Enhanced Drought Analysis in Bangladesh: A Machine Learning Approach for Severity Classification Using Satellite Data

Tonmoy Paul, Mrittika Devi Mati, Md. Mahmudul Islam · 2025

Read Enhanced Drought Analysis in Bangladesh: A Machine Learning Approach for Severity Classification Using Satellite Data by Tonmoy Paul, Mrittika Devi Mati, Md. Mahmudul Islam (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2506-04696.txt.

Introduction

Tonmoy Paul, Mrittika Devi Mati, and Mahmudul Islam wrote a paper from 2025 on grading drought severity across Bangladesh with machine learning. Drought hits the country's farms, food supply, and economy hard. Classic drought indexes lean on rain alone and miss soil moisture and heat. Past machine learning work had not woven satellite data across many districts. The team set out to close that gap.

They built a data-driven frame covering thirty-eight districts and the years 2012 to 2024. Weather feeds gave humidity, soil moisture, and temperature for each district. Unsupervised methods first grouped districts by likeness. Then four classifier models sorted each district into severity levels. The models sorted the levels well. The map that emerged varied sharply by region, with some districts far more exposed than others. The team argues such maps can guide where aid and water planning should go first.

Catalog record

CREATORTonmoy Paul, Mrittika Devi Mati, Md. Mahmudul Islam
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE21 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2506-04696

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Enhanced Drought Analysis in Bangladesh: A Machine Learning Approach for Severity Classification Using Satellite Data Tonmoy Paul, Mrittika Devi Mati, Md. Mahmudul Islam Abstract Drought poses a pervasive environmental challenge in Bangladesh, impacting agriculture, socio-economic stability, and food security due to its unique geographic and anthropogenic vulnerabilities. Traditional drought indices, such as the Standardized Precipitation Index (SPI) and Palmer Drought Severity Index (PDSI), often overlook crucial factors like soil moisture and temperature, limiting their resolution. Moreover, current machine learning models applied to drought prediction have been underexplored in the context of Bangladesh, lacking a comprehensive integration of satellite data across multiple districts. To address these gaps, we propose a satellite data-driven machine learning framework to classify drought across 38 districts of Bangladesh. Using unsupervised algorithms like K-means and Bayesian Gaussian Mixture for clustering, followed by classification models such as KNN, Random Forest, Decision Tree, and Naive Bayes, the framework integrates weather data (humidity, soil moisture, temperature) from 2012-2024. This approach successfully classifies drought severity into different levels. However, it shows significant variabilities in drought vulnerabilities across regions which highlights the aptitude of machine learning models in terms of identifying and predicting drought conditions. Disclaimer: This is the preprint version of a paper accepted at the 2024 27th International Conference on Computer and Information Technology (ICCIT), organized by IEEE Bangladesh Section. The final version will be published in IEEE Xplore. I Introduction Drought is an inherently catastrophic event induced by the climatological factors, landscape, climate, topography, and water demand of a specific region. Drought is more frequent in Bangladesh due to high temperatures and low rainfall. Bangladesh is known as one of the largest deltas in the world, extremely vulnerable to Natural Disasters due to its Strategic location, Low-lying, Flat terrain, Population density, Poverty, Lack of Institutional setup, etc [ 1 ] . When precipitation levels are consistently below average, it disrupts the natural water cycle, leading to diminished groundwater recharge. Soil moisture is a critical factor in mitigating drought, as it stores water that plants rely on during dry periods. Reduced soil moisture due to lack of precipitation or high temperatures can lead to agricultural drought [ 2 ] . Machine Learning(ML) is nowadays a very popular method for analyzing complex datasets and solving problems. Drought analysis itself, a complex task, needs a huge amount of time and effort to interpret its characteristics. Author in [ 3 ] showed how ML can be treated to forecast drought. Unsupervised ML models have been introduced in drought analysis several times as shown by the authors of [ 4 ] [ 5 ] . However, this study tries to inaugurate an innovative approach to drought severity classification, utilizing the machine learning technique to analyze satellite-driven weather dataset. This framework combines unsupervised clustering algorithms with supervised classification techniques to accurately predict drought severity. We have organized this paper as follows: Section II reviews previous works and their limitations. Our proposed methodologies are explained in Section III , and the machine learning algorithms we used are detailed in Section IV . The analysis and discussion of the results are presented in Section V . Finally, Section VI provides the overall conclusion of this paper. II Literature Review The literature on drought analysis from the perspective of Bangladesh is comparatively infrequent. Indeed, most of the studies traditionally focus on flood risk assessments because of this country’s historical inclination to flooding. Recent studies and research projects have focused on drought forecasting, particularly on drought indices employing machine learning. The concepts, characteristics, complex nature of drought and the various environmental factors that influence drought; drought indicators are also identified and predicted by implementing Prediction Models and Adopted Technologies [ 6 ] . Several indices of drought such that; Standardized Precipitation Index (SPI) [ 7 ] ,Palmer Drought Severity Index(PDSI), Standardized Precipitation-Evapotranspiration Index (SPEI) [ 8 ] and Vegetation Health Index (VHI) [ 9 ] . TABLE I: Advantages and Limitations to previously used drought index Index Focus Calculation Advantages Limitations SPI Precipitation anomalies Measures deviation from long-term averages Simple to calculate, widely used Does not consider other factors like temperature PDSI Moisture balance Considers precipitation, temperature, potential evapotranspiration, and soil moisture Accounts for multiple factors, provides comprehensive assessment Requires more data and complex calculations SPEI Moisture balance and temperature Combines precipitation and potential evapotranspiration More sensitive to drought in regions with high temperatures Requires more data and complex calculations VHI Vegetation response to drought Uses remote sensing data to assess vegetation health Provides a direct measure of drought’s impact Can be influenced by factors other than drought The table I presents four commonly used drought indices: SPI, PDSI, SPEI, and VHI. Each index focuses on a different aspect of drought, brief overview of the calculation methods involved, advantages and limitations. For example, SPI is simple to calculate but does not consider temperature, while PDSI provides a comprehensive assessment but requires more data. The preference of index depends on the earmarked research question and the available data. Moreover, types of drought Meteorological Drought [ 10 ] , Hydrological Drought [ 11 ] are also assumed by implementing ML. This [ 3 ] study shows that these sorts of studies have been frequently conducted based on certain locations. Clustering algorithms, such as K-Means and Gaussian Mixture Models have been introduced in environmental studies around the world. In the example, Authors in [ 12 ] presented the application of the K-Means clustering model for Drought analysis and authors of [ 13 ] explored the Gaussian Mixture Models for environmental analysis. These studies highlight the efficacy of clustering techniques in environmental analysis, yet their application to analyze drought conditions in Bangladesh remains underexplored. This study tries to address the limitations of previous approaches, particularly their reliance on limited datasets and lack of incorporated critical environmental parameters such as temperature, soil moisture, and humidity, which are vital for a more comprehensive understanding of drought conditions. III Methodologies This research aims to represent an in-depth analysis of drought conditions across BD, emerging with advanced Machine Learning algorithms to cluster drought characteristics, identify and predict drought conditions. This study leverages a satellite-extracted dataset from [ 14 ] on the time span of 2012 to 2024 (daily), composed of several environmental factors, i.e. solar radiation, humidity, temperature, soil moisture and wind-speed. By applying unsupervised machine learning algorithms like K-means clustering, Bayesian Gaussian Mixture, this study categorizes drought severity and tries to provide a framework to understand drought patterns. The objective of this study is to introduce a new classification framework which describes drought severity and predicts the drought scenarios based on satellite based data in Bangladesh. III-A Data Collection The dataset has been collected from [ 14 ] across 38 different district-wise locations of Bangladesh based on important weather parameters. The Fig. 1 shows the locations that were collected , table II describes which weather parameters has been introduced to further analysis. Fig. 1: Data collected across 38 districts of Bangladesh. TABLE II: Model parameters Parameters Acronyms Locations Latitude, Longitude Year (2012-2024) Year Day of Year DOY All Sky Surface Shortwave Downward Irradiance (MJ/m 2 )/day) ALLSKY_SFC_SW_DWN Temperature at 2 Meters (°C) T2M Dew/Frost Point at 2 Meters (°C) T2MDEW Earth Skin Temperature (°C) TS Specific

Read the complete text