Classifying Mangrove Forests and Aquaculture Ponds using Multi-Layer Perceptron Neural Networks in Southeast Asia
Organization: Clark Labs/IDRISI
Position: Remote Sensing Research Assistant
Project: Mangrove Forest and Pond Aquaculture Classification funded by The Gordon and Betty Moore Foundation
Date: July 2018
Background: Clark Labs, in partnership with the Gordon and Betty Moore Foundation and in support of the Foundation’s Oceans and Seafood Markets Initiative, is mapping an inventory of pond aquaculture and coastal habitats in Bangladesh, Cambodia, Ecuador, India, Indonesia, Malaysia, Myanmar, Thailand, and Vietnam. Landcover maps are in development for the years 1999, 2014, and 2018 for these countries.
Data: Landsat 4-5 Imagery from USGS Earth Explorer
Methods: Using USGS Earth Explorer, Tier-1 Landsat scenes were obtained and downloaded. Three dates of imagery are being classified for Indonesia. The dates include 1999, 2014, and 2018. The required classes are mangrove, coastal wetland, pond, water, other land cover, and missing (clouds or transmission error). The protocol for classification of all classes is the same across dates, but the data and date ranges differ for 1999 from 2014 and 2018. The classification uses several methods: Isoclust, Mahalanobis, and MLP. This summary will show the 1999 classification of an area in Indonesia that required two different images, as the area of dense mangrove forest was mostly covered with clouds.
Outcome: The 1999 land cover map is based on the Landsat 5 imagery from 1997-1999 shown below. Additionally, the database includes a mapping of present-day vulnerability for further conversion to pond aquaculture and a mapping of the risk of landcover conversion to pond aquaculture by 2050.
View the current published version of the Southeast Asia project Here.
Position: Remote Sensing Research Assistant
Project: Mangrove Forest and Pond Aquaculture Classification funded by The Gordon and Betty Moore Foundation
Date: July 2018
Background: Clark Labs, in partnership with the Gordon and Betty Moore Foundation and in support of the Foundation’s Oceans and Seafood Markets Initiative, is mapping an inventory of pond aquaculture and coastal habitats in Bangladesh, Cambodia, Ecuador, India, Indonesia, Malaysia, Myanmar, Thailand, and Vietnam. Landcover maps are in development for the years 1999, 2014, and 2018 for these countries.
Data: Landsat 4-5 Imagery from USGS Earth Explorer
Methods: Using USGS Earth Explorer, Tier-1 Landsat scenes were obtained and downloaded. Three dates of imagery are being classified for Indonesia. The dates include 1999, 2014, and 2018. The required classes are mangrove, coastal wetland, pond, water, other land cover, and missing (clouds or transmission error). The protocol for classification of all classes is the same across dates, but the data and date ranges differ for 1999 from 2014 and 2018. The classification uses several methods: Isoclust, Mahalanobis, and MLP. This summary will show the 1999 classification of an area in Indonesia that required two different images, as the area of dense mangrove forest was mostly covered with clouds.
Outcome: The 1999 land cover map is based on the Landsat 5 imagery from 1997-1999 shown below. Additionally, the database includes a mapping of present-day vulnerability for further conversion to pond aquaculture and a mapping of the risk of landcover conversion to pond aquaculture by 2050.
View the current published version of the Southeast Asia project Here.