International Journal of Hydrology Science and Technology (5 papers in press)

Regular Issues

• Advancing streamflow prediction with Kolmogorov-Arnold network and LSTM: a comparative analysis  
  by Vipul Santrambhai Varma, Jayantilal N. Patel 
  Abstract: Accurate streamflow forecasting is a prerequisite for integrated water resource management and effective flood mitigation strategies. This study presents a comparative evaluation of the Kolmogorov-Arnold network (KAN) and the long short-term memory (LSTM) model for daily discharge prediction. The novelty of this research lies in its pioneering application of KAN, an innovative architecture based on learnable B-spline activation functions on edges to river discharge forecasting across both regulated (Garudeshwar) and natural (Handia) flow regimes in the Narmada River Basin, India. Utilising longitudinal data from 1974 to 2018, the results demonstrate that KAN achieved higher predictive fidelity than the LSTM benchmark, specifically in regulated environments. Quantitatively, at the Garudeshwar station, KAN achieved a coefficient of determination (R2) of 0.7756 and reduced the root mean square error (RMSE) to 819.11, compared to the LSTMs R2 of 0.6975 and RMSE of 935.34. Furthermore, KAN mitigated the systemic peak-smoothing observed in gated architectures, providing a nearly 44% reduction in overall mean absolute error (MAE), largely attributed to improved accuracy during high magnitude events. Significantly, KAN achieved these results with a parameter count reduction of up to 98%, establishing it as a parsimonious and computationally efficient tool for real-time hydrological forecasting.
  Keywords: streamflow; Kolmogorov-Arnold network; KAN; long short-term memory; LSTM; forecasting; machine learning.
  DOI: 10.1504/IJHST.2026.10078563
   
  
• Application of RUSLE for estimating medium- and long-term water erosion in the Agneby watershed (southeast Cote dIvoire)  
  by Ahou Sabine Kouadio, Ehouman Serge Koffi, Brahima Kone, Amidou Dao, Bamory Kamagate 
  Abstract: The Agneby watershed (8,450 km2), located in southeastern Cote dIvoire, has undergone major transformations driven by agro-ecosystems that promote sheet erosion. This study aims to estimate and map erosion at medium- and long-term scales using the RUSLE model integrated into GIS, based on five factors: rainfall erosivity, soil erodibility, slope and slope length, vegetation cover, and soil conservation practices. The results reveal a progressive degradation between 1988, 2050 and 2080. In 1988, the very low class overwhelmingly dominates (98.52%; 8,088.50 km2; 1.37 t/ha/yr), while the higher classes remain marginal. In 2050 and 2080, this dominance slightly decreases (95.76% then 95.61%), in favour of the low and moderate classes (up to 3.43% and 0.78%). The high and very high classes also increase, reaching 26.99 and 59.38 t/ha/yr in 2080, respectively, indicating a growing vulnerability of soils to erosion.
  Keywords: sheet erosion; RSULE; spatial modelling; multi-criteria analysis; Agneby watershed; Cote d’Ivoire.
  DOI: 10.1504/IJHST.2026.10080027
   
  
• Leaf litter (allochthonous energy) input manipulation in water quality management of non-perennial streams  
  by Manimeldura D. D. Perera, Pattiyage I.A. Gomes, Wei Zhao 
  Abstract: The abundance of non-perennial streams is expected to increase due to climate change and anthropogenic activities. Therefore, their management is a timely need. Drying non-perennial streams was simulated using a series of mesocosm experiments with an indicator fish species, aiming to investigate the water quality changes and the effect of varying allochthonous energy input (i.e., leaf-litter content from 22 grams dry/m2 to 1,440 grams dry/m2 for a water column of 0.025 L). Dissolved oxygen and ammonia decreased with increasing litter content, showing strong and significant linear correlations (R2 > 0.7; P < 0.05). However, electrical conductivity, nitrite, phosphate and nitrate did not show notable correlations or a leaf litter content-based response. Mesocosms with higher leaf litter content showed earlier fish birth but with fewer offspring, earlier fish death, and fungal spots. This study gave insights into the feasibility of leaf litter control in managing non-perennial streams.
  Keywords: ammonia; blackwater events; dissolved oxygen; drying of streams; fish; mesocosms; stream management.
  DOI: 10.1504/IJHST.2026.10080193
   
  
• Data-driven groundwater monitoring strategies to enhance water resource sustainability  
  by Deshmukh Sachin Chandrakant, Gosavi Shrikrishna Avinash, Azharoddin Abdul Qadar Peerampalli, Shailaja Amrutlal Doddi, Anant Pradeep Pandit 
  Abstract: To enrich the ground water (GW) stability, sustainability knowledge is required in every climate and season conditions. To well establish the sustainability, different intelligent models were executed in past, but those sustainability outcomes are not evaluated under the different seasonal changes or climate changes conditions. The present research work has aimed to model novel intelligent solution called as pipe fish sequential neural framework (PFSNF) for the prediction of water sustainability. The chief process like data initialisation, filtering, hyperparameters tuning, feature extraction, prediction and alert generation has been executed in this present work. Here, the SNN parameters are regulated by the fitness of the pipe fish best solution for enhancing the feature analysis outcome. It offered the finest prediction outcome about 0.99R-square, and the recorded RMSE and MAE are 0.61 and 0.45 respectively. Compared to other baselines, the novel solution performance was improved by 4%.
  Keywords: ground water sustainability; pipe fish optimisation; sequential neural network; ground water monitoring process; feature analysis.
  DOI: 10.1504/IJHST.2026.10080308
   
  
• Understanding the impact of geodetic measurements uncertainty on drought assessment: a case study of Europe  
  by Artur Lenczuk, Christopher Ndehedehe, Anna Klos, Janusz Bogusz 
  Abstract: A changing climate reveals that drought are becoming more severe and frequent, with cascading impacts on the environment, and people. As opposed to meteorological-based drought indices, GRACE and GPS observations have provided new capabilities and insights that improve understanding of drought dynamics. However, there are also limitations in the use of geodetic data. Thus, we examine the impact of GRACE and GPS residual errors on defined droughts estimating DSI and its uncertainty based on various approaches. Our results indicate the largest uncertainties above 35 degree. We observe a good spatial consistency in displacement uncertainties between GPS and both GRACE-type data for Western Europe, with relative RMS values over 20%. In comparison to original DSI, uncertainties contain 20% and 100% of DSI signal at 48% and 5% of stations for MSC JPL and GPS, respectively. For TCH method, we identify region-dependent hot spots and north-south strips for displacements and DSI, respectively.
  Keywords: global positioning system; GPS; GRACE; vertical displacements; DSI uncertainty; three-corner hat method; drought.
  DOI: 10.1504/IJHST.2026.10080309