Tarawa and Rising Sea Levels: Analyzing Land Change via Remote Sensing

For my bachelor's thesis during the fall of 2014 at Rowan University, I conducted independent research on land surface area and coastline changes over a 15 year period for the island of Tarawa (Republic of Kiribati) using applied remote sensing analysis techniques with Idrisi Selva, Landsat 7 and 8 satellite imagery, and locally provided tidal data. 

This brief summary includes slides and raster images used during my presentation to help explain the objectives, data, methods, and results of my research.

Tarawa is a remote central Pacific island apart of the Gilbert Islands archipelago and capital of the Republic of Kiribati. The island is approximately 190 sq mi. and is home to a population of about 60,000 people. Kiribati, along with many other island nations worldwide, is seen as especially vulnerable to climate change and rising sea levels among geographers and climate scientists. The island was the subject of similar research to this one, though only studied in less detail while being part of a larger general study area.
 


A web map showing the Republic of Kiribati. The country can be roughly divided into three main archipelagos and one isolated island. Together, the Gilbert, Phoenix, and Line islands along with isolated Banaba Island make up the 33 islands that comprise the nation of Kiribati.

 

A web map showing how and where the population of Kiribati is dispersed. Half of Kiribati's 103,000 population live on the island of Tarawa, also its capital. Tarawa was chosen as the primary island subject for this study not only for its large population (compared to the rest of Kiribati's islands) but due to its vulnerability to rising sea levels.

 

Objectives

The main objectives of my research were to identify areas of land change on Tarawa, analyze the change (gain, loss, or neutral), and quantify the change. Prior research by other academics suggested that, like most other low lying islands in the Pacific, Tarawa too was slowly shrinking and losing its land mass to the ocean. This research sought to study Tarawa in greater detail with more recent data.

Data & Methods

NASA & USGS's Landsat 7 and 8 imagery (30m² resolution, data hosted here) was primarily used in sourcing the oldest and most recent images of Tarawa. The sporadic and large cloud cover in many of the images however posed a challenge in analysis as it often does in remote sensing, specifically in the classification of land vs water surfaces. However, dense cloud coverage not directly over the island was easily 'vectored out'. After producing NDVI and tasseled cap vegetation, brightness, and moisture index images, I narrowed down the best images to either the NDVI or the T-Cap's vegetation index. I eventually chose an NDVI image (with values multiplied by 100 to allow for more precise classification) due to its higher accuracy in separately classifying land and water for this location. The NDVI process was also more successful in capturing inter tidal zones where the tasseled cap process could not easily do so. Accurately identifying and classifying these inter tidal zones was a critical step in this research. Comparing an image from 2000 to 2014 and determining a significant change in land coverage had occurred could be incorrect if tides were not taken into account.

True color composite image of Tarawa, Kiribati - January 27, 2000.

NDVI

True color composite image of Tarawa, Kiribati - November 9, 2014.

Tasseled cap greenness index

In a simpler location, land change analysis wouldn't involve map algebra more complex than overlaying the two images and subtracting the earlier image from the later one, but the low-lying islands of Kiribati are subject to significant daily tides which adds a challenge. Most of Tarawa sits under 3 meters in elevation with the high tide typically coming in with a maximum height above local mean sea level at 1 - 1.25 meters. The significant tidal action here would have to be accounted for in calculating the final land coverage change. Fortunately, the NOAA does maintain accessible tide data for Tarawa (and many other places) and I was able to make use of it for my research. 

Tarawa in green, as seen in this NDVI derived image. At first glance it appears the island has grown over a 15 year period. However the fact that the images were taken at different tide intervals can give misleading conclusions. There is a 0.9 meter difference in tide levels between the two images.

After producing accurate 2000 and 2014 NDVI x100 images with reduced cloud coverage, compensating for tides, classifying land & water coverage, and overlaying the two images for a map algebra process, the final result showed which parts of the island had seen changes. The areas in red shown below are areas where Tarawa has shrunk. In other words, places that are now permanently inundated. Whereas 15 years ago these places would have only been under water intermittently- during high tide. This amounts to a 3.2% decrease in total land coverage for Tarawa over a 15 year period, results that are consistent with previous studies of similar island atolls in the Pacific.

The results of this study can even be seen with Digital Globe commercial imagery hosted on Google Earth. The slides below show key areas on Tarawa affected by the coastline change, and accompanying decrease in total land coverage.

Landsat 7 true color images from 2003 & 2006 of Tarawa's coastline.

Landsat 8 true color images of same locations in 2013 (identical scales and viewing widths). Some of the most significant changes in the coastline can be seen here.