Analysis Of The Variation In Land Cover Classes On Agriculture Land

Introduction

Despite the fact that the terms i. e. land cover and land use are used together simultaneously both have a distinct meaning. Land cover refers to the surface cover on the ground, including vegetation, built-up, water, and barren land. Identification, delineating and mapping of land cover is significant for regional as well as global monitoring studies, planning activities, and resource management. Remote sensing plays a vital role in land cover monitoring. Satellite sensor remote sensing images due to their synoptic view, map like the format and repetitive coverage are a viable source of gathering effective land cover information.

PCA is a cast-off machine learning algorithm and corporate in remote sensing by providing high classification accuracy while using small training data set. Assumes that multi-temporal data are highly correlated and change information can be highlighted in the new components. Two ways to apply PCA for change detection are: (1) put two or more dates of images into a single file, then perform PCA and analyses the minor component images for change information and (2) perform PCA independently, then subtract the second-date PCA image from the corresponding. Analyst’s skill in identifying which component best represents the change and selecting thresholds. Remote sensing (RS) data have been one of the most important data Sources for studies of LC spatial and temporal changes. Till date as per the review of literature, very few works have been carried out SAR Data and principal component analysis to monitoring land cover changes. That’s why using that to generate an exclusive outcome of that study.

REVIEW OF LITERATURE: There is so many research works concentrating on the LAND COVER study. To assess the temporal variation in land cover classes studies has to be completed. Scientific papers based on the same topic are reviewed to Fulfill the objective of this project. Some useful concepts and related studies are summarized below:

1. Abdikan, Sanli, Ustuner, & Calò, (2016), used dual polarized Interferometric Wide swath mode (IW) sentinel-1 SAR DATA for mapping of land cover over built-up area of Istanbul megacity, Turkey. For classification of dataset, support vector Machine technique has been applied as pixel-based image classification to determine the different scenarios of that data. For analyzing the VV& VH polarization applying different set. To evaluate the result of the classified image by using Kappa coefficient and overall classification accuracy. It shows the result that, utilize the combination of dual polarimetric data, the overall accuracy increases up to 93. 28% against 73. 85% and 70. 74% of using individual polarization VV and VH, respectively. The target of the study is that characterized the capability of utilizing SAR DATA for megacities.
2. Chauhan & Srivastava, (2016), conducted a study over relative assessment of the sensitivity of multi-polarized SAR & optical Data for various Land Cover (forest, wheat, sugarcanes, built up, agriculture) of Haridwar district located in Uttarakhand. For land cover mapping examined the abilities of the dual-polarimetric (VV &VH) Envisat-1 ASAR and Landsat ETM+ data. In this study, the different band combination and combined the MDPI band to the multi-polarised SAR data, &NDVI band to optical data are used to retrieve the maximum information. By using the Transformed Divergence (TD) and Maximum Likelihood supervised classifier to explore the different class pairs and classification respectively. The maximum classification accuracy was 91. 25% with kappa coefficient of 0. 90 provided by using the fusion of Landsat & VH backscatter data. The outcomes demonstrated that buildup area is more precisely notable in VV band due to corner reflector effect & vegetation is very subtle in VH band due to volume scattering.
3. Deng, Wang, Deng, & Qi,( 2008) discussed the change detection technique which is related to multi-temporal, multispectral, and multi-sensor imagery that provided and delivered complete data for planning and decision-making over decades. Based on principal-component analysis (PCA) and hybrid classification methods Using multi-temporal and multi-sensor data (SPOT-5 and Landsat data) to observe land-use variations in an urban environment over the region of Hangzhou City, the capital of Zhejiang Province. From stacked imagery developed the change information using PCA technique & after that to recognized and quantify land use changes by hybrid classifier. The conclusion of this study land use changes have happened that is associated with the rapid economic development and urban expansion & shown maximum changes happen in cropland due to urban Encroachment.
4. In this study Hajj et al. , (2017), examined of sentinel-1 radiometric stability and quality for land monitoring by SAR temporal series data. Mean σ̋ value of first and third sub time series have the same value (i. e. 0. 3dB) and lower than the second sub-time series is approximately 0. 9dB.
5. In this study Sonobe, Tani, & Wang,( 2015) using extreme learning machine(i. e. newly developed single hidden layer neural ) as a supervised classifier using potential for multi-temporal ALOS/PALSAR images for classification of agricultural fields in the western To kachi plain, Hokkaido, Japan. Also compared the result of the k-nearest neighbor algorithm (k-NN) and ELM that are used for the classification of different type of crop in respective field a data. The results demonstrated that ELM classification (i. e. based on machine learning)was better than K-NN classification with overall accuracy of 79. 3% & for agricultural application L band & ALOS/PALSAR images is very advantageous.
6. Fichera, Modica, & Pollino, (2012), has analysed the land cover & its changes using REMOTE SENSING technique as GIS and landscape metrics utilize multitemporal data of aerial photos and Land sat TM & ETM+ data for the region of Avellino (Southern Italy) i. e affected by calamitous Irpinia

Materials and Methodology:

Dataset: For this research work Sentinel-1A satellite data has been used and about the dataset is discussed in detail: Sentinel-1A: ESA’s Sentinel-1A satellite was launched on 3 April 2014 and have Constellation of two satellites with the main objectives of Land and Ocean monitoring. It carries C-band Synthetic Aperture Radar (SAR). Each SENTINEL-1 satellite (1A or 1B) will be in a near-polar, sun-synchronous orbit, with a 12-day repeat cycle and 175 orbits per cycle.

A single SENTINEL-1 satellite is potentially able to map the global landmasses in the Interferometric Wide swath mode once every 12 days, in a single pass either ascending or descending. Sentinel-1 carries an advanced radar instrument to deliver an all-weather, day-and-night supply of imagery of Earth’s surface. Advantage of Sentinel-1A sensors is that it is freely available and has open access with dual polarization configuration (VV-VH).

Software: We can work on SNAP, NEST and PolSARPro software toolboxes for working with SENTINEL-1 data products. In this project we are using: SNAP: The Sentinel Application Platform (SNAP) is a collection of executable tools and Application Programming Interfaces (APIs) which have been developed to facilitate the utilization, observing and processing of a variety of remotely sensed data.

The functionality of SNAP is retrieved through the Sentinel Toolbox. SNAP toolbox (S1TBX) contains multiple processing tools i. e. applicable not only for SENTINEL-1 but also for ENVISAT ASAR ERS-1 & 2 RADARSAT-2 TerraSAR-X/TanDEM-X ALOS 1 & 2 COSMO-Skymed and many more. SNAP tool supports all type of SAR DATA functionality like Calibration Speckle Filtering, Terrain Correction,Ellipsoid Correction, SAR Simulation,Mosaicking,Reprojection, Coregistration,Interferometry.

Proposed methodology: Description:

• The planned workflows for this project initiated with the acquiring the Sentinel-1 satellite Data that required to calibrate after that using refined lee, speckle filtering was applied & for geometric correction choose TERRAIN correction.
• While stacking the all the imagery using nearest neighborhood method of resampling images were co-registered.
• After that identify the classes of temporal SAR data and create the ROIs & parallel work on principal component analysis to generate the PCA of year 2016-17 & 2017-18 and after that change detection technique is used for monitoring the various land cover classes
• Extraction of signature using the different classes of ROIs and study over that signature.
• At last,all these steps are used to conclude the final outcome. So, this study is quantitative in nature and carried out in a positivistic pattern. It experimented with multiple classifying techniques and scenarios, which were then assessed to determine their efficacy for differentiating crops or other classes within the study area.

EXPECTED OUTCOMES: The study is based on monitoring of land cover classes. As,the project estimates the monitoring of various land cover classes over Agra & Mathura region in a year (2016-2018) using multitemporal dual polarized Sentinel-1A SAR data.