FEASIBILITY TO DETECT SIGNS OF POTENTIAL CO2 LEAKAGE WITH MULTI- TEMPORAL SPOT SATELLITE VEGETATION IMAGERY IN OTWAY, VICTORIA

Abstract. This paper presents image processing results for the OtwayCO2storage site, a demonstration project of CO 2 sequestration in south-western Victoria, Australia. These results were derived from SPOT-VGT S10 datasets of 2001 to mid 2011. Over 65,000 tonnes of CO 2 -rich gas stream was injected into a depleted gas reservoir at a depth of 2050 meters at the site since 2008. Over time, CO 2 migration up-dip within the 31 m thick reservoir sandstone capped by the impervious thick seal rock has been recorded. But no top soil contamination has been discovered. This study has analysed the site vegetation growth using NDVI as a measure on a pixel by pixel basis. The multi-year time series result shows that NDVI values at the site regularly vary according to the seasons. Furthermore, precipitation levels were fluctuating in the past 10 years, especially in the years of 2002 and 2006, which correlated with low NDVI measuring results. But there are detected hot spots that cannot be linked with rainfall. Authors have found that some hot spots correspond with site well drilling and pipelines construction periods and locations. While others might be due to image data biased. Therefore, certain low NDVI spikes in the temporal evolution results cannot be attributed to only drought or pasture grazing. These subtle changes detected in the NDVI index prove the ability to use satellite image for providing valuable information to decision makers in relation to CO2 sequestration site environmental safety monitoring for searching CO 2 leakage signals.


INTRODUCTION
Using underground reservoirs to store CO 2 has been tested globally as an effective technique to reduce greenhouse gas emissions.But in the meantime, the public have raised concerns about the stability of geologic formations to hold large amounts of carbon dioxide for the long term.People worry that CO 2 could migrate or leak into the top soil through geological faults or cracks, fractures of rocks, or an abandoned well connected with the reservoir.The leaked CO 2 gas can displace oxygen from the roots of plants at the surface and thus cause the plants to die.This phenomenon has been discovered from natural CO 2 gas seepage caused by a volcano eruption (USGS, 2000).Also the leaked CO 2 could change the pH and redox of soil and alter natural microbial environments (Noomen et al., 2009).Consequently, over time, vegetation in CO 2 contaminated soil can display characteristics such as stunted growth, reduced water content, or decreased leaf chlorophyll concentrations.Asides from CO 2 gas killing vegetation, it also can lead to asphyxiation in humans and animals at higher concentrations.These problems have driven public concern over the environmental safety of CO 2 sequestration sites.In response, scientists are motivated to develop effective methodologies for confirming the stability and suitability of selected geological formations to store CO 2 and to monitor for any signs of CO 2 leakage.In this paper, the authors have proposed to apply the SPOT-VGT S10 images to indirectly detect the presence of CO 2 by tracking vegetation stress changes through Normalized Difference Vegetation Index (NDVI) analysis in OtwayCO2CRC site.However, the vegetation growth signature value of NDVI from optical image can be impacted by differing factors such as the seasons, weather, precipitation, and human activity.Therefore, the temporal evolution results of NDVI need to be assisted by a trend value derived from regression analysis with multiple years' statistical model.

Otway CO2CRC site
The CO2CRC Otway storage Project (Figure 1) is the first advanced geosequestration project in Australia and the world's largest research and demonstration project.Work at the Otway site commenced in as early as 2002 (Figure 2).It can be seen that vegetation has been cleared at the construction site and along the route for the pipeline.Since April 2008 about 100,000 tonnes of CO 2 have been injected and stored in a depleted gas reservoir 2km deep underground and further injections into different geological formations are being planned.Because NDVI is strongly influenced by precipitation, temperature, and also grazing and farming activities, multi-year NDVI time series have to be used so that each contributor can be identified accurately.
The VGT-sensor registers 1728 pixels in line, with a sub-nadir resolution of 1.15 km and a swath with of about 2200 km.Furthermore, the VGT-sensor simultaneously registers in 4 short wave bands: BLUE, RED, NIR and SWIR.Table 1 gives the specifications of SPOT-VGT.The raw VGT-S10 data are in Plate carrée projection with a 0.0089285714-degree pixel resolution.These multi-temporal data were transformed and resampled using a nearest neighbour operator into the UTM projection based on WGS84 spheroid at 1km resolution by using in house automated GEOS-VGTS10-NDVI software.

METHODOLOGY
Using NDVI index to determine vegetation condition is based on the fact that healthy vegetation has a low reflectance in RED (visible) electromagnetic spectrum due to plants' chlorophyll absorption, and high reflectance in NIR (near infrared) because of internal reflectance by mesophyll of the green leaves.We calculate the NDVI index using SPOT-VGT S10 image pixel's reflectance values of RED and NIR by equation ( 1): Where, NIR is near infrared reflectance in band 3 RED is visible light reflectance in band 2 The value of NDVI is limited between +1 and -1.Normally, the NDVI of vegetation will vary between 0 to +1.The value closing to +1 (e.g.0.8 -0.9) represents the highest possible density of green leaves.And the value would be close to zero if there are no green leaves.A zero therefore means no vegetation.
Besides, the healthy vegetation absorbs most of the visible light and reflects a massive portion of the near-infrared light.And unhealthy vegetation will reflect more in visible light and less in near-infrared.

Image processing flowchart with GEOS-VGTS10-NDVI
The commercial-off-the-shelf remote sensing software packages such as ENVI calculate NDVI on a patch by patch basis for selected interested area.The output result of this area is the average NDVI of multiple pixels.Hence any change in less than 1km 2 scale because CO 2 leakage will be smoothed out.Also the result can only be done manually one image at a time.It will take too long to process hundreds of satellite images.In order to derive NDVI time series for areas around the CCS sites on a pixel by pixel basis in a reasonable range, an automated software tool GEOS-VGTS10-NDVI has been developed NDVI time series of 25 pixels or more for areas around the Otway CCS site according to the each pixels location have been extracted from the 378 SPOT images.Figure 3 is the flowchart for processing the VGT-S10 images.

Vegetation cover localized NDVI signature analysis using GEOS-VGTS10-NDVI
In order to track possible CO 2 leakage around the CO 2 injection and storage site, a patch of 5 by 5 pixels has been selected in the SPOT image covering 5km by 5km centred at the CO 2 injection well (Figure 4).The sub image thus contains a total of 25 pixels with 1km resolution.In Figure 5 each pixel has been numbered.The yellow boundary represents the extent of CO2CRC activity areas for drilling and construction.The yellow pin is the location of injection well which is right on the boundary between pixels 13 and 18.

RESULT AND DISCUSSION
A total of 378 SPOT VGT-S10 images are processed for vegatation growth analysis at this ccs site.The temporal evlution of NDVI value in the past 10.5 years for each pixels are given by figures of 5 and 6. Figure 5 gives the time series result in 2D .It can be seen a clear seasonal sin wave change of NDVI in figure 5, and also with many sharp spikes indecating large amount of NDVI drop.

Figure 4 .
Figure 4.The 25 pixels around the Otway CCS site (selected based on the coordinates of the injection well; latitude: -38.51786, longitude: 142.82143)

Figure
Figure 7-B.Seasonal variations removed 3D NDVI time series.

Figure 9 .
Figure 9. Annual rainfall in Victoria in the last 100 years (BOM, 2011)