Stratification of density in dry deciduous forest using satellite remote sensing digital data—An approach based on spectral indices.

Forest density expressing the stocking status constitutes the major stand physiognomic parameter of Indian forest. Density and age are often taken as surrogate to structural and compositional changes that occur with the forest succession. Satellite remote sensing spectral response is reported to provide information on structure and composition of forest stands. The various vegetation indices are also correlated with forest canopy closure. The paper presents a three way crown density model utilizing the vegetation indices viz., advanced vegetation index, bare soil index and canopy shadow index for classification of forest crown density. The crop and water classes which could not be delineated by the model were finally masked from normalized difference vegetation index and TM band 7 respectively. The rule based approach has been implemented for land use and forest density classification. The broad land cover classification accuracy has been found to be 91·5%. In the higher forest density classes the classification accuracy ranged between 93 and 95%, whereas in the lower density classes it was found to be between 82 and 85%.

Biomass estimation using satellite remote sensing data—An investigation on possible approaches for natural forest.

Vegetation type and its biomass are considered important components affecting biosphere-atmosphere interactions. The measurements of biomass per unit area and productivity have been set as one of the goals for International Geosphere-Biosphere Programme (IGBP). Ground assessment of biomass, however, has been found insufficient to present spatial extent of the biomass. The present study suggests approaches for using satellite remote sensing data for regional biomass mapping in Madhav National Park (MP). The stratified random sampling in the homogeneous vegetation strata mapped using satellite remote sensing has been effectively utilized to extrapolate the sample point biomass observations in the first approach. In the second approach attempt has been to develop empirical models with satellite measured spectral response and biomass. The results indicate that there is significant relationships with spectral responses. These relationships have seasonal dependency in varying phonological conditions. The relationships are strongest in visible bands and middle infrared bands. However, spectral biomass models developed using middle infrared bands would be more reliable as compared to the visible bands as the later spectral regions are less sensitive to atmospheric changes. It was observed that brightness and wetness parameters show very strong relationship with the biomass values. Multiple regression equations using brightness and wetness isolates have been used to predict biomass values. The model used has correlation coefficient of 0·77. Per cent error between observed and predicted biomass was 10·5%. The biomass estimated for the entire national park using stratified and spectral response modelling approaches were compared and showed similarity with the difference of only 4·69%. The results indicate that satellite remote sensing data provide capability of biomass estimation.

Space remote sensing for spatial vegetation characterization.

The study area, Madhav National Park (MP) represents northern tropical dry deciduous forest. The national park, due to its unique location (nearest to township), is under tremendous biotic pressure. In order to understand vegetation structure and dynamics, vegetation mapping at community level was considered important. Prolonged leafless period and background reflection due to open canopy poses challenge in interpretation of satellite data. The vegetation of Madhav National Park was mapped using Landsat TM data. The ground data collected from sample points were subjected to TWINSPAN analysis to cluster sample point data into six communities. The vegetation classification obtained by interpretation (visual and digital) of remote sensing data and TWINSPAN were compared to validate the vegetation classification at community level. The phytosociological data collected from sample points were analysed to characterize communities. The results indicate that structural variations in the communities modulate spectral signatures of vegetation and form basis to describe community structure subjectively and at spatial level.