← All proceedings

TheIIER INTERNATIONAL CONFERENCE

Proceedings cover: TheIIER INTERNATIONAL CONFERENCE

Zurich, Switzerland 16-01-2026

2 papers

  1. ENHANCING FLOOD ROUTING SIMULATION OF DAM SAFETY REMOTE MONITORING SYSTEM BY MACHINE LEARNING TECHNIQUE

    KANOKVATE TUNGPIMOLRUT, RANGSARIT VANIJJIRATTIKHAN, JITTIWUT SUWATTHIKUL, MANUT TANTICHAROEN, CHIRAPOT VUTHIPRAPHAN, ARNAN JOMTARAX, SEUBSUANG KACHAPORNKUL, SHINICHI MIYAKAWA, SHINSUKE MIWA, TOSHIYUKI MIYACHI, JESSADDA KARNJANA

    Paper ID II-RIETZRCH-160126-31912 · Pages 1-7

    Abstract

    In this paper, a concept for integrating machine learning–based inflow forecasting into the flood routing simulation of a dam safety remote monitoring system is proposed. A Fully Connected Neural Network (FCNN) is selected as the forecasting model for reservoir inflow at the Ubol Ratana Dam, used here as a case study. Based on nine years of recorded daily historical environmental parameters at the dam site, preliminary results indicate that the model achieves improved performance, with a correlation coefficient ranging from 0.60452 to 0.82921. This is higher than the results obtained from a model based on monthly and global parameters, which produced a correlation coefficient of approximately 0.40. The 15-day-ahead forecasted inflow aligns well with the actual inflow. When integrated into the flood routing simulation, a similar pattern of spillway gate operations is observed, and the reservoir water level can be effectively managed. These findings validate that the proposed concept can enhance flood routing simulations by expediting the process of identifying suitable hydrographs that align with realworld conditions—traditionally done manually and randomly—thereby making the system more adaptive.

    Keywords - Dam Safety Management, Flood Routing Simulation, Machine Learning, Reservoir Operation.

    Download PDF
  2. HOW LONG DO READING GAINS HOLD? META-ANALYTIC EVIDENCE OF FADEOUT TRAJECTORIES

    FLORINA ERBELI, YING XU

    Paper ID II-ELTZRCH-160125-32051 · Pages 25

    Abstract

    Purpose:
    Reading interventions improve short-term reading outcomes (i.e., post-test effects). However, positive impacts of these interventions fade out over time (i.e., follow-up effects). The purpose was to examine the extent to which reading intervention gains persist or fade out over time.
    Method:
    We systematically collected experimental studies of reading interventions reporting the same reading outcome at post-test and follow-up. We grouped follow-up assessments into four bins (1 week–4 months, 5–8 months, 9–12 months, 12–24 months). Hedges’ g was the effect size. To examine fadeout trajectories, we computed average weighted effects at each follow-up bin. We then estimated metaregression models linking post-test effects to follow-up effects, where the slope captured the proportion of post-test gains that carried into follow-up, whilst the intercept represented additional follow-up effects not explained by post-test effects.
    Results:
    Thirty-five studies were included (117 effect sizes). The average post-test effect was 0.34, declining to 0.29 at 1 week–4 months and 0.28 at 5–8 months. At 1 week–4 months, the slope was nonsignificant. At 5–8 months, the slope was significant, indicating that 52% of post-test gains persisted. Small nonsignificant intercepts at 1week–4 months and 5–8 months indicated limited evidence of benefits beyond those tied to post-test gains. Results for estimates greater than 8 months are not reported here.
    Conclusions:
    Reading intervention effects declined over time, with roughly half of the initial gains persisting by 5-8 months. Findings highlight that sustainedbenefits depend strongly on the magnitude of post-test effects.

    Download PDF