Mapping the spatio-temporal dynamics of drought in Northeast Thailand

การวิเคราะห์และทำแผนที่พลวัตเชิงพื้นที่และเวลาของภัยแล้งในภาคตะวันออกเฉียงเหนือของประเทศไทย

Name: Phongphat Japhichom

LCSH: Burapha University -- Geoinformatics

Classification :.DDC: 526

LCSH: Cartography

LCSH: Remote sensing

LCSH: Maps

LCSH: Droughts

LCSH: Machine learning

Abstract: Drought, a globally significant natural disaster, imposes considerable economic and environmental impacts, severely impacting agriculture and socioeconomic systems annually. The frequency of global drought occurrences can be attributed to the impacts of climate change and human activities. This study aims to investigate the spatiotemporal dynamics of drought in Northeast Thailand by integrating remote sensing (RS) and ground observations with machine learning models. This study specifically focused on northeast Thailand. This area is situated within the tropical zone, characterized by mainly sandy soil that has a limited capacity to retain water. Therefore, effective water resource management and drought monitoring efforts are needed in northeast Thailand. The main contents of this thesis include: 1) Investigation of spatio-temporal drought patterns (shorth term and long term) of the study area from 2014 to 2023 using ML modeling from Landsat 8 satellite and ground observation data. 2) Performance comparison of machine learning (ML) models for monitoring drought in Northeast Thailand. 3) Mapping spatial distribution of drought events in the Northeast of Thailand from 2014 to 2023. This study leverages the fusion of RS and ground data to enhance drought monitoring. Ground indicators offer precision but have limited coverage, while RS indices cover larger areas with less accuracy. ML algorithms were used to combine these data sources, improving spatial resolution and accuracy. The study used five RS parameters such as The Vegetation Condition Index (VCI), The Enhanced Vegetation Index (EVI), The Temperature Condition Index (TCI), Topography, Precipitation, combined with ground data as The Standardized Precipitation Evapotranspiration Index (SPEI). ML techniques, including XGBoost, Random Forest, and Extra Trees, assessed the relationship between variables. Additionally, cross-validation techniques were utilized to validate the model performance. The optimal model was used to generate a spatial distribution of drought, contributing to more effective drought management strategies, and enhancing drought dynamics in the region. The results demonstrate that the Extra Trees model is outperform for accurate drought index prediction. For short-term, the results show an R² ranging from 65.26% to 94.28%, an RMSE between 1.58% to 33.28%, and an MAE ranging from 0.09% to 18.55%. Similarly, for long-term, the results show an R² ranging from 78.73% to 94.8%, an RMSE between 4.55% and 31.93%, and an MAE ranging from 0.45% to 18.14%. In particular, the variables contributing to model accuracy include precipitation (27%-67%), topography (19%-37%), and land surface temperature (6% 21%). The feature importance values of these variables enhance the model performance. The study examines both short-term and long-term precipitation patterns using the Standardized Precipitation Evapotranspiration Index (SPEI) to assess drought conditions. Short-term analysis identified significant drought occurrences in June 2015 and April 2016, with recurrent drought periods observed in late 2018 and 2019, as well as the beginning of 2020 and 2021. These findings underscore the cyclic nature of decreased precipitation and the associated risk of water scarcity within shorter time frames. Moreover, long-term precipitation trends analyzed through SPEI indicated sustained negative values from mid-2015 to 2016, indicating the onset of drought conditions. Particularly noteworthy was the persistent negativity of SPEI values from mid-2018 to 2020, indicating an extended drought period spanning multiple months. indicating the severity and duration of the drought. The main initiatives of the thesis are as follows: 1) Developed method that fuses the drought index using remote sensing (RS) data from the Landsat 8 satellite and ground observations. This provides insights into drought-related environmental parameters and precise meteorological measurements. 2) Compared the performance of three ML models to identify the mosteffective method for drought monitoring in the study area. 3) Explored spatiotemporal trends in drought distribution to inform water management and mitigation strategies. In conclusion, the study provides a framework for strategic planning in drought management by integrating RS and ground observation data. Future work could explore deep learning or neural networks to enhance drought monitoring and understanding of regional environmental implications.

Burapha University. Library

Address: CHONBURI

Email: buulibrary@buu.ac.th

Name: Shao, Zhenfeng

Role: Principal advisor

Name: Phattraporn Soytong

Role: Co-advisor

Created: 2025

Modified: 2025-12-13

Issued: 2025-12-13

วิทยานิพนธ์/Thesis

application/pdf

CallNumber: Th 526 Ph574M

eng

Spatial: Thai (Northeast)

DegreeName: Master of Science

Level: Master's Degree

Descipline: Geoinformatics

Grantor: Burapha University

©copyrights Burapha University

ลำดับที่.	ชื่อแฟ้มข้อมูล	ขนาดแฟ้มข้อมูล	จำนวนเข้าถึง	วัน-เวลาเข้าถึงล่าสุด
1	65910037.pdf	3.03 MB

ใช้เวลา

0.044366 วินาที