Multisensor estimates of land loss and fragmentation in Deltas using Google Earth Engine
Abstract
Almost one-half billion people live on and near deltas worldwide. These populations are increasingly under threat from changes to riverine sediment load, increased storm surge, and sea-level rise. Over 85% of deltas worldwide have experienced severe flooding in the past decade. These risks are particularly true in the Mississippi River Delta, Louisiana, USA, where coastal land loss is occurring at an alarming rate and has already displaced people from their homes. An essential part of combating land loss on the Louisiana coast is making accurate measurements of land loss. Up to now, land loss estimates primarily come from Landsat satellite data. However, at a spatial resolution of 30-m per pixel and a temporal resolution of almost 16 days, land loss across small, inland areas that cause delta fragmentation can be missed, making accurate short-term estimates difficult. Given that climate change will increase the frequency of extreme events along the coast, an existing challenge is to improve land-loss estimates at finer spatial and temporal scales to better protect vulnerable populations living on and near deltas.
Given the dynamism of deltaic land change, we tackle the need for progressively finer spatial and temporal resolution from Landsat, Sentinel-2, and near-daily PlanetScope imagery. Earth Engine allows for the creation of image composites from these three datasets at monthly and yearly intervals. Intertidal and submerged land in deltas requires the use of sub-pixel classification where pure endmembers are harder to resolve. Our approach contrasts (1) random forest pixel-level classification against (2) constrained spectral unmixing for sub-pixel classification. Earth Engine allows us to compare land loss and fragmentation across our three sensors, which expands on previous research that refined overall land loss estimates and where fragmentation (e.g., land-to-water conversion) occurs. This approach allows us to refine land loss estimates by filling in spatio-temporal and spectral windows and better account for inland fragmentation that can accelerate overall land loss in deltas.- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2018
- Bibcode:
- 2018AGUFMEP53C..33R
- Keywords:
-
- 0480 Remote sensing;
- BIOGEOSCIENCESDE: 0758 Remote sensing;
- CRYOSPHEREDE: 1640 Remote sensing;
- GLOBAL CHANGEDE: 1855 Remote sensing;
- HYDROLOGY