High-Resolution Rainfall Estimation Using Ensemble Learning Techniques and Multisensor Data Integration

Abstract

In Indonesia, the monitoring of rainfall requires an estimation system with a high resolution and wide spatial coverage because of the complexities of the rainfall patterns. This study built a rainfall estimation model for Indonesia through the integration of data from various instruments, namely, rain gauges, weather radars, and weather satellites. An ensemble learning technique, specifically, extreme gradient boosting (XGBoost), was applied to overcome the sparse data due to the limited number of rain gauge points, limited weather radar coverage, and imbalanced rain data. The model includes bias correction of the satellite data to increase the estimation accuracy. In addition, the data from several weather radars installed in Indonesia were also combined. This research handled rainfall estimates in various rain patterns in Indonesia, such as seasonal, equatorial, and local patterns, with a high temporal resolution, close to real time. The validation was carried out at six points, namely, Bandar Lampung, Banjarmasin, Pontianak, Deli Serdang, Gorontalo, and Biak. The research results show good estimation accuracy, with respective values of 0.89, 0.91, 0.89, 0.9, 0.92, and 0.9, and root mean square error (RMSE) values of 2.75 mm/h, 2.57 mm/h, 3.08 mm/h, 2.64 mm/h, 1.85 mm/h, and 2.48 mm/h. Our research highlights the potential of this model to accurately capture diverse rainfall patterns in Indonesia at high spatial and temporal scales.

Keywords: ensemble learning; multisensor; rainfall.

Figure 1

Architecture of the ensemble learning-based rainfall estimation model using multiple instruments: rain gauges, integrated weather radars, and weather satellites.

Figure 2

A map of the study area in Indonesia with three different rain patterns and a map of the weather radar network used in this research.

Figure 3

Scatter plots and correlation of GPM and rain gauge data in (a) Lampung, (b) Banjarmasin, (c) Deli Serdang, (d) Pontianak, (e) Gorontalo, and (f) Biak.

Figure 4

Scatter plots and correlations between GPM and Himawari satellite data in (a) Lampung, (b) Banjarmasin, (c) Deli Serdang, (d) Pontianak, (e) Gorontalo, and (f) Biak.

Figure 5

Scatter plots and correlations between GPM and weather radar data in (a) Lampung, (b) Banjarmasin, (c) Deli Serdang, (d) Pontianak, (e) Gorontalo, and (f) Biak.

Figure 6

Comparison of (a) PDF and (b) CDF graphs between measured rainfall, uncorrected GPM, and corrected GPM. The x-axis represents the possible amount of rainfall. The y-axis represents the probability value.

Figure 7

Data fusion technique for multiple weather radars with frequency, polarization, and coverage range differences. There are several coverage overlaps of the weather radars with the data from each radar, which are combined into a single data output. The dot is the weather radar location, and the circle is the weather radar range.

Figure 8

XGBoost hyperparameter optimization graph using Bayesian optimization.

Figure 9

Rainfall estimation products in mm/h using ensemble learning techniques and multisensor data integration produced in this study.

Figure 10

Image capture of GPM, radar, satellite, and estimation model results as well as comparison graphs of rain between rain gauge, GPM, and rainfall estimations during ongoing rain that occurred during very heavy rain (>20 mm/h) in Bandar Lampung on 16 December 2022 between 13.00 and 14.30.

Figure 11

Image capture of GPM, radar, satellite, and estimation model results as well as comparison graphs of rain between rain gauge, GPM, and rainfall estimation during ongoing rain which occurred during heavy rain (10–20 mm/h) in Biak on 26 December 2022 between 16.30 and 18.00.

Figure 12

Image capture of GPM, radar, satellite, and estimation model results as well as comparison graphs of rain between rain gauge, GPM, and rainfall estimation during ongoing rain that occurred during moderate rain (10–20 mm/h) in Gorontalo on 17 December 2022 between 11.30 and 12.30.

Figure 13

Image capture of GPM, radar, satellite, and estimation model results as well as comparison graphs of rain between rain gauge, GPM, and rainfall estimation during ongoing rain which occurred during light rain (5–10 mm/h) in Banjarmasin on 5 December 2022 between 05.30 and 06.30.

Figure 14

Image capture of GPM, radar, satellite, and estimation model results as well as comparison graphs of rain between rain gauge, GPM, and rainfall estimation during ongoing rain which occurred during very light rain (<1 mm/h) in Pontianak on 23 December 2022 between 08.00 and 09.00.