the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 21 Aug 2026
ALERT: A Scalable Cloud-Native Framework for Satellite Rainfall-Driven Landslide Early Warning in Data-Scarce Regions
Abstract. Rainfall-triggered landslides cause thousands of fatalities each year, yet effective early warning remains challenging across many mountainous regions because of sparse rainfall observations and limited forecasting infrastructure. Here we present ALERT (Automated Landslide Early Risk Tracker), a scalable, cloud-native framework that integrates NASA IMERG satellite precipitation, ECMWF operational forecasts, terrain susceptibility, rainfall intensity–duration thresholds, debris-flow runout modelling, and building exposure to deliver near-real-time, impact-based landslide warnings. ALERT incorporates a globally transferable catalogue of 24 rainfall thresholds while allowing user-defined thresholds and susceptibility layers, enabling deployment across diverse climatic and geomorphological settings. Event-based validation achieved a sensitivity of 0.80 and an overall accuracy exceeding 0.70, demonstrating effective detection of hazardous rainfall conditions for landslide initiation. By reducing dependence on dense ground-based rain-gauge networks, ALERT provides a computationally efficient and scalable pathway towards regional scale landslide early warning and climate-resilient disaster risk reduction, particularly in data-scarce regions.
- Preprint
(2437 KB) - Metadata XML
- BibTeX
- EndNote
Status: open (until 02 Oct 2026)
- RC1: 'Comment on egusphere-2026-4588', Anonymous Referee #1, 21 Aug 2026 reply
-
RC2: 'Comment on egusphere-2026-4588', Anonymous Referee #2, 29 Aug 2026
reply
The manuscript presents an interesting and potentially valuable framework for rainfall-triggered landslide early warning, particularly through the integration of satellite rainfall observations, numerical weather forecasts, susceptibility information, runout modelling, and exposure assessment within a cloud-based environment. The topic is relevant and the proposed framework has good potential for application in data-scarce regions.
The following points should be addressed to improve the clarity and robustness of the manuscript:
Clarify the methodology used to derive the rainfall intensity–duration (I–D) threshold. In particular, please provide a clearer justification for the selection of the 10th percentile and briefly discuss how this choice may affect the resulting threshold.
Clarify the independence between calibration and validation. Please specify which landslide events were used to derive the threshold and which were used for validation, and discuss whether any overlap exists between these datasets.
Expand the discussion of rainfall uncertainty. Since IMERG and ECMWF products have considerably coarser spatial resolutions than individual rainfall-triggering cells, it would be useful to discuss more explicitly how this scale mismatch and potential precipitation biases may affect threshold exceedance and warning performance.
Moderate the claims regarding global transferability. The framework is clearly designed to be transferable to other regions; however, its performance has been demonstrated primarily for the Western Ghats. It would therefore be useful to distinguish between the framework's potential for global transferability and its empirical validation at the global scale.
Strengthen the discussion of rainfall threshold approaches. I recommend considering the recent work of Silva et al. (2026), which proposes a combined empirical rainfall-triggered landslide threshold incorporating preparatory and triggering rainfall conditions. This study may provide a useful comparison with the I–D approach adopted here and would strengthen the discussion of alternative threshold formulations.
Silva, R. F., Marques, R., & ZĂŞzere, J. L. (2026). A New Approach for Developing Combined Empirical Rainfall-Triggered Landslide Thresholds: Application to SĂŁo Miguel Island (Azores, Portugal). Water, 18(6), 673. https://doi.org/10.3390/w18060673
Citation: https://doi.org/10.5194/egusphere-2026-4588-RC2
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 205 | 64 | 25 | 294 | 22 | 22 |
- HTML: 205
- PDF: 64
- XML: 25
- Total: 294
- BibTeX: 22
- EndNote: 22
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
This is an interesting and potentially useful framework, but the manuscript currently overstates its performance. In its present form, ALERT is a promising web application, not a validated operational landslide early-warning system.
Here, my main concerns are:
There are also several inconsistencies: accuracy is exactly 0.70, not exceeding 0.70; 5,474 + 7,707 equals 13,181, not 13,188; the abstract reports 24 thresholds whereas the manuscript reports 23; and Figure 9 contains only 65 landslide samples despite stating that 104 were available.