Inventory of Irrigated Rice Ecosystem Using Polarimetric Sar Data

An attempt has been made in the current study to assess the potential of polarimetric SAR data for inventory of kharif rice and the major competing crop like cotton. In the process, physical process of the scattering mechanisms occurring in rice and cotton crops at different phonological stages was studied through the use of temporal Radarsat 2 Fine quadpol SAR data. The temporal dynamics of the volume, double and odd bounce, entropy, anisotropy, alpha parameters and polarimertic signatures, classification through isodata clustering and Wishart techniques were assessed. The Wishart (H-Į) classification showed higher overall as well as rice and cotton crop accuracies compared to the isodata clustering from Freeman 3-component decomposition. The classification of temporal SAR data sets independently showed that the rice crop forecasting can be advanced with the use of appropriate single date polarimetric SAR data rather than using temporal SAR amplitude data sets with the single polarization in irrigated rice ecosystems


INTRODUCTION
In India, rice is grown under diverse agro-ecological conditions and Irrigated rice ecosystem contributes about 45% to the rice cropped area.Hence, monitoring of these ecosystems in timely and efficient manner is very important for agriculture, environmental sustainability, food and water security.Multispectral data from IRS series of satellites have been effectively used in India for crop inventory but its use is limited for kharif rice due to non-availability of sufficient cloud free data.In this context, microwave remote sensing offers great potential for monitoring crops especially during the monsoon season due to capability of radar systems to acquire data under all weather conditions.
Studies conducted in India and elsewhere have shown the potential of radar remote sensing techniques to crop management (Macelloni, 2002;Brown et al 1999;and Shao Yun et al, 2001).Further, studies conducted in India have demonstrated the utility of temporal SAR data for rice crop monitoring.Temporal SAR data sets of amplitude component are being used to monitor the paddy crop (Panigraphy et al, 1999, andChakraborty et al, 2005).Research has demonstrated that the additional polarizations will increase the information content in a SAR dataset similar to using multispectral approaches in the optical region (Mc Narain andBrisco, 2004, Feilong et al, 2005).The multi-polarized configurations provide more information related to crop structure and condition.The phase and polarimetric parameters does help in understanding the different scattering mechanisms from different surface features in general and crops in particular.The objective of the current study is to assess the potential of fully polarimetric data for inventory of irrigated rice ecosystem.

METHODOLOGY
The fine quadpol data from RADARSAT-2 covering Guntur (2 nd and 26 th October and 19 th November, 2010 with incidence angle 34.12 0 ) and Anantapur test sites (22 nd October, 2010 with incidence angel of 24.5 0 ) of Andhra Pradesh were used.The rice and cotton crops are the major crops grown in irrigated and rainfed regions respectively in the Guntur test site.While, the Anantapur test site covers only rainfed areas and the rice crop grown is confined to narrow valley regions where ground water is available for irrigation.Synchronous ground truth data like crop type, phenology, leaf area, plant height, fresh biomass and soil moisture were collected from the test sites on the day of data acquisition.
PolSARPro and ENVI s/w were used for processing of the data sets.The raw data was imported into the coherence matrix as well as amplitude format.The data was geometrically corrected using the GCP's given in the header file using ASF s/w module.The Gaussian-Boxcar Speckle filter with 5*5 size was applied to remove the speckle.The Entropy-Anisotropy-Alpha (H-A-Į) (Cloude and Pottier, 1996) and Freeman 3-component (Volume-Odd-Double) decomposition (Freeman and Druden, 1998) technique was applied on the polarimetric SAR data.The Wishart (H-Į) classification algorithm was implemented on the (H-A-Į) decomposed image, while isodata clustering was performed on Freeman 3-component decomposition image.The results were compared with the ground data collected synchronous to the date of pass.

RESULTS AND DISCUSSION
The incoherent decompositions methods viz.It was observed in cotton initially, the odd predominant and with the increase in leaf increase in the volume scattering has been is no significant change in the doubl mechanism.In the case of rice crop who vertical, there has been a significant increas followed by double bounce scattering mec of the temporal behaviour of the pa Anisotropy and mean alpha (Į) angle gener Į decomposition showed that the mean alph for the paddy crop compared to cotton, entr for cotton and rice crop with the age o   While, the cotton crop could be delineated with more than 87.5 percent accuracy and the highest accuracy was achieved in the month of November when the crop attained maximum vegetative cover.The higher accuracy with Wishart classifier could be mainly because, the coherence matrix derivable from Polarimetric SAR data nearly follows Wishart distribution and uses the whole Polarimetric information.Chen et al. ( 2007) also observed that Wishart classifier has achieved the best classification performance for rice class ranging from 62 to 92% using a multi temporal dual polarization sets.
Table 1.Classification accuracies for rice and cotton crops

CONCLUSION
The Wishart (H-Į) classification showed higher overall as well as rice and cotton crop accuracies compared to the isodata clustering from Freeman 3-component decomposition.The use of alpha and entropy parameters provide substantial information on the polarimetric properties of a scene and can be used to classify an image in a very simple way.The higher accuracies from Wishart H-Į classification could be mainly because of the use of whole polarimetric information.With the consistent higher rice crop accuracies, the rice crop forecasting can be advanced with the use of appropriate single date polarimetric SAR data rather than using temporal SAR amplitude data sets with the single polarization in irrigated rice ecosystems.For cotton crop, depending upon the crop growth, an early assessment can be made.These studies need to be tested in the rainfed upland rice, other rice ecosystems, as well as upscaling the study to cover large areas.

Wishart (H-Į) classification Isodata clustering on
Fig 1. FCC images of Freeman (a) and H-A- Fig 2. Scatterplot of Alph Fig 5. Classified image of Rice and Cotton

Freeman 3-component decomposition Rice crop accuracy % Cotton crop accuracy % Overall K^ statistic Rice crop accuracy % Cotton crop accuracy %
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XXXVIII-8/W20, 2011 ISPRS Bhopal 2011 Workshop, 8 November 2011, Bhopal, India