Coastal mangroves have been lost to deforestation for anthropogenic activities such as agriculture over the past two decades. The genesis of methane (CH4), a significant greenhouse gas (GHG) with a high potential for global warming, occurs through these mangrove beds. The mangrove forests in the Indian Sundarban deltaic region were studied for pre-monsoonal and post-monsoonal variations of CH4 emission. Considering the importance of CH4 emission, a process-based spatiotemporal (PBS) and an analytical neural network (ANN) model were proposed and used to estimate the amount of CH4 emission from different land use land cover classes (LULC) of mangroves. The field work was performed in 2020, and gas samples of various LULC were directly collected from the mangrove bed using the enclosed box chamber method. Historical climatic data (1960-1989) were used to predict future climate scenarios and associated CH4 emissions. The analysis and estimation activities were carried out utilizing satellite images from the pre-monsoonal and post-monsoonal seasons of the same year. The study revealed that pre-monsoonal CH4 emission was higher in the south-west and northern parts of the deforested mangrove of the Indian Sundarban. A sensitivity study of the anticipated models was conducted using a variety of environmental input parameters and related main field observations. The measured precision area under curve of receiver operating characteristics was 0.753 for PBS and 0.718 for ANN models, respectively. The temperature factor (Tf) was the most crucial variable for CH4 emissions. Based on the PBS model with coupled model intercomparison project-6 temperature data, a global circulation model was run to predict increasing CH4 emissions up to 2100. The model revealed that the agricultural lands were the prime emitters of CH4 in the Sundarban mangrove ecosystem.
Keywords: Deforestation; Greenhouse gases; LULC; Mangroves; Neural network.
Copyright © 2023 Elsevier B.V. All rights reserved.