DEVELOPMENT OF MULTIVARIATE LINEAR MODEL TO ESTIMATE PM10 CONCENTRATION IN THE AMBIENT AIR OF BANDUNG

• Main Author: RISKY DIWITA (NIM : 15312041), NABELA
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/29435
• Institution: Institut Teknologi Bandung

Particulate matter were proven to have negative impacts on human health when exposed to high concentration that exceeds the threshold limit value. Numerous studies has shown a connection between particulate matter and health or <br /> <br /> environmental problems. Health complications caused by particulate matter were found to be highly linked to large populations that lives in big cities. Meteorological conditions holds an important role in the transport and dispersion of airborne pollutants and were correlated to PM levels. Multivariate models can explain the connection between PM10 concentration and meteorological conditions linearly. Data used for this research includes PM10 monitoring results in the year 2014-2015. Linear models were developed by multivariate linear techniques such as multiple linear regression and principal component regression with PM10 concentration as the response variable and meteorological conditions such as temperature, humidity, dew point, air pressure, solar radiation, and wind speed as the predictor variables. Predictor variables were reduced by principal component analysis and stepwise regression in advanced to multivariate modelling. Both modelling methods were then compared and verified with recent PM10 monitoring data taken in 2016. PM10 concentration were found to be <br /> <br /> significantly correlated with humidity, air pressure, and wind spreed. Later results showed that multiple linear regression gives better predictions compared to principal component regression. Adjusted R2 values for multiple linear regression were able to explain 50,16% of total PM10 variance while principal component regression explains as much as 48,68% of total variation.