Change Detection between Landsat 8 images and Sentinel-2 images

The technology of change detection is a technique by which changes are verified in a certain time period. Remote sensing images are used to detect changes in agriculture land for the selected study area located south of Baghdad governorate in Agricultural Division of AL-Rasheed district because this method is very effective for assessing change compared to other traditional scanning techniques. In this research two remotely sensed images for the study area were taken by Landsat 8 and Sentinel-2, the difference between them is one month to monitor the change in the winter crops, especially the wheat crop, where the agriculture began for the wheat crop there in the Agricultural Division of AL-Rasheed district at 15/11/2018. The first preprocessing procedure was the extraction of the NDVI (Normalized Difference Vegetation Index) values for the two scenes of Landsat 8 and the two scenes of Sentinel-2B and then using the change detection between them to compare the changes in agriculture land. Also, change detection was implemented between NIR bands because they are most severely affected by biomass or the amount of available chlorophyll-containing in plant structures. The results of the change detection for Sentinel-2B were more accurate than for the Landsat 8 as demonstrated by field visits for the study area, where the changes in the distribution of vegetal cover (wheat and other winter crops) were clear and accurate in the image of Sentinel-2B, as opposed to Landsat's 8 image, where the variation in vegetation cover was not accurate, especially for the change detection between NIR bands.


Introduction
The change detection is the process of identifying differences in the characteristics of the Earth by observing them at different times. This process can be carried out manually or by remote sensing programs (S. Bazgeer, 2005) [1] . By analysis of atmospheric photographs, where the analysis is used to perform the equation between image and image and detect the change zones. (Wright, Lillesand, & Kiefer, 1980) [2] To detect the variable after analysis, two different dates are categorized independently. The area of change is then extracted by a direct comparison of the categorization results (Duro, 2012) [3]. One of the benefits of detecting post-categorization is that it transcends difficulties in detecting change associated with the analysis of acquired images at various times of the year or by various sensor (Richards, Jia, Richards, & Jia, 2013) [4] . The flaw of the postcategorization approach include more computational requirements and distinguish, high sensitivity to the precision of the individual categorization, as well as difficulties in accurately evaluating historical data sets. Because all optimization methods depend on processes or pixel scenes as well as pixel operations, accuracy in registering images and basic registering is more important to these methods compared to other methods (Clark, 2002) [5].
The changes detection have been done between the images of Landsat 8, and between the images of Sentinel-2B. The Landsat program is a large project to operate satellite images of the Earth. The satellite was launched on July 23, 1972, for land resources technology. The development of Landsat 8 was the result of communion between the US space agency (NASA) and the US Geological Survey (USGS)(Rachel, Paul, Rowland, & Ross, 2000) [6]. The selection of Landsat spectral bands is directed at different types and sizes of plants. One of the most important strengths of Landsat, in general, is that it re-inspect every spot on earth every 16 days, and has a long-term data archive so that the images captured by it can be compared from 1982, and also has relatively rich spectral information [7]. As for the satellite Sentinel-2B, It is an Earth-monitoring expedition from the (EU) Copernicus Program that obtains methodical optical imagery with high spatial resolution (10m to 60m) above the land and coastal waters. The launch of the first satellite, Sentinel-2A, occurred 23 June 2015 at 01:52 UTC on a Vega launch vehicle. Sentinel-2B was launched on 7 March 2017 at 01:49 UTC, also aboard a Vega rocket. Where this family of missions consists of two twin satellites (Sentinel-2A and Sentinel-2B). This task provides very wide services and applications such as agricultural monitoring, emergency administration, land cover classification or water quality [8].

The study area
The study area is selected for this research because it is famous for the cultivation of the wheat crop which is considered a strategic crop and is a major source of food security for the country. In addition, the ease of communication with the study area is important for the purpose of conducting field studies and identifying the features of agricultural land. The selected study area is located south of Baghdad governorate in Agricultural Division of AL-Rasheed district, where it showed in green as shown in Figure-

Methodology of Work
The entirety of this study was conducted using ArcGIS software specifically, and ERDAS program. Repeated imaging for the study area enables the estimation of changes in the type or condition of surface features. This is one of the most important of all analyses in remote sensing, typically called change detection. Many of these analyses use images acquired at two points in time, the primary focus of this research. The Change Detection, between NDVI imageries for the two of Landsat 8 scene on different dates, and also the Change detection, between NDVI imageries for the two of Sentinel-2B scene on different dates, was extracted. On the other side the Change detection, between the two of Landsat 8 scenes at the near-infrared region, and the Change detection, between the two of Sentinel-2B scenes at the near-infrared region, was extracted. For change difference for each band has been illustrated in the below equation (1) (Muhsin, 2016) [9].
(1) Where: The NDVI (band ratio) is a structure from every input image, and the difference can be taken between the band ratios at different times. The change detection can be used on many other remote sensing methods. As in equation (2) (M. K. Alfarttoosi, 2016) [10].

Change Detection between NDVI imageries
After extracting the NDVI values of satellite images of the Landsat 8 satellite and Sentinel-2 satellite for the study area (in the Agricultural Division of AL-Rasheed district), Change detection of the NDVI imageries is obtained with different dates, where they included one scene for Landsat 8 and included one scene for Sentinel-2B. a. NDVI scenes from Landsat 8 include Figure-2

Abd-Alwahab and Ghazal
Iraqi Journal of Science, 2019, Vol.60, No.8, pp: 1868-1876 1874 Change detection, between NDVI imageries for Sentinel-2B scenes for the Agricultural Division of AL-Rasheed district, as shown in Figure -7. It can be observed that there is a significant increase in the growth of winter crops as shown in dark green and light green.

The Change detection, between NIR bands for satellites images
Change detection is obtained for the NIR (near-infrared region) bands for satellite images of Landsat 8 and Sentinel-2B where they included one Scene for Landsat 8 and for Sentinel-2B.  b. The Change detection, between Sentinel-2B images at near-infrared region (0.842-0.957 µm) in the Agricultural Division of AL-Rasheed district, as shown in Figure-8. It produces a very accurate scene of the change detection in the study area. Where it illustrates an increase in the area of cultivated land (wheat and other winter crops) and the shrinking of uncultivated land. It also shows a scene similar to what it is in change detection, between NDVI imageries of Sentinel-2B. 3. Sentinel-2B imageries are very active in observation of the crop health status, the account of crop area, and crop productivity for the regular and irregular farms because of its high spatial resolution compared to Landsat 8 4. The results of the change detection for the study area showed four classifications, according to the increase and decrease in the vegetation cover. A. A significant increase in winter crop growth such as Lettuce B. Some increase in the growth of some winter crops such as wheat. C. A significant decrease in summer crop growth due to harvest D. Some decrease in the growth of weeds and jungles because it was treated by farmers. 5. Field visits proved that these results were more accurate for the image of Sentinel-2B compared to the results shown by the image of Landsat 8 in respect of the change detection between NDVI imageries. 6. The NIR band of the Sentinel-2B is influenced by the amount of available plant structures containing chlorophyll more than NIR band of Landsat 8, so the results of change detecting between NIR bands were inaccurate for the Landsat 8, as shown in Figure-8.