Vol. 1 No. 1 (2026): Journal of Earth and Planetary System Science

Background: Tropical cyclones forming within or near the Malacca Strait can trigger extreme rainfall over North Sumatra through interactions between cyclone dynamics, atmospheric moisture transport, and the orographic effects of the Bukit Barisan mountain range. However, the mechanisms linking Tropical Cyclone Senyar to remote orographic extreme rainfall remain poorly understood.
Aims: This study investigates the relationship between the structure of Tropical Cyclone Senyar and remote orographic extreme rainfall over North Sumatra.
Methods: ECMWF reanalysis data were processed using Python-based computational and visualization tools. The analyzed variables included total precipitation, 10-m zonal and meridional wind components, mean sea level pressure, specific humidity, and vertical velocity at 37 pressure levels. Topographic influences were represented using the 30-m SRTM Digital Elevation Model
Result: During 21–30 November 2025, Tropical Cyclone Senyar produced total rainfall accumulations of 117.16–871.48 mm (mean: 340.25 mm) and a maximum daily rainfall of 327.01 mm day⁻¹, with a minimum mean sea level pressure of 1009.64 hPa. Extreme rainfall (>150 mm day⁻¹) occurred mainly over Sibolga, North Tapanuli, Langkat, Medan, and Deli Serdang. Strong moisture transport, organized moisture flux corridors, localized moisture convergence, and persistent negative omega values at 500 hPa promoted deep convection. Rainfall maxima coincided with the orographic forcing zone along the Bukit Barisan range.
Conclusion: The integrated analysis demonstrates that Tropical Cyclone Senyar, moisture transport, vertical atmospheric dynamics, and orographic forcing acted synergistically to generate remote extreme rainfall over North Sumatra, providing new insights into extreme precipitation mechanisms in the equatorial maritime region.

Background: Climate change has affected rainfall patterns, particularly extreme rainfall events. As a region with high levels of tourism and socio-economic activity, the province of Bali is vulnerable to these changes. However, studies on trends in extreme rainfall in the Bali region remain limited.
Aims: This study aims to determine the trends and magnitude of changes in extreme rainfall indices (R20mm, RX1day, PRCPTOT, and CDD) in the Province of Bali during the period 1994–2023.
Methods: The data used consists of daily rainfall records from 17 BMKG stations, with a data completeness rate exceeding 90%. Extreme rainfall indices were calculated using the ETCCDI method. Trend analysis was conducted using the Mann-Kendall test and Sen’s Slope estimator. Additionally, monthly trend analysis was performed as a follow-up.
Result: The results indicate an increase in intensity-based extreme rainfall indices (RX1day, R20mm, and PRCPTOT), whilst the CDD index shows a decreasing trend. Several stations showed a significant upward trend, particularly in the PRCPTOT and RX1day indices. Monthly analysis also indicated increased rainfall in certain months, including the period from May to September.
Conclusion: These results highlight the importance of climate adaptation strategies and the mitigation of hydrometeorological disasters, particularly in the Bali region.

Background: Water inundation is a recurring environmental hazard in Semarang Regency that frequently affects settlements, agricultural land, and infrastructure. Mapping inundation recurrence is essential for improving flood risk management and supporting spatial planning. The availability of cloud-based geospatial platforms such as Google Earth Engine (GEE) enables efficient multitemporal analysis of satellite imagery for monitoring inundation dynamics.
Aims: This study aims to identify and classify water inundation-prone areas in Semarang Regency by analyzing the spatial distribution and recurrence of inundation using Landsat observations acquired from 2020 to 2025.
Methods: Landsat 8 and Landsat 9 surface reflectance imagery, the administrative boundary of Semarang Regency, and SRTM Digital Elevation Model (DEM) data were processed in the Google Earth Engine environment. The Modified Normalized Difference Water Index (MNDWI) was calculated from the green and shortwave infrared bands to produce annual water masks. Pixel-based inundation frequency was summarized over six years and classified into vulnerability categories.
Results: Annual inundation extent fluctuated between 8.63 km² in 2020 and a maximum of 23.77 km² in 2021, before decreasing and increasing again to 15.16 km² in 2025. Recurrently inundated areas were mainly concentrated around Rawa Pening and surrounding lowlands characterized by flat to gently sloping terrain and dominated by agricultural and settlement land cover.
Conclusion: The proposed framework effectively integrates Landsat-derived MNDWI and Google Earth Engine to identify recurrent inundation patterns and produce spatially explicit vulnerability maps. The results provide valuable information for flood risk assessment, land-use planning, and disaster mitigation in Semarang Regency

Background: Weather information plays an important role in supporting people's daily activities, including travel, agriculture, outdoor activities, and decision-making. However, raw meteorological data delivered through Application Programming Interface (API) services is commonly presented in JSON format, which is highly technical and difficult for the general public to interpret.
Aims: This research aims to democratize access to raw weather data by developing an interactive and personalized Telegram chatbot that serves as an intermediary system (middleware) between users and external weather data APIs. The system is designed to provide weather information in a natural, contextual, and user-friendly form.
Methods: The proposed system implements a Two-Stage Large Language Model (LLM) Pipeline architecture utilizing the Gemini 3.1 Pro model. The architecture separates the LLM workload into two specialized stages. The first stage, the Router Engine, functions as a probabilistic semantic classifier to identify and categorize user intentions. The second stage, the Formatter Engine, acts as a contextual text generator that processes factual JSON weather data obtained from external APIs and transforms it into concise, casual, and understandable narratives
Result: The evaluation results demonstrate high system stability, with context recognition accuracy reaching 98.8%. The proposed processing architecture also achieves network response latency below 2.5 seconds. Furthermore, the graceful degradation mechanism demonstrates strong fault tolerance when the system encounters API quota limitations.
Conclusion: The separation of cognitive roles within the LLM, combined with external factual data extraction, effectively reduces cognitive barriers to weather information and mitigates the risk of hallucination.

Background: Monthly rainfall forecasting is important for water-resource management, agricultural planning, and hydrometeorological risk assessment. However, reliable forecasting requires a temporally consistent dataset and appropriate representation of seasonal rainfall variability.
Aims: This study aims to reconstruct and audit a monthly rainfall dataset for the Indragiri region, identify its temporal and seasonal characteristics, evaluate alternative Seasonal Autoregressive Integrated Moving Average (SARIMA) models, and assess their forecasting performance and residual adequacy.
Methods: Monthly rainfall data derived from reanalysis were reconstructed for January 2014-June 2025, resulting in 138 monthly records. Temporal auditing identified June 2025 as incomplete, containing only 20 of the expected 60 timestamps; therefore, 137 complete months through May 2025 were used for model development. Autocorrelation and partial autocorrelation analyses were conducted to identify temporal and seasonal structures. Eight SARIMA candidates were compared using AIC, BIC, RMSE, MAE, sMAPE, and Ljung–Box residual diagnostics.
Results: The rainfall series had a mean of 16.15 mm, standard deviation of 13.19 mm, and maximum of 49.59 mm, with strong annual seasonality. SARIMA(0,0,2)(2,1,0)[12] achieved the lowest test RMSE (10.36 mm), MAE (6.98 mm), and sMAPE (48.17%). However, significant residual autocorrelation remained at lags 6, 12, 18, and 24 (p < .001).
Conclusion: The selected SARIMA model provides a useful seasonal forecasting baseline but does not fully capture the dependence structure of monthly rainfall. Forecasts for July 2025–June 2027 indicate persistent seasonal variability, highlighting the need for additional climatic predictors or alternative models to improve forecasting accuracy.