Jump to:

Wetland inundation dynamics

Monitoring surface inundation in wetlands is important for understanding the ecosystem services they offer, including the regulation of regional hydrology, water quality and greenhouse gas cycling. Existing methods for detecting wetland inundation using satellite imagery are often inadequate due to the complexity of wetlands in both space and time. Funded by NASA's Land Cover and Land Use Change (LCLUC) programme, my research involves the use of data from a diverse set of Earth Observation satellites, including NASA's Landsat and the European Space Agency's Sentinel-1 and Sentinel-2 satellites. By combining optical (Landsat and Sentinel-2) with radar (Sentinel-1) data, we can track inundation dynamics over time at much higher temporal frequency than is possible with only one sensor type. Below is a demonstration of how these data sources can be used together to characterize temporal inundation profiles in the Florida Everglades.

inundation time series over a part of the Everglades

Above: Time series of surface inundation over a part of the Florida Everglades, using combined optical and radar data (Landsat, Sentinel-1 and Sentinel-2). The bottom panel shows the average water fraction for each image in the time series.

Read more about this project at here.

Back to top

Flood disaster monitoring

In my current research, I am also interested in the use of radar data to track flood disasters in near real-time. Imagery from radar satellites are not affected by cloud cover, and are therefore instrumental in high-frequency monitoring of flood events, during which cloud cover is often problematic for optical sensors. I am developing methods to use Sentinel-1 radar data on the Google Earth Engine to map potentially catastrophic floods in near real-time. Below is an example of how this novel platform can facilitate disaster response through the rapid processing of image time series data

screenshot of a Google Earth Engine flood monitoring app

Above: Screenshot of a Google Earth Engine app using Sentinel-1 radar data to map "new" floods (red) and historical Landsat data to map previously inundated areas or permanent water (blue) around Houston, Texas following Hurricane Harvey on 2017-08-29. A time series plot of radar data for a single pixel is shown in the left-hand panel.

Back to top

Forest change in the tropics

My PhD dissertation describes approaches to monitoring change dynamics in tropical forests using dense Landsat time series and ground observations. In my research, I explored ways in which these data can be used to monitor deforestation, degradation and regrowth. Below is an example of the bfastmonitor method over a site in southern Ethiopia. In this study, I showed that by using all Landsat observations, small-scale deforestation could be tracked fairly well. Without reliable ground-based observations, however, degradation was nearly impossible to track. Working with local communities, I showed in another study that high-detail characterization of deforestation and degradation in these complex forest systems is indeed possible with dense Landsat time series when used together with observations from the ground.

forest change in southern Ethiopia

Above: Demonstration of the bfastmonitor method over a site in southern Ethiopia, where in-migration caused successive conversion of forest to cropland between 2005 and 2012.

In addition to my work in southern Ethiopia, I published another of my PhD chapters describing a method for the automated detection of post-disturbance regrowth using Landsat time series. I demonstrated the method in Madre de Dios, southern Peru, an area characterized by rapid forest change in recent years due to conversion to pasture and small-scale gold mining operations.

Back to top

Change monitoring resources

Alongside my research activities, I have produced or collaborated on a number of open-source packages. You can read more about them here.

Back to top