Coastal Ecosystem Monitoring Using Long-Term Satellite Data

Vivek G. 1 , Santonu Goswami 1 , R.N. Samal 2 , S.B. Choudhury 1 3 4 1 Ocean Colour Applications and Measurement Division, Earth and Climate Sciences, National 5 Remote Sensing Centre, Dept. of Space, Hyderabad – 500037, India. 6 2 Chilika Development Authority, Dept. of Forest and Environment, Govt. of Odisha – 751014, 7 India. 8 9 Corresponding Author: 10 Santonu Goswami 11 Ocean Colour Applications and Measurement Division, Earth and Climate Sciences, National 12 Remote Sensing Centre, Dept. of Space, Hyderabad – 500037, India. 13 Email address: santonu.isro@gmail.com 14 15 Abstract 16 Changing trend in coastal ecosystem can be quantified by performing time series analysis. Time 17 series analysis performed using long term remote sensing data will help us to identify the 18 dynamic changes happening in the coastal ecosystem and its surrounding regions. In the present 19 study we performed time series analysis on northern sector of Chilika Lake and its nearby 20 regions of Odisha, which is situated in the east coast of India using three decades of freely 21 available Landsat archive data. In order to detect dynamic changes trend parameters were 22 calculated by using available data sets from Landsat Thematic Mapper (TM) and Operational 23 Land Imager and Thermal Infrared Sensor (OLI/TIRS) for the observation period from 1988 to 24 2017. Two multi-spectral indices i.e. NDVI and EVI were generated from the available data sets 25 and the trend analyses were performed using Theil-Sen (T-S) regression method by identifying 26 robust trend parameters (slope and pvalue). The average mean values for NDVI and EVI were 27 recorded at 0.4 and 0.2. Significant positive trend was observed in both vegetation indices 28 (NDVI and EVI) with a mean slope value of 0.004 and 0.003 and pvalue of 0.03 and 0.02 29 respectively. 30 Introduction 31


54
NDVI is often used due to its 'ratio' properties which enable to eliminate the noise such as 55 (topography, sun angle and clouds) present in the data (Matsushita B. et al., 2007). EVI is 56 considered as a modified NDVI due to its improved sensitivity towards higher biomass region 57 and reduction in atmosphere influences (Huete A.R. and Justice C., 1999). Typical NDVI and 58 EVI values ranges between -1 to +1, negative values represents snow and water while the 59 positive values for NDVI represents vegetated areas and soil, with classified ranges for; sparse 60 vegetation from 0.2 to 0.5 which is considered as moderate vegetation, from 0.6 and above are 61 considered as dense vegetation while in EVI healthy vegetation falls between 0.2 to 0.8. 62 The time series analysis technique with temporal data can be used to perform seasonal and 63 annual variations. This type of analysis provides a benefit in various fields such as land use/land 64 cover change, agriculture, forest management, ecosystem changes and water management 65 (Muttitanon W., 2004). To perform such analysis in remote sensing based studies, we require 66 high spatial and temporal resolution satellite data such as Landsat (TM, ETM+ and OLI) at 30m 67 spatial resolution and provides 16 days temporal resolution, which is provided by USGS at free 68 of cost and it can be quite useful in performing time series analysis to identify seasonal and 69 annual trends over decades. However Landsat is as an optical sensor and presence of cloud can 70 profoundly affect the quality of data and time series analysis performed using this data. For 71 detecting the seasonal response of vegetation, both NDVI and EVI time series are considered as 72 efficient data products (Wardlow et al., 2007), while for the short term interannual variability;   Table 1.
Where L λ is the spectral radiance at the sensors aperture, d is the Earth-Sun distance, ESUN λ is 135 the mean solar exo-atmospheric irradiances, and θ s is the solar zenith angle in degrees.

136
The TOA reflectance was converted to surface reflectance with Dark Object Subtraction (DOS1) 137 correction technique. DOS is an image based atmospheric correction method which rectifies the 138 image containing pixels which are completely covered by the shadow and radiance received by 139 the satellite due to atmospheric scattering. The path radiance was calculated by the 'Equation
Where L min is the radiance that corresponds to a digital count value, L DO1% is the radiance of 143 dark object. Finally the surface reflectance was calculated by using the 'Equation (4)'. 144 Where L p is the path radiance, T v is the atmospheric transmittance in the viewing direction, T z is 146 the atmospheric transmittance in the illumination direction and E down is the down welling diffuse  Landsat TM and OLI/TIRS data are having different spatial extent, to achieve common spatial 159 extent we created an area of interest (AOI) boundary for study region and all the data sets were 160 subset to that boundary. Using the subset data we calculated vegetation indices NDVI and EVI.  In final process all datasets were stacked to make a one temporal image and trend were 179 calculated for each multi-spectral index in the temporal domain by using robust Theil-Sen (T-S) 180 regression method (Sen, 1968;Theil, 1992). The T-S regression method was applied on these 181 spectral indices to calculate trend parameters such as slope and pvalue. T-S calculation was carried out in R Studio v 1.0.153 software followed by using 'trend' package to get the trend 183 statistics.

185
Vegetation/greening 186 The Chilika region exhibits a moderate vegetation trend which can be seen in both the vegetation 187 indices NDVI and EVI (Fig. 4). Annual mean NDVI and EVI value ranged between 0.2 and 0.87     Trend slope maps were generated for both multi-spectral indices which are shown in (Fig. 5). 206 Graph show in (Fig. 6) depicts temporal trend of annual mean NDVI and EVI composites and we   Chilika lagoon has change significantly in the past four and half decades due to constant pressure 219 of natural and anthropogenic activities. These changes can be seen visually on the images shown 220 in (Fig. 7). The lagoon is highly infected with Fragmites Karka which is deposited on northern