BRIN Researcher Reveals Bias in Weather Predictions in Indonesia, Global Models Miss Peak Rainfall by 5 Hours

The accuracy of weather predictions in the Maritime Continent of Indonesia sometimes still misses the mark. The latest research from BRIN researcher Sopia Lestari reveals a persistent bias in global atmospheric models that fail to capture peak rainfall times by up to a five-hour difference. Through the topic of Seasonal Variability of Stratiform and Convective Rainfall Structure During MJO Phases over the Maritime Continent (MC), Sopia notes that the rainfall cycle in the maritime continent generally always has a persistent bias, making it intriguing to study. “The focus of my study is because the rainfall cycle in the maritime continent generally always has a persistent bias,” she emphasised during the PRIMA-BRIN Research Collaboration Workshop with the Center for Orbital and Suborbital Observation, Institute for Space-Earth Environmental Research (ISEE), Nagoya University, on Tuesday (31/3/2026). Sopia explains that the MJO is an atmospheric wave that moves from west to east, bringing active convection areas (clouds and heavy rain) and quiet areas (dry and clear) alternately. In the atmosphere, this phase is a large-scale pattern of tropical climate variability that mainly occurs in tropical regions and affects global weather and climate. Sopia presented several simulation model displays and MJO observations. “We know that the MJO is a large-scale oscillation that passes through Indonesia and also has phases 1 to 8. We can see that the peak rainfall observed by WARP has two peaks, whereas that observed by the STEM satellite has only one peak. So, from this image, we can show that the model captures the peak rainfall phase inaccurately during the MJO,” she explained, quoted from the BRIN website on Monday (6/4/2026). Furthermore, Sopia showed other results comparing the model with observations, namely in the case of Sumatra and Borneo. According to her, the Y-axis shows the average hourly rainfall, and the X-axis shows the time of day. “So, the dashed line shows the average hourly rainfall, the thick line and thin line; the thick line shows observations, the thin line shows the model. So, we can see here that the peak rainfall shown by the model and also by observations has quite a different time gap,” she emphasised. This indicates that the peak rainfall according to observations occurs around late night, but according to the model, it appears that there is no peak, but rather it gets lower. “So from here we can see there is an error in determining global rainfall time and a maximum of four to five hours, which shows that separating rainfall types is very important to simulate the rainfall cycle phase correctly, especially over the maritime continent,” Sopia revealed. Furthermore, she highlighted that based on this, the challenge is that Indonesia has complex topography, while data and research related to rainfall types are still scarce.