159
through the study site, acquired on 22nd February 2016, the details of which is provided in Chapter 3, was ordered for this work from National Remote Sensing Centre, India (https://www.nrsc.gov.in).
Methodology to derive the Crop Parameters from high spatial Multispectral data
8.3.1 Index based fusion approach for hyperspectral and multispectral data
An effort has been made to derive narrow band vegetation indices by adopting fusion of Hyperion narrow band with LISS IV broad band. The hyperspectral bands of Hyperion L1R Hyperion data are of 10 nm bandwidth, whereas LISS IV data has three bands: b2 (0.52-0.59 µm, green), b3 (0.62-0.68 µm, red) and b4 (0.77-0.86 µm, NIR). This fusion approach will be further implemented to retrieve chlorophyll and nitrogen content at plot scale within the rice field.
160 The relationship is stated in the following expressions.
1
1
GC G IH (8.1)
2
2
RC R IH (8.2)
3
3
NC N IH (8.3)
where,
n i
i
IH n 1
1
IH= broad band equivalent to integrated narrow bands of Hyperion image
i= broad band reflectance at i (i at a step size of 10 nm)
, = linear coefficients and GIH,RIH,
NIHare the three spectral (LISS IV) broad bands equivalent to integrated Hyperion narrow bands calculated for green, red and NIR regions of the spectrum, respectively, and
GC,
RC ,
NCare the critical narrow bands of green, red and NIR regions of the spectrum, respectively.
It was found that some of the critical bands were highly correlated to the broad bands equivalent to the integrated narrow bands, whereas others failed to reciprocate much.
Therefore, a higher linear coefficient of determination (R2 > 0.95) between critical bands and broad bands equivalent to integrated narrow bands derived from Hyperion bands was considered (Figure 8.1). By adopting band average concept method, bands like 533 nm, 565 nm, 681 nm, 705 nm, 717 nm, 750 nm and 800 nm were found to be significant amongst all the critical bands. Furthermore, the multispectral bands of LISS IV image were converted into three synthetic multispectral bands, which were derived as a function of Hyperion narrow bands sensitive to paddy crop.
TH-2211_126104014
161
Figure 8.1 : Correlation between critical bands and broad bands equivalent to integrated narrow bands, green: (a)-(b), red: (c)-(d), NIR: (e)-(f).
(f) R2=0.96 (e) R2=0.97
(c) R2=0.98 (d) R2=0.99
(b) R2=0.95 (a) R2=0.99
critical bands vs integrated narrow bands reflectance
TH-2211_126104014
162
For this LISS IV image representing the study area dominated with paddy crop cultivation, a linear relationship was established between multispectral broad band (so called synthetic band;
IV
SLISS ) and broad band equivalent to integrated narrow bands of Hyperion image (IH).
The relationship is represented in the following expressions.
G IH G
IV
LISS G
GS (8.4)
R IH R
IV
LISS R
RS (8.5)
N IH N
IV
LISS N
NS (8.6)
where,
and are linear coefficients, andIV
GSLISS ,
IV
RSLISS ,
IV
NSLISS are the three multispectral synthetic bands derived for Hyperion narrow bands and
GIH ,
RIH, NIHare the three broad bands equivalent to integrated narrow bands in green, red and NIR regions of the spectrum, respectively.
8.3.2.2 Spectral Shape Function Concept
The hyperspectral narrow bands are nothing but integration of measurements of the spectrum.
The bands illustrate the shape of the spectrum over a certain spectral region (Fensholt and Sandholt, 2003; Khanna et al., 2007), and this is used as a major source of information. In this context, the reflectance of the narrow bands, that capture spectral information of paddy crop, were used to fit a spectral shape, in order to add a new dimension to the biophysical parameter studies in case of paddy crop. The spectral shape function can be further used to get the critical wavelengths, which will provide more detailed information on paddy crop parameters in the field level. To identify the sensitive bands for establishing the spectral shape function, band to band correlation (Figure 8.2) was done for all the critical bands (533 nm, 565 nm, 574 nm, 681 nm, 695 nm, 705 nm, 709 nm, 717 nm, 740 nm, 750 nm, 800 nm) as per findings from Chapter 6. By considering the coefficient of determination, R2 > 0.95, significant wavelengths were selected. An interesting result was achieved showing that the wavelengths 533 nm, 681 nm, 705 nm, 717 nm, 750 nm and 800 nm wave bands were found critical from band to band correlation analysis.
TH-2211_126104014
163
0.9 0.92 0.94 0.96 0.98 1
0.52 0.53 0.54 0.55 0.56 0.57 0.58 0.59 0.60 Coefficient of determination (R2)
Wavelength (nm)
Green band 533 nm 565 nm 574 nm
0.8 0.84 0.88 0.92 0.96 1
0.62 0.63 0.64 0.65 0.66 0.67 0.68 0.69 0.7 Coefficient of determination (R2)
Wavelength (nm)
Red band 681 nm 695 nm 705 nm
0.7 0.75 0.8 0.85 0.9 0.95 1
0.76 0.77 0.78 0.79 0.80 0.81 0.82 0.83 0.84 0.85 0.86 Coefficient of determination (R2)
Wavelength (nm)
NIR band 709 nm 717 nm 740 nm 750 nm 800 nm
Figure 8.2 : Band to band correlation analysis to get the critical bands from spectral shape function method, (a) green, (b) red, (c) NIR
(a)
(b)
(c)
TH-2211_126104014
164
Results and Discussion
Due to non-availability of Hyperion data with good temporal resolution, predominantly it is not possible to monitor the rice crop parameters covering all the stages of rice crop age period. Secondly, it is a difficult task to find the paddy crop parameters like chlorophyll and nitrogen at field level from space platform. Therefore, an earnest effort has been made to estimate the parameters from multispectral data at plot scale within the crop field. To do so, index based fusion approach for multispectral and hyperspectral data was adopted by incorporating band average method and spectral shape function method. For the present study, estimation of paddy crop parameters from LISS IV imagery (of the Study site 2) acquired on 22nd February 2016 was carried out by using narrow band index regression models. These index models were specifically derived for paddy crop in Indian rice agriculture system from hyperspectral data.