Enhancement of Forecasting Accuracy for Mass Concentration of Ultrafine Particles using WRF-Chem and Hybrid Machine Learning with SHAP-Based Explainable Analysis

Laddawan Noynoo

doi:10.1016/j.apr.2026.103110

Enhancement of Forecasting Accuracy for Mass Concentration of Ultrafine Particles using WRF-Chem and Hybrid Machine Learning with SHAP-Based Explainable Analysis

2026-06-16

0

PRE

AI

L

Laddawan Noynoo

T

Thanathip Limna

P

Perapong Tekasakul

B

Bukhoree Sahoh

C

Chidchanok Choksuchat

K

Korakot Wichitsa-Nguan Jetwanna

J

John Morris

T

Thi-Cuc Le

C

Chuen-Jinn Tsai

T

Thanita Areerob

A

Aree Choodum

Y

Yutthapong Pianroj

S

Saysunee Jumrat

R

Racha Dejchanchaiwong *

DOI：10.1016/j.apr.2026.103110

Original

Original request for help

Save

Abstract

En 中文

• BO-optimized XGBoost-LSTM model achieved the highest accuracy in predicting PM0.1. • WRF-Chem-BO-XGBoost-LSTM model achieved high accuracy, with R2 up to 0.99. • Hybrid model can reliably forecast PM0.1 concentration up to 7 days in advance. • PM0.1 was driven by pollutant predictors linked to local sources.

Enhancement of Forecasting Accuracy for Mass Concentration of Ultrafine Particles using WRF-Chem and Hybrid Machine Learning with SHAP-Based Explainable Analysis

Abstract

Journal

Organization

Cited Papers

Citing Papers