Biophysical Spectral Response Modeling Approach for Forest Density Stratification
3.2 Methodology
The digital image processing for evergreen forests of South Andaman Forest Division has been done using PC based MAI (Modular GIS Advanced Imager) of Intergraph package on Windows NT. The Landsat TM bands TM bands (except bands 6) were normalized using linear transformation. The temperature calibration using coefficients for Landsat 5 was done to estimate ground temperature. The temperature data has only been used to separate soil and non-tree shadow. The colour images produced from Landsat TM raw bands 4,3,2 and 5,4,3 provide valuable information on the forest cover type distribution.
The digital image processing for dry deciduous forests of central India has been performed on IBM RS-6000 series using EASI PACE software. The spectral dataset is subjected to physical transformation using enhancement techniques. Attempts have been made to isolate vegetation cover, soil background influence and canopy shadow been made to isolate vegetation cover, soil background influence and canopy shadow
pattern from Landsat TM data. The vegetation feature space data was stratified based on the 'texture' of the data as influenced by the canopy shadow. Finally a rule based logic is implemented to achieve land cover and forest density classification. In both the case the results have been improved by using water mask from Landsat TM band7.
The process and steps involved in calculating Biophysical Spectral Indices are given as below (Anon., 1993 and Rikimaru, 1996):
3.3 Normalisation of Landsat TM Bands
The Landsat TM bands (ecept band 6) were normalized using linear transformation (equations 1 and 2)
| a |
= (Y1-Y2)/ (X1-X2) |
| |
= (20-220) / {m-2s) - (m
+2s)} …………………….eq. 1 |
| |
= 50/s
|
| X1
= m-2s
X2
= m+2s
|
| Y1 = 20 Y2 = 220 |
| B = -ax1 +y1 or Y= aX +b ……eq. 2 |
Y will normalize the Landsat TM band and X is the original Landsat data.
3.4 Temperature calibration
The normalization operation is not conducted for band 6 due to treatment of temperature calibration. The temperature calibration of the thermal infrared band into the value of ground temperature has been done using equation 3 and 4. The calibration coefficients used for the study are as per Pat, 1989.
| L = L min + ((L max-Lmin)/255) x Q ………….. eq. 3 |
| L: value of radiance in thermal infared |
| Q: digital record |
| L max: value of radiance = 1.500 mw/cm2/str(Q=0) |
| L min : value of radiance = 0.1238 mw/cm2/str (Q=255) |
| T = A2/in (Al /L +1) |
| T: ground temperature (k) |
3.5 Advanced vegetation index
NDVI is unable to highlight subtle differences in canopy density. It has been found to improve by using power degree of the infrared response. The index thus calculated has been termed as advanced vegetation index (AVI). It has been more sensitive to forest density and physiognomic vegetation classes. AVI has been calculated using equation 5.
| AVI = {(B4 +1) x (256-B3) x B4/3]1/3 …………..eq. 5 |
| B43=B4-B3 |
| IF (B43.LT.O) GO TO f: if B4<B3 after normalization, f AVI = 0 |
3.6 Bar soil index
The bare soil areas, fallow lands, vegetation with marked background response are enhanced using this index. Similar to the concept of AVI, the bare soil index (BI) is a normalized index of the difference sums of two separating the vegetation with different background viz. completely bare, sparse canopy and dense canopy etc. BI has been calculated using equation 6 and 7.
| BI = BIO x 100 +100 …………………………………… eq. 6 |
| BIO = {(B5 +B3) ((B4 + B1)(/((B5 +B3) + (B4+B1)}. Eq. 7 |
3.7 Canopy shadow Index
The crown arrangement in the forest stand leads to shadow pattern affecting the spectral responses. The young even aged stands have low canopy shadow index (SI) compared to the mature natural forest stands. The later forest stands show flat and low spectral axis in comparison to that of the open area. SI has been calculated using equation 8.
SI={(256-b1) x (256-B2) x (256-B3)}1/3 …………………..eq. 8
Canopy shadow index has been further improved by calculating logarithmic amplification of canopy shadow index. The index has been calculated using equation 9.
| SIL = {(256-BL1) X (256 - BL2) x (256 -BL3)}1/3 ………….eq. 9 |
| BLI = 100 x Log (B1-B1 min +1) |
| BL2 = 100 x Log (B2-B2 min +1) |
| BL3 = 100 x Log (B3-B3 min +1) |
The maximum value of SI within the 3 x 3 spatial filter is incorporated into the central cell of the filter to calculate the textured advance shadow index (ASIO)
| AIO = Max (SI1, SI2,SI3, SI4, SIO, SI5, SI6, SI7, SI8) …eq. 10 |
| IF (AVIO.LT.THV) ASIO = 0: |
If the AVI value is lower than the standard value, it is not recognized as forest shadow, It may be a gap in the forest canopy. |
| IF (TO.T.THT) ASIO = 0: |
If the ground temperature is higher than the standard value, is not recognized as forest shadow. |
3.8 Classification
Two methods have been followed to achieve the classification of forest type and density. In the first approach forest density stratification has been achieved using three way model using AVI, BI and SI. In the second approach the rule -based approach has been followed wherein the AVI, BI and SI database has been subjected to clustering. The clusters have been reclassed using the ground truth data to label them. The density classes are presented along with the forest type and their succession stages for information completeness. The hybrid classification approach (combination of unsupervised and supervised) has been used for classifying forest types. Keeping forest type map in the background the forest density map has been reclassified using logical operator to prepare final forest type and density map.
