Professional Documents
Culture Documents
Assement of Sentinel-2 Vegetation Indices For Plot Level Tree AGB Estimation
Assement of Sentinel-2 Vegetation Indices For Plot Level Tree AGB Estimation
By : Mehboob Alam
MS 3 RS & GIS, 140311060
Supervisor: Dr Saleem Ullah,
Co Supervisor Dr Waqas Qazi
1. Introduction
2. Literature Review
3. Data Sets
4. Methodology
5. Results
6. Conclusion
7. Discussion
Definitions
Non Destructive:
Utilize Allometric relationships between
the tree stand parameters (DBH, Height, CPA)
that approximates the biomass of the tree
component or the total biomass of single
trees according to more easily measured
variables. (Chave et al., 2005
Limitations of Traditional Methods
Tiwari & Singh (1984) estimate the Indian forest AGB using
aerial photographs.
The passive satellite sensor data with visible and NIR reflectance
from the earth surface have formed the basis for most of current
global scale mapping and forest biomass estimation.
These are very common datasets and many scientists have
assessed the capability of passive satellite data for biomass
estimation using different types of optical data and image
processing techniques (Foody et al., 2003; Goetz et al., 2009; Lu,
Passive RS Satellite Data Contd
The studies about forest biomass using optical datasets are mostly
involved in establishment of a relationship between the
commonly used vegetation indices with biomass or some
biophysical factors (Olander et al., 2008).
Passive sensors are generally characterized by their spatial,
spectral, and radiometric performances. Satellites are typically
categorized according to their spatial resolution into
environmental satellites (NOAA, GOES, Meteosat), medium
resolution satellites (IRSI, Landsat TM, MSS, etc.); and high
resolution satellites (Quickbird,Pleiades, SPOT, IKONOS, ERS-
1.
Passive RS Satellite Data Contd..
The most commonly used optical datasets for biomass estimation
are Landsat TM(Foody et al., 2003; Lu, 2005b) Landsat ETM+
(Cartus, Kellndorfer, Rombach, & Walker, 2012; Morel, Fisher, &
Malhi, 2012; Zheng et al., 2004), IKONOS (Abbas et al., 2013; P.
S. Thenkabail et al., 2004) Quickbird (Peter Hyde et al., 2006),
SPOT-5 HGR, NOAA AVHRR, MODIS (Baccini, Friedl,
Woodcock, & Warbington, 2004), and ASTER (Muukkonen &
Heiskanen, 2005).
introduced for biomass assessment from the reflectance images,
and different types of statistical modelling were also applied to
correlate NDVI values or band ratio values with field
measurements. These several studies shown that approaches using
passive remote sensing data are not suitable for most terrestrial
biomass densities because of the saturation problem (Sarker,
Optical Medium Resolution Data Contd
Many different models are available which can be develop for
AGB estimation (Rosillo-Calle, 2007).
Whereas NDVIs have been recommended to remove
inconsistencies caused by soil background, canopy geometry, sun
view angles, and Atmospheric conditions, while measuring
biophysical characteristics of the trees (Elvidge & Chen, 1995;
Blackburn & Steele, 1999).
However, not all NDVIs are expressively correlated with forest
biomass. In general, NDVIs can partially reduce the impacts on
reflectance caused by terrain and canopy shadows, environmental
conditions, and consequently improving correlation between
AGB and NDVIs, particularly at those sites which have
multifaceted vegetation stand structures (Lu, Mausel, Brondizio,
& Moran, 2004).
High Spatial Resolution Data
Deodonia Viscosa
Vegetation Red
B08A 20 865 20
Edge/NIR
Date: 2016-10-08T05:47:22.000Z
Identifier: S2A_OPER_PRD_MSIL1C_PDMC_20161008T115539_
R048_V2061008T054722_20161008T055257
Instrument: MSI
Satellite: Sentinel-2
Index
MTCI , The Meris Terrestrial (NIRRE)/ (RE R) (B6B5)/(B5B4) (Dash and Curran 2007)
NDI45, (NIR R)/ (NIR +R) (B5 B4)/ (B5 + B4) (Delegido et al. 2011b)
NDVI, Normalized Difference (NIR R)/ (NIR + R) (B7B4)/(B7+B4) (Rouse et al. 1973; Wan, Wang,
REIP Red-Edge Inflection Point 702+40((R +NIR/2)- NIR / RE2RE1) 702+40((B4+B7/2)-B5/B6-B5 ) (Samalca 2007;Timothy et al.
S2REP, Sentinel-2 Red-Edge 705+35*((((NIR+R)/2) RE1)/(RE2RE1) 705 + 35 * ((((B7 + B4)/2) B5)/ (B6 (Frampton et al. 2013),
Position B5))
SAVI, Soil Adjusted Vegetation (RNIRRR)/ (RNIR + RR + 0.5) 1.5 (B8- B4)/ (B8- B4+.5) *1.5 (Heiskanen 2006; Huete et al.
Field Data
The AGB values from Field Processing Was recorded as attribute In the Point
Shape file Containing the Location of Measured Plots.
Shapefile was overlayed over the Each vegetation index calculated.
Through Spatial analysis the Cell value was extracted and recorded as attribute.
The Attribute table containing the AGB Value, and Cell value was exported as
Excell and then to matlab for regression analysis.
Spectral Profile of Cheer Pine
Chir Pine
Spectral Profile For Paper Mulberry
Paper Mulberry
Spectral Profile Of All Species Combined
All Species Combined
Conclusion
This study is concluded by revisiting the Questions asked and the objective outlined in
the first chapter.
How is much is the Biomass measured in the field correlated with changes in the spectral
response?
The AGB is poorly related to the spectral response.
What are the best spectral bands for different species which explains best the variation
in the biomass for that specie?
For Cheer Pine SAVI, For Paper Mulberry IRECI while for combined species SAVI
produces better result compared to the other indices.
Does the generalized model for different species gives any promising results better than
the model for individual tree species?
No, it does not produce any better result. As clear from the tables and figures in the
result section, the correlation further becomes weak when all the species are modelled in
the bulk.
The findings reject the hypothesis that was assumed in the chapter 1. It therefore
concludes that biomass cannot be calculated based on Vegetation indices alone.
Possible Reasons.
Above Ground Tree biomass is directly related to the Count of trees present in the area, the tree size and
height. For the better relationship, these characteristics needs to be captured with reasonable level of
accuracy
High resolution satellite imageries of Quick Bird, Ikonos, Spot or other satellite possibly can have
promising correlation due to their high spatial resolution
Due to their high resolution, they can better represent the canopy structure, which is as much important as
the spectral response characteristics of the canopy
Lidar gives high resolution horizontal and accurate vertical information (Lim et al,
2003). Hyde et al. (2006) indicates that LiDAR alone can best estimate canopy height
and biomass information, and the addition fusion of sentinel 2 can improve the results.
One of the possible option for Biomass Estimation is UAV Remote sensing. The UAVS
gives the user the control over the flight time and the choice of sensors which are
suitable to the study. Compared to the satellite data they are cheaper and can be
configured to need of the phenomenon to be studied.
Thank You