Page 37 - IPCoSME 2021

1 INTERNATIONAL POSTGRADUATE CONFERENCE ON SCIENCE AND MARINE ENVIRONMENT 2021
st
(IPCoSME 2021)
“Environmental Sustainability Enhancement Through the Collaboration of Sciences”

RD-05

CHANGE DETECTION OF MANGROVE FOREST IN TRANG PROVINCE,
THAILAND USING REMOTE SENSING AND MACHINE LEARNING

1*
2
BUTCHANOK KONGKET , PONLACHART CHOTIKARN AND SUTHINEE
SINUTOK 3

1 Faculty of Environmental Management, Prince of Songkla University, 90110 Songkhla,
Thailand
2 Coastal Oceanography and Climate Change Research Center, Prince of Songkla University,
90110 Songkhla, Thailand
3 Marine and Coastal Resources Institute, Faculty of Environmental Management, Prince of
Songkla University, 90110 Songkhla, Thailand

*Corresponding author email: Butchanokk1998@gmail.com

Abstract: Mangrove forests have significant environmental and ecological values that fulfil

many important functions, such as providing habitat for organisms and mitigating climate
change. Machine learning coupled with remotely-sensed images has been applied to mangrove

mapping and monitoring. In this study, we use the medium-resolution satellite images from

Landsat-5, 7 and 8 and random forest classifier (RF) to perform change detection on mangrove
forests in Sikao, Kantang, Hadd Samran and Palian districts, Trang Province. Mangrove

coverage and mangrove change maps were generated to identify both mangrove distribution
and temporal variation in 1990, 2001, 2016 and 2019. We found that approximately 4% of

mangrove area in 2019 was lost from 1990 (1,258 ha). For the classification methods, the RF
classifier achieved overall accuracy greater than 97%. This study presented a suitable approach

for mangrove mapping and change detection analysis. The maps provide an up to date

information for the study areas and can be used for future comparative studies.

Keywords: landsat, mangrove classification, pixel-based classification, random forest

37