Soil Moisture, Not Daily Rainfall, Marks the Location of Monsoon Road-Landslide Closures in Uttarakhand: A Low-Cost, Reproducible Validation Using Government Road-Closure Records

Monsoon road-landslides cause recurrent loss of life and connectivity across the Indian Himalaya, yet real-time, spatially resolved risk information rarely reaches those most exposed. We developed an accessible, no-login, phone-based tool (CREST) that monitors road-landslide risk in Uttarakhand in real time from public weather data. Treating 1,900+ distinct Uttarakhand PWD road-closure records as ground truth, we find that at closure sites soil moisture sat at approximately the 62nd percentile of the day's hill-point distribution (chance = 50th), whereas rainfall reached only the 55th–57th — replicated across nine independent samples with a significance exceeding 6σ. In this v3 we add a point-level validation using 100 exact-coordinate sites taken from PWD road-closure field-photos (which carried both the image and its GPS coordinate), across 12 distinct monsoon days and many districts: the event-day mean soil percentile was 70.0 (chance 50), with a significance of about 7.9σ. We note explicitly that these two forms — name-matched records and coordinate-bearing photos — come from the same source (PWD road-closure material), not two independent sources; the coordinate-bearing photos simply enable point-level (~5 km) validation that name-matched records (road name + KM, no coordinate) cannot. For events the tool flagged, it had signalled elevated risk in that valley about 3 days in advance on average. We make no claim of novel discovery — the role of antecedent moisture is established science. Our contribution is threefold: a low-cost validation method using self-generating government records; measured negative results; and an honest characterisation of limits. The tool is available without login, on a phone, in both Hindi and English — reaching communities that larger systems often do not. The signal resolves the dangerous valley (~5 km), not the exact slope, and does not detect cloudbursts, flash floods, tunnel failures, or earthquakes. Uttarakhand is a pilot; the framework can extend across the Himalayan belt, provided each region is re-validated against its own ground truth.