A Review on Classification of Time Series Hyper-Spectral Images for Automated Carbon Stock Estimation

The Earth’s climate is changing, with anthropological activities having a considerable impact on its delicate equilibrium. One of the most significant concerns in recent decades has been the steady rise in atmospheric carbon dioxide (CO2) levels. The movement of carbon in earth system that is known as the global carbon cycle plays an important role in the regulation of the Earth’s climate through stabilizing the CO2 concentration in the atmosphere. The emission of CO2 in the atmosphere leads to an increase in carbon levels, which has a negative impact on global warming by increasing global temperature and influencing other environmental factors. An accurate and current data on forest area, extent, fragmentation, and change must be provided to support regional sustainable development, including forest management, carbon emission estimation, habitat planning and biodiversity conservation. Carbon stock estimation has been studied for decades by determining forest composition and quantifying the amount of carbon stored in various components, such as trees, vegetation, soil, and organic matter. The common methods to estimate carbon stock can be time-consuming and costly, especially in collecting data for forest inventory and sampling. Understanding the composition of a forest can be achieved using hyperspectral imaging, a method that can record a wide range of spectral bands of the electromagnetic spectrum. This paper presents a review on the classification of hyperspectral images for automated carbon stocked estimation. This review includes the approaches used to estimate the carbon stock Allometric Equations and Hyperspectral Images approaches and the applications of deep learning in estimating the high stock carbon. Finally, the paper concludes by summarizing the findings obtained from this comprehensive review.