipitation were examined for several trees in Japan. A significant (1 σ (standard deviation) and 2 σ) ε_Tree increase is observed after rainfall at 89.8% and 90.5% probability. However, rainfall does not always induce significant ε_Tree increases. Rainfall of more than 5 mm/day can induce 1 σ ε_Tree Tree increase at a 59.6% probability. In order to examine whether the increase in εTree affects the L-band σ⁰ variation in a forest, the four-year temporal variation of the L-band backscattering coefficient (σ⁰) was estimated from observations by the Advanced Land Observing Satellite Phased Array type L-band Synthetic Aperture Radar. Observed maximum absolute deviations from the mean over the forest area were 1.0 and 1.2 dB for σ_HH⁰ and σ_HV⁰, respectively, and 4.0 and 3.0 dB over open land. σ⁰ and rainfall correlations show that ε_Tree and σ_Forest⁰ are proportional to precipitation integrated over seven or eight days; ε_Soil and σ_{Open land}⁰ are proportional to precipitation integrated over three days. This finding indicates that ε_Tree variations influence σ_{Forest areas}⁰. A stronger correlation between σ_HV⁰ and precipitation is observed in several sites with low σ_HV⁰, where less biomass is expected, and several sites with high σ_HV⁰, where more biomass is expected. A weaker correlation between σ_HV⁰ and precipitation is observed for several sites with high σ_HV⁰. These differences may be explained by the different contributions of double bounce scattering and potential transpiration, which is a measure of the ability of the atmosphere to remove water from the surface through the processes of transpiration. The two other results were as follows: 1) The functional relation between aboveground biomass and σ⁰ showed dependence on precipitation data, this being an effect connected with seasonal changes of the ε_Tree. This experiment reinforces the fact that the dry season is preferable for retrieval of woody biomass from inversion of the functional dependence of SAR backscatter and for avoiding the influence of rainfall. 2) The complex dielectric constant for a tree trunk, which is measured between 0.2 and 6 GHz, indicates that free water is dominant in the measured tree.

Dependency of forest biomass on full Polarimetric parameters obtained from L-band SAR data for a natural forest in Indonesia

Aboveground (AG)-biomass was estimated from a field biomass collection and LiDAR observations for a natural forest in Indonesia. The derived AG-biomass data were plotted against full polarimetric parameters calculated from Polarimetric and Interferometric Airborne Synthetic Aperture Radar L2 (Pi-SAR-L2) and PALSAR data. The α°-AG-biomass curve shows saturation by around 100 tons/ha, while for entropy, correlation is observed up to 200 tons/ha. The largest coefficient of determination (R2 = 0.2348) was observed for the range with AG-biomass of more than 100 tons/ha for the relation between AG-biomass and entropy. The α°HV-biomass plot derived from PALSAR data shows smaller variance in the dry season than in other seasons, indicating that dry season data is preferable for a more accurate estimate of AG-biomass.