selection of an appropriate classification technique for coastal biomass mapping using high and low...
TRANSCRIPT
Selection of an appropriate classification technique for coastal biomass mapping
using high and low resolution dataset
Urooj Saeed, GIS CoordinatorWorld Wide Fund for Nature – Pakistan
Presented at 2nd International Conference on Advances in Space Technologies29-30 November, 2008; Islamabad, Pakistan
2nd International Conference on Advances in Space Technologies
Presentation Outlines
Research Questions Study Area Methodology Analysis and Results Conclusions and recommendations
2nd International Conference on Advances in Space Technologies
Research Questions
Maximum Likelihood Classification (MLC) or Sub Pixel Classification (SPC)- which is time and cost efficient technique for coastal biomass mapping?
Can medium resolution satellite data be used as an alternate of better-resolution satellite data?
2nd International Conference on Advances in Space Technologies
Sandspit is located at the west of Liyari river, near Hawks Bay along the Karachi coast line.
Area = 1618 ha
Major vegetation types : Mangrove Saltbush Algae
Study Area
2nd International Conference on Advances in Space Technologies
Satellite Data Used Landsat image of 30 m resolution
Spatial resolution 30m Acquisition date 6th October, 2001 Tide Height Value2m
Quickbird image Spatial resolution 2.4m Acquisition date 27th April, 2003 Tide Height Value2m
2nd International Conference on Advances in Space Technologies
Data acquisition & preprocessing Field Survey Image Classification
Maximum Likelihood – Landsat and QuickBird
Sub Pixel – Landsat Comparison and Analysis
Methodology
2nd International Conference on Advances in Space Technologies
Field Visit Avicennia marina – dominant mangrove
species, height varies from 5 to 20 ft. More the distance from the creeks less is the
density/height of the mangroves Floating algae was also observed GPS coordinates of different forest density
classes were recorded
2nd International Conference on Advances in Space Technologies
Maximum Likelihood Classification (MLC) MLC calculates different statistical
parameters from the inputs known as training areas and on the basis of these parameters it assigns a specific class to certain pixel
In this study MLC technique was used to develop output maps of both the datasets i.e Quick Bird and Landsat
2nd International Conference on Advances in Space Technologies
2nd International Conference on Advances in Space Technologies
MLC results – Landsat MLC results – QuickBird
2nd International Conference on Advances in Space Technologies
Land cover Classes QuickBird (ha) Landsat (ha)
Tall Mangroves 57.45 87.83
Medium Mangroves 125.9 216.41
Small Mangroves 180.89 113.68
Regeneration 5.35 Nil
Sub-Total for Mangroves 370 417.92
Salt Bushes 38.57 Nil
Floating Algae 131.03 Nil
Algae on mud 42.82 Nil
Sludge/wet soil 11.18 66.11
Water 364.04 378.91
Settlements 45.86 119.35
Salt pans 25.61 116.56
Mudflats 149.52 155.39
Land soil 444.66 398.04
Tabular comparison of MLC results
2nd International Conference on Advances in Space Technologies
Sub Pixel Classification (SPC)
SPC is used to map the landcover classes which are smaller than the pixel size
Five major modules were used to run this algorithm
In this study SPC was applied only on Landsat satellite image
2nd International Conference on Advances in Space Technologies
2nd International Conference on Advances in Space Technologies
SPC results for MangrovesSPC results for SaltbushSPC results for Algae
2nd International Conference on Advances in Space Technologies
M
A
S
Association map of Mangroves, Saltbush and Algae
2nd International Conference on Advances in Space Technologies
LandcoverClasses
QuickBird(ha)
Landsat(ha)
Difference(ha)
MLC MLC SPC Landsat (MLC) – QuickBird (MLC)
Landsat (SPC) – QuickBird (MLC)
Mangroves 376.65 417.92 375.56 41.27 - 1.09
Saltbush 26.05 Nil 26.33 - 0.28
Algae 42.82 Nil 67.65 - 24.83
Tabular comparison of SPC and MLC
2nd International Conference on Advances in Space Technologies
Conclusions and Recommendations
MLC when applied on Landsat failed to separate a very important vegetationcomponent into a different landcover classes.
By using SPC, mixed pixel problem was satisfactorily overcome by classifying high spatialfrequency vegetation classes i.e. saltbush and algae
SPC proved to be a time effective technique as less training areasfor Material of Interest (MOI) were required
Due to the time and cost effectiveness, it is highly recommended to use SPC on medium resolution data for coastal biomass mapping at large area.
It is also suggested to use selective patches of high resolution satellite image forthe ground truthing and MOI definition.
2nd International Conference on Advances in Space Technologies
2nd International Conference on Advances in Space Technologies
Thank Thank youyou