RAINFALL ESTIMATION ON C-BAND WEATHER RADAR IMAGE BASED ON MACHINE LEARNING AND AUTOMATIC RAIN GAUGE

• Main Author: Ananda, Naufal
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/81507
• Institution: Institut Teknologi Bandung

Weather radar is a remote sensing instrument that can estimate rainfall. The output of weather radar observation is in the form of image data that displays reflectivity variables. This research aims to develop the Convolutional Neural Network (CNN) method for estimating rainfall. CNN has advantages in reading and detecting objects in the form of images. The performance of rainfall estimation using CNN is tested to other estimator models and actual rainfall data of automatic rain gauge. Other rainfall estimator models use the Z-R Marshall Palmer and Rosenfeld Tropical equations. Both Z-R equation models are used in weather operational services. The data period used in this study is December 1, 2022, to February 28, 2023. The study used 12 automatic rain gauge spread within the maximum radius range of weather radar observation, which is as far as 220 km. Based on the performance test shows that the influence of distance and elevation is a contributing variable in increasing the error, and the influence of the Z-R constant in estimating rainfall is a factor that can contribute to the estimation error. Rainfall estimation data on weather radar can only reach areas with automatic rain gauge. Keywords: rainfall, CNN, Marshall-Pallmer, weather radar, Rosenfeld Tropical.