Water Depth Measurement and Cross-Section Extraction of Concrete Open Channel using ICESat-2 ATL03 Photon-Counting Laser Altimetry

Agricultural water infrastructure requires continuous monitoring of channel geometry and water surface conditions to ensure stable water distribution. This study applies NASA's ICESat-2 ATL03 geolocated photon product, derived from the ATLAS 532 nm photon-counting laser, to an agricultural open channel in central Japan. Eighty-four ATL03 granules acquired between 2018 and 2026 were identified as intersecting the channel corridor and processed to extract along-track photon elevations, converted from ellipsoidal to orthometric heights by geoid correction; where the photon count of a beam within the study area exceeded a threshold, the photons were subsampled. Runs of consecutive retained photons within a narrow corridor about the channel centreline, excluding tunnel and bridge segments, were treated as beam–channel intersections. Within each run, the water surface was taken as the median elevation of the photons passing screening of ATL03 signal confidence for the inland-water surface type. A channel-bed return was then sought among all photons of the run lying beneath that level, without confidence screening, and accepted where the modal bin of their elevation histogram met a minimum photon count; the resulting depth was corrected for in-water refraction. Intersections with anomalously large photon dispersion were rejected, leaving 163 intersections over 45 acquisition dates with a median within-intersection dispersion of 0.29 m. Bed returns were obtained at 38 of these intersections, giving depths of 0.21–3.15 m with a median of 0.39 m. Detection statistics are reported for the 80 intersections processed without subsampling. Among these, bed returns were obtained at 16 of 49 night-time intersections against 4 of 31 daytime ones, and at 9 of 25 strong-beam intersections against 11 of 55 weak-beam ones; neither contrast reaches significance at this sample size, and acquisition season shows no difference. At 14 locations observed on two or more dates between 2018 and 2025, the retrieved water surface elevation spans 0.05–1.11 m, with a median span of 0.16 m and eight locations within 0.2 m; the same median span is obtained from the unsubsampled intersections alone. These spans describe the repeatability of the retrieval. Absolute accuracy is not assessed, as no in situ water level or discharge record was used. At these locations, the observation dates cluster into a small number of narrow windows of the year rather than sampling them uniformly, so the observed differences are not attributed here to irrigation-related operational changes.