Anomaly Detection in Selected Aerosol Optical Properties and Associated Climate Variables Using a Multivariate Hidden Markov Model: A Case Study over Kenya

Abstract

Understanding aerosol climate interactions is crucial for monitoring atmospheric changes and supporting climate resilience efforts, particularly in vulnerable regions such as Kenya. This study applies a Multivariate Hidden Markov Model (HMM) to detect anomalies in key Aerosol Optical Properties (AOP) i.e., Aerosol Optical Depth (AOD), Single Scattering Albedo (SSA), and Ångström Exponent (AE) alongside associated climate variables; Surface Air Temperature (SAT) and Rainfall Rate (RR), over the period 2000-2022. Satellitebased datasets from MODIS, MERRA-2, and TRMM were used to derive monthly means, and descriptive statistics and linear regression were initially employed to characterize long-term variability. The objectives of this study were to examine the temporal and spatial variability of key aerosol and climate parameters over Kenya, detect and classify anomalies in the multivariate dataset using HMM and to interpret the climatic and environmental implications of detected anomalies and their possible causes. The HMM approach successfully identified temporal patterns and hidden states, enabling the detection of significant anomalous periods, particularly between 2010 and 2016, which aligned with regional biomass burning events and transboundary pollution episodes. Results indicate that AOD and SSA anomalies correspond with periods of elevated temperature and reduced rainfall, highlighting potential climate-aerosol feedbacks. The findings demonstrate the utility of multivariate HMMs in capturing the complex dynamics of aerosol-climate interactions and provide a foundation for improved air quality monitoring and climate impact assessments in Kenya which is critical for improving environmental monitoring and enhancing regional climate adaptation strategies.