<![CDATA[Flood early warning in Indonesian watersheds is constrained by sparse hydrological instrumentation and by the absence of quantitative evidence linking upstream deforestation to downstream flood response. This study reports the design and field validation of a low-cost LoRaWAN sensor network intended as the acquisition layer of a watershed digital twin, together with a decadal deforestation assessment for the Krueng Peusangan watershed in Aceh, Indonesia. Land cover was classified from Sentinel-2 and Landsat imagery for 2015 to 2025 using Google Earth Engine, and the resulting curve numbers were propagated through an SCS-CN rainfall-runoff model. Two sensor nodes were deployed and operated continuously for eight days, generating 22,122 raw records. Analysis indicates forest cover declining by an estimated 38.4 percent, with the derived runoff coefficient rising from 0.43 to 0.59 and simulated peak discharge increasing by 66 percent under a 120 mm design storm. Network evaluation separates measurement, transmission, and storage success: 84.3 percent of scheduled five-minute slots were populated at the upstream node, with a mean of 5.96 stored records per slot indicating substantial retransmission overhead. Atmospheric sensors showed a temperature-humidity correlation of -0.979 with a 95 percent confidence interval of -0.980 to -0.978, and diurnal temperature patterns correlated at 0.988 across observation days. The turbidity channel saturated at its upper bound in all records and is therefore unusable without site-specific recalibration. The validated acquisition layer corresponds to Technology Readiness Level 3; predictive modelling and warning performance evaluation remain future work.]]>
Copyrights © 2026
