TROPOSPHERIC WATER VAPOR ESTIMATION FROM HIMAWARI-8 / ADVANCED HIMAWARI IMAGER (AHI) AND MICROWAVE RADIOMETER PROFILER (MRP)

• Main Author: Cika Nur Fatihah, Shaffira
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/50148
• Institution: Institut Teknologi Bandung

Tropospheric water vapor can be estimated using various methods. Himawari-8 / Advanced Himawari Imager (AHI) observe the presence of water vapor through channel 8?10 (6.2, 6.9, and 7.3 ?m) with high spatial and temporal resolution. However, the common way of AHI to express water vapor content is by imagery interpretation in the form of brightness temperature (BT). This study tried to find a way how AHI can produce a quantitative value of water vapor content in the form of water vapor density (WVD), in the lower and middle-upper layer of troposphere. The estimation method approached by creating a linear regression model by combining AHI data with the Microwave Radiometer Profiler (MRP) data. Types of model are ensemble regression model (ERM) of a multiple linear regression (MLR) and a combination of univariate regression. Both models were approached with the K-Fold cross-validation technique on hourly data and data that had been convoluted. The results of testing the application of the two types of regression models in Serpong (6.36°S and 106.67°E), Jakarta Soekarno-Hatta (6.11°S and 106.65°E), and the western part of Java Island, show that the ERM MLR provides water vapor estimation with higher accuracy, but lower variance ratio with the observed values than the ERM combination of univariate regression. Meanwhile, the ERM combination of univariate regression gives estimation results that more reliable with physically consistent regression coefficients and has a similarity in variance ratio with the observed value. So, it can show a similar pattern that can represent observation data. However, there is a bias that needs to be corrected and it has a lower accuracy than the ERM MLR.