Pre-registration: L-band detection of alternate wetting and drying in a Balinese subak (Subak Kloncing, Bali, Indonesia)

Abstract

This deposit states, in advance of the observations that would test it, what would count as satellite detection of alternate wetting and drying (AWD) in a smallholder rice landscape. Subak Kloncing, Bali, has 32 farmers enrolling 1.63 ha of AWD within a 101 ha subak — median plot 5 are, on terraces, with treatment and control plots interleaved at 11-16 m. Sentinel-1 C-band cannot separate the two arms at this geometry: a self-calibrating classifier that successfully labelled 700 wet and 1,232 dry pixels from Sentinel-2 optical then found the corresponding VV distributions overlapping and inverted, with a gap of -2.80 dB and medians 0.30 dB apart. NISAR L-band GCOV at 10 m posting, HH and HV, yields 11 unambiguous cells per arm after ground contested between arms is removed and the remainder eroded by half a resolution cell. Seven pre-drainage acquisitions from 23 June to 10 August 2026 give an arm difference in HH of +1.51 dB with a standard deviation across acquisitions of 0.52 dB. Nothing was drained on any of these dates, so this offset is structural and stable; ascending and descending passes twelve hours apart on 23 June agree to 0.01 dB. The detection criterion is fixed here and not to be revised: a post-drainage acquisition constitutes evidence of detected drainage if the AWD minus FLD difference in HH exceeds +2.6 dB, with no corresponding shift in HV. Both conditions must hold. The first acquisitions that could test this fall on 3 and 15 September 2026. The design supports an arm-level comparison and makes no per-plot claim. The deposit contains no personal data. Version 1.1.0 adds an Indonesian translation (PRAREGISTRASI_ID.md). The English text remains authoritative; in case of discrepancy the English governs. No figures, thresholds, or analysis commitments have changed from version 1.0.0.