Understanding drivers of changing fire activities: from the perspective of causality

Understanding the drivers of environmental changes is critical for predicting and managing this rapidly evolving world. One of the challenges commonly faced by existing studies of analyzing the drivers of an environmental change, be it at a small scale or a large one, is that the drivers presented a...

Full description

Saved in:
Bibliographic Details
Main Author: Wu, Sifeng
Other Authors: Chew Lock Yue
Format: Thesis-Doctor of Philosophy
Language:English
Published: Nanyang Technological University 2023
Subjects:
Online Access:https://hdl.handle.net/10356/166585
Tags: Add Tag
No Tags, Be the first to tag this record!
Institution: Nanyang Technological University
Language: English
Description
Summary:Understanding the drivers of environmental changes is critical for predicting and managing this rapidly evolving world. One of the challenges commonly faced by existing studies of analyzing the drivers of an environmental change, be it at a small scale or a large one, is that the drivers presented are not necessarily the true causes that have causal relations with the target of their study. In this thesis, I used state-of-the-art techniques from causal inference to understand the drivers of changing fire activities at different scales. Fire is a fundamental process of the earth system, and the widely observed changing patterns of fire behavior is an important aspect of the global changes that can have profound impacts on ecosystems and human societies. Therefore, understanding the drivers of the changing fire behavior is of vital importance. I first propose a causal framework that integrate models that can detect causality among variables, and interpretable machine learning models to select and quantify the impact of drivers on fire emissions. I tested this framework on 12 selected regions, and the results showed that the framework can effectively select the true causes as drivers of fire emissions, and also provide informative evaluations of their impacts on fire emissions. Then, I apply this causal framework to a global analysis for drivers of fire emission trend at a scale specific to biome and geological continents. Global fire-derived carbon emissions (fire emissions hereafter) are relatively stable over the period 2001-2019. This was mainly caused by the decrease in African savannas and the increase in Asian boreal forests. The main drivers for the decrease of fire emissions in African savannas are decreased vegetation caused by anthropogenic intervention, mainly through agricultural expansion. The increasing fire emissions in Asian boreal forests was mainly driven by agricultural activities and changing climates, especially drier climates in this region. For the other parts of the world, their drivers differ. In general, vegetation is the most widely observed driver which usually has a positive impact; climates is also widely observed as a fire emission driver, while the impact of the several aspects of climate, namely temperature, humidity, water availability and wind, differ among areas; and anthropogenic interventions were relatively less important because it was identified as a driver for fewer locations. I also apply scenario analysis to a theoretical grass-savanna-tree model to under- stand the role of climate change and anthropogenic interventions on the process of forest degradation where fire plays a critical role. I found that for scenarios with high level of climate change, the system displays bistability and hysteresis; while for scenarios with low level of climate change, the system responds to in- creasing anthropogenic interventions nonlinearly but does not display bistability. A tree dominated system will degrade into a grass dominated one under a high level of anthropogenic intervention, and the degradation process can be accelerated and worsened by higher level of climate change. At last, I proposed an indicator to evaluate the risk of irreversible degradation for a system with hysteresis using Bayesian inference for model parameterization. I apply this indicator to a real world case and found that the risk of irreversible degradation can indeed be high. The risk of irreversible degradation should be taking into account and measured in future management decisions due to the consequences and potential loss of irreversible degradation. This thesis investigated the drivers of fire activities with emphasis on climate change and anthropogenic intervention as potential drivers. With the causal framework and its application to understanding drivers of fire-derived carbon emissions at the global scale from 2001 to 2019, climate change towards drier and warmer climates and anthropogenic intervention through deforestation are of vital importance in driving increasing carbon emissions in forest such as Asian boreal forests and Amazon forests. And the forest-savanna transition model showed the synergetic effects between the two drivers that can make the transitioning process from forest to savannas and grasslands in a shorter time and to a more degraded state. In the end, I proposed an indicator that can be adopted to measure the risk of irreversible degradation from forests to savannas, which is a promising research for my future plan.