In this study, we have performed an analysis between the L-band backscattering intensity derived from the slope corrected ALOS PALSAR remote sensing data and the in-situ stand biophysical parameter of Sugi ( Cryptomeria japonica ) and Hinoki ( Chamaecyparis obtusa ) trees at the forests of Chiba Prefecture, Japan. Diameter at breast height (DBH), tree height, and stem volume were statistically compared with the slope corrected sigma naught backscattering in an empirical approach. It was found that the relationship between the backscattering and the stand characteristics was strongly dependent on species showing different trends between the Sugi and Hinoki trees. The Hinoki trees showed an increasing backscattering with increasing parameters (higher DBH, higher Tree height and higher stem volume), as it was mentioned on various researches, while the Sugi tree showed and decreasing backscattering with increasing parameters. We have also found for the Sugi trees that the backscattering is affected strongly by the number of stems. We have assumed that this is because of the characteristics of the Sugi trees which have high moisture content in the heartwood of the stem, compared with other tree species in Japan. The results pave the way to the possibility for estimating biophysical parameters within the forests of Japan by considering such trends and at highly rugged areas by using slope corrected imagery of the SAR data.
Synthetic Aperture Radar (SAR) application in land remote sensing is becoming one of the top methods chosen among the researchers for solving issues that were facing difficulties when only optical data were selected. The field varies from land cover change, detection analysis, to disasters such as land degradation, earthquakes, and to the application in forest management. Developing algorithms for extracting information’s from the Earth’s surface (e.g. biophysical parameters) is especially of high interests because of the potential of the remote sensing technique where we can interpret areas which are remote and difficult to access, and furthermore, for the use as performing continuous monitoring of our forests resources.
The use of SAR for a biomass/biophysical parameter application is likely to be processed with the relationship between the radar backscattering and the parameters of the forests. This works out because of the characteristics of the radar information; radar waves are longer than the optical waves. As a result, the scattering of the radar contributes from not only the surface of the top layer (e.g. canopy) but also from the medium of the objects (e.g. branches, stems) for the L-band radar [
Studies applied in the forests of Japan are rare, Motohka et al. [
Leblon et al. [
A number of studies have been implemented throughout various regions in the world, but not many studies have been applied at the forests of Japan for understanding the trends between the structure of the Japanese forests and SAR information. Even the work that has been implemented by Motohka et al. [
Therefore, our objective is to analyze the relationship between backscattering information with the biophysical parameters (diameter at breast height (DBH), tree height, stem volume) from different tree types (Sugi (Japanese cedar: Cryptomeria japonica) and Hinoki (Japanese cypress: Chamaecyparis obtusa) at different polarizations in the forests of Japan.
Our study area is focused on the Prefectural owned forests of Chiba Prefecture, Japan. Chiba Prefecture is located on the east coast of Japan along the Pacific Ocean, just east of Tokyo metropolitan area where the peninsula sticks out, approximately between 139.75˚E and 140.88˚E, 34.89˚N and 36.10˚N, where a total land area of 5156 km2 (
Climatic condition shows a warm oceanic climate, which is a condition that is high humidity, high precipitation on summer, and low humidity, low precipitation on the winter. However, compared to the southern and north eastern region of Chiba where the climate is warm throughout the year, the inland area shows more diversity such as a higher temperature drop in the winter. Annual precipitation shows highest at in the southern area with more than 2000 mm and next the northern area which is has about 1400 mm to 1600 mm. The difference in precipitation clearly shows the distribution of the forests being more dense at the south and sparse on the north.
For our analytical purpose, a microwave satellite data was considered in use. We have chosen to apply the Advanced Land Observing Satellite (ALOS) Phased Array L-band Synthetic Aperture Radar (PALSAR) provided by Japan Aerospace Exploration Agency (JAXA). Remote Sensing Technology Center of Japan (RESTEC) has started to provide the PALSAR Global Mosaic (PGM) product which covers global range in the ground range pixel spacing of 10 m or 25 m, along with the process of ortho-rectification and slope correction (selective) from the beginning of 2013. Since the topographic effects that causes distortions to the observed SAR data is critical, we have selected both 10 and 25 m pixel spacing data of the study area for the comparison with the slope corrected option. The observation date for the PALSAR image is July to September, 2009, where ranges in month occurs because the product is a mosaic image of the area with multiple scene tiles. The year 2009 data was used because both 10 m and 25 m data was available only for that year for the comparison, so that data was utilized for our analytical purpose. The PALSAR specification of the study area is at ascending Fine Beam Dual (FBD) polarization, characterized by 34.3 degree of off nadir angle.
We have collected a forest inventory data provided by the Prefectural Government of Chiba, which was obtained from the field observation implemented by the Chiba Prefectural Government, Agriculture, Forestry and Fisheries Department, Forest Division. Observations were made through 2011 and 2012 academic year at the prefectural owned forests located central to southern regions of the Prefecture. The data compiles with the information of tree type (Sugi or Hinoki), tree age, mean diameter at breast height (DBH), mean tree height, stem volume per unit area, stem density per unit area, mean basal area, and some other parameters which indicates the geological position of where the observations were made in terms of aspect and the position of whereabouts on the mountain (e.g. ridge). Information of the trees are collected within a plot area (0.01 ha) with the shape being similar to the satellite image pixel, which is normally square, and the coordinates of the plot is recorded only at the centre of the plots. Total observation plot size results up to 1939 plots; 838 for year 2011 (Central: 588 Southern: 250) and 1101 for 2012 (Central: 672 Southern: 429). Central and southern plot data on 2011 and the plot data at the central of 2012 will be used for the statistical analysis between satellite data and the field observation data.
Vegetation Continuous Fields (VCF) data developed by the university of Maryland [
The PALSAR image used in this study was converted from the provided format which is in the Digital Number (DN) values, to the backscattering intensity information also known as backscattering coefficients or the Normalized Radar Cross Section expressed using Equation (1):
where σ0 is the backscattering intensity represented in decibel units (dB) and CF is the calibration factor for the data obtained, depending on the observation period and polarization [
We have carried out a statistical analysis for formulating relationships among the field observation data of the forest stand characteristics and the remotely sensed microwave satellite data. The process will be taken by analyzing the relationships between SAR backscattering intensity (σ0) and the stand characteristics (DBH, tree height, and stem volume) for both 10 m and 25 m pixel spacing PALSAR image to compare the differences of the results from higher and lower pixel spacing. The procedure for the relationship analysis was taken through multiple approaches:
a) Combine all the observation plots and make the analysis;
b) Divide the plots into per forest percentage cover;
c) Divide the plots to different stem volume range (lower range or higher range);
d) Divide the plots into separate species (Sugi and Hinoki).
We attempted to do this because the scattering mechanisms of the SAR backscatter in the vegetated areas are very complex; usually it is difficult to see any trends when we come across making relationships with the stands in various range of the structure when combined [
Statistical relationship between backscattering intensity (σ0) and each forest stand characteristics (DBH, Tree height, stem volume) is investigated by using least-square method on the basis of the field observation plot. For reference, a regression line is drawn using second order polynomial. First, we have used all the plot information and made the relationship analysis (
It is obvious that we are seeing no relationship between the two. Usually, when empirical approach is taken, it doesn’t show much relation because of the complexity in the scattering at the forested areas [
1) Diffuse scattering from the ground (no vegetation);
2 and 3) Direct scattering from various vegetation components;
4) Double-bounce vegetation?ground interaction;
5) Corner reflector between tree trunks and ground;
6) Direct backscatter from the forest canopy;
7) Volume scattering from within the forest canopy;
8) Diffuse scattering from the ground (with vegetation);
9) Shadowing by parts of the forest canopy of other parts of the canopy or the ground.
When we are making relationship with a forested area that is varied with different growth stages and at different density of the trees and different structure of the trees, it is very difficult to see a trend because of all of those different scatterings, which likely results as on
Iizuka and Tateishi [
Figures 5(A)-(C) show the relationship between the backscattering and the stand characteristics for DBH, tree height and stem volume respectively for the 10 m and 25 m pixel spacing data on both HH and HV polarizations, on the area where the forest cover percentage is below 65% coverage and at lower volume ranges (0 - 500 m3/ha).
For DBH, 25 m pixel spacing HH polarization showed the highest correlation (R2 = 0.109) among the others, which is similar for Tree height (HH: R2 = 0.189), but for Tree height, also the 10 m pixel spacing data showed some correlation too (HH: R2 = 0.1) although it is a very low trend. Of them all, stem volume shows the highest correlation with the backscattering, and in this case the HH polarization showed better correlation than the HV (HH: R2 = 0.216, HV: R2 = 0.107) at 25 m pixel spacing data. In the overall trend, we can confirm that comparing with the 10 m and 25 m pixel spacing data, the 25 m data shows higher correlation for all the biophysical parameters. We believe this is occurring from the smoothing of the local backscattering area generated at the low pixel spacing images. The differences to the local scattering would average out when the pixel spacing becomes lower; as a result, errors reduce and better correlations would be seen. Some studies also confirm the increase of the accuracy and higher correlation when the pixel spacing reduces [
Correlation between the backscatter and stand biophysical parameter showed better result in the 25 m pixel spacing PGM. So the third approach of the analysis will be based on the 25 m pixel spacing PGM (Figures 6(A)-(C)). The figures show the relation between the backscatter for the same as Figures 5(A)-(C), but the plots were separated into differences of the species: Sugi (Cryptomeria japonica) and Hinoki (Chamaecyparis obtusa) trees. The relationship was analyzed for these species individually for each polarizations.
Compared to the analysis made using both species combined, we can see better correlation for both species and the characteristics of those are conspicuous. For the Hinoki trees, increase in backscatter shows clearly with the increase in biophysical parameters, showing highest correlation for the tree height for HH (R2 = 0.443) and stem volume for HV (R2 = 0.282) however, for stem volume HH shows higher correlation (R2 = 0.385) than the HV. While for the Sugi trees, stand biophysical parameter decreases with increasing backscatter but at the highest backscattering point, the parameters increase. This phenomenon for the Sugi trees shows only at the lower volume case, and at the higher stem volume range (over 550 m3/ha), the relation of the biophysical parameter and the backscattering becomes obvious; backscattering decreases with increasing biophysical parameters (
Statistical relationship between the Sugi stand characteristics and the backscattering showed an astonishing result, showing that backscattering was decreasing with an increasing of the biophysical parameters. However, there were some questioning results for the Sugi trees at the lower volume range plots, where an increase in the parameter showed at the end of the increasing backscatter. We wanted to understand why this was happening, so closer analysis on those plots was carried out. From
(A)
(B)
(C)
(A)
(B)
(C)
dB | Mean DBH (cm) | Mean Height (m) | Density (N/ha) | Volume | Age | |
---|---|---|---|---|---|---|
Point 1 | −2.13 | 16.56 | 11.96 | 2500 | 471 | 20 |
Point 2 | −8.76 | 26.545 | 15.18 | 1100 | 475 | 57 |
Point 3 | −3.54 | 12.76 | 8.76 | 2900 | 177 | 22 |
Point 4 | −6.94 | 14.84 | 11 | 1900 | 197 | 18 |
from the table what we can clearly indicate is the difference in the stem density (number of trees) of the plot area, where higher stem density has higher backscattering information.
When we say stem volume per unit area, we can describe this in various ways from the differences in the thinning process of the forests [
One reason for this phenomenon could be considered from the differences in the moisture content in the stem of the Sugi and Hinoki trees. Minato et al. [
The Hinoki relationship with the backscattering is very similar to what is performed by other researchers with their relationships with the forests. Where, the backscattering increases with the increasing biophysical parameters. Compared to the Sugi trees, stem density was not much influencing the trend, and on HH polarization we could see that stem density was decreasing with increasing backscattering. When time passes and the trees grow, DBH and tree height increases and eventually the number of stems will reduce due to the thinning process. As an overall trend for both Sugi and Hinoki, this is likely to be the case. However, for the Sugi case backscattering decreased with decreasing stem density, but for the Hinoki trees, even when the number of stems reduced, the backscattering increased. For the Hinoki case, this trend could be answered that how it is described by Kobayashi et al. [
As mentioned in the former section, backscattering is occurring from various mechanisms within the forested area. For example, Ulaby et al. [
In our study, the relationship between the biophysical parameters and backscattering was performed in very limited conditions, such as areas that are less than 65 % forest cover considering the attenuation from the dense forests. It is said that P-band SAR can sense more to the parameters than the L-band because it has longer wavelength. We guess this is true if we look at the studies that use P-band radar for the relationship analysis, indicating higher saturation points than the L-band data [
In this study, we have performed a statistical analysis in relation of the backscattering intensity from the L-band PALSAR and the in-situ stand characteristics of Sugi (Cryptomeria japonica) and Hinoki (Chamaecyparis obtusa) trees. The relation showed an acceptable result for the Hinoki trees, where backscatter increased with increasing biophysical parameters (DBH, tree height, stem volume). However, for the Sugi trees, the trend showed completely opposite where the backscattering decreased with increasing biophysical parameters. It was surprising that totally opposite trends showed for the same coniferous trees. This difference is assumed that it is coming from the characteristics of the Sugi tree, where it contains of high moisture content resulting with high dielectric constant to affect the backscattering with the increasing stem density, and attenuations caused from increasing DBH and tree height. However, there are still needs of validation for this hypothesis and we need to verify on more relationship analysis for different species other than the Sugi and Hinoki trees because it might have different trends in the scattering we would like to consider this as our future works.
We thank the Chiba Prefectural Government, Agriculture, Forestry and Fisheries Department, Forest Division for providing the forest inventory data (field observation data).