{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,9,17]],"date-time":"2026-09-17T22:00:17Z","timestamp":1789682417876,"version":"4.0.1"},"reference-count":59,"publisher":"IWA Publishing","issue":"9","license":[{"start":{"date-parts":[[2025,10,25]],"date-time":"2025-10-25T00:00:00Z","timestamp":1761350400000},"content-version":"vor","delay-in-days":0,"URL":"http:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"funder":[{"DOI":"10.13039\/100000192","name":"National Oceanic and Atmospheric Administration","doi-asserted-by":"publisher","award":["NA22NWS4320003"],"award-info":[{"award-number":["NA22NWS4320003"]}],"id":[{"id":"10.13039\/100000192","id-type":"DOI","asserted-by":"publisher"}]}],"content-domain":{"domain":["iwaponline.com"],"crossmark-restriction":true},"short-container-title":[],"published-print":{"date-parts":[[2025,11,1]]},"abstract":"<jats:title>ABSTRACT<\/jats:title>\n                  <jats:p>Large Language Models (LLMs) have shown strong performance across natural language processing tasks, yet their general-purpose embeddings often fall short in domains with specialized terminology and complex syntax, such as hydrology and environmental science. This study introduces HydroEmbed, a suite of open-source sentence embedding models fine-tuned for four QA formats: multiple-choice (MCQ), true\/false (TF), fill-in-the-blank (FITB), and open-ended questions. Models were trained on the HydroLLM Benchmark, a domain-aligned dataset combining textbook and scientific article content. Fine-tuning strategies included MultipleNegativesRankingLoss, CosineSimilarityLoss, and TripletLoss, selected to match each task's semantic structure. Evaluation was conducted on a held-out set of 400 textbook-derived QA pairs, using top-k similarity-based context retrieval and GPT-4o-mini for answer generation. Results show that the fine-tuned models match or exceed performance of strong proprietary and open-source baselines, particularly in FITB and open-ended tasks, where domain alignment significantly improves semantic precision. The MCQ\/TF model also achieved competitive accuracy. These findings highlight the value of task- and domain-specific embedding models for building robust retrieval-augmented generation (RAG) pipelines and intelligent QA systems in scientific domains. This work represents a foundational step toward HydroLLM, a domain-specialized language model ecosystem for environmental sciences.<\/jats:p>","DOI":"10.2166\/wst.2025.156","type":"journal-article","created":{"date-parts":[[2025,10,25]],"date-time":"2025-10-25T11:32:12Z","timestamp":1761391932000},"page":"1328-1342","update-policy":"https:\/\/doi.org\/10.2166\/iwapcrossmarkpolicypage","source":"Crossref","is-referenced-by-count":4,"title":["Domain-specific embedding models for hydrology and environmental sciences: enhancing semantic retrieval and question answering"],"prefix":"10.2166","volume":"92","author":[{"ORCID":"https:\/\/orcid.org\/0000-0002-2432-2472","authenticated-orcid":false,"given":"Ramteja","family":"Sajja","sequence":"first","affiliation":[{"name":"a IIHR \u2013 Hydroscience and Engineering, University of Iowa, Iowa City, IA 52242, USA"},{"name":"b ByWater Institute, Tulane University, New Orleans, LA 70118, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0003-1516-8335","authenticated-orcid":false,"given":"Yusuf","family":"Sermet","sequence":"additional","affiliation":[{"name":"a IIHR \u2013 Hydroscience and Engineering, University of Iowa, Iowa City, IA 52242, USA"},{"name":"b ByWater Institute, Tulane University, New Orleans, LA 70118, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-0461-1242","authenticated-orcid":false,"given":"Ibrahim","family":"Demir","sequence":"additional","affiliation":[{"name":"a IIHR \u2013 Hydroscience and Engineering, University of Iowa, Iowa City, IA 52242, USA"},{"name":"b ByWater Institute, Tulane University, New Orleans, LA 70118, USA"},{"name":"c River-Coastal Science and Engineering, Tulane University, New Orleans, LA 70118, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"945","published-online":{"date-parts":[[2025,10,25]]},"reference":[{"issue":"2","key":"key-10.2166\/wst.2025.156-1","first-page":"90","article-title":"A comprehensive review of ontologies in the hydrology towards guiding next generation artificial intelligence applications","volume":"42","author":"Baydaro\u011flu","year":"2022","journal-title":"Journal of Environmental Informatics"},{"key":"key-10.2166\/wst.2025.156-2","doi-asserted-by":"publisher","DOI":"10.1109\/TKDE.2023.3235779"},{"key":"key-10.2166\/wst.2025.156-3","first-page":"8","article-title":"The language of engineering: training a domain-specific word embedding model for engineering","author":"Braun","year":"2021"},{"key":"key-10.2166\/wst.2025.156-4","doi-asserted-by":"publisher","DOI":"10.1007\/s10994-021-05997-6"},{"key":"key-10.2166\/wst.2025.156-5","doi-asserted-by":"publisher","DOI":"10.1109\/TKDE.2025.3559023"},{"key":"key-10.2166\/wst.2025.156-6","doi-asserted-by":"publisher","DOI":"10.1145\/3625224"},{"key":"key-10.2166\/wst.2025.156-7","first-page":"1","article-title":"A multilevel center embedding approach for sentence similarity having complex structures","author":"Dubey","year":"2023"},{"key":"key-10.2166\/wst.2025.156-8","unstructured":"Eitan D., Pirchi M., Glazer N., Meital S., Ayach G., Krendel G., Shamsian A., Navon A., Hetz G. & Keshet J. (2023) Combining language models for specialized domains: a colorful approach. arXiv (Cornell University). https:\/\/doi.org\/10.48550\/arxiv.2310.19708."},{"key":"key-10.2166\/wst.2025.156-9","doi-asserted-by":"crossref","first-page":"295","DOI":"10.18653\/v1\/P19-2041","article-title":"Embedding strategies for specialized domains: Application to clinical entity recognition","volume-title":"Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics: Student Research Workshop","author":"El Boukkouri","year":"2019"},{"key":"key-10.2166\/wst.2025.156-10","doi-asserted-by":"publisher","DOI":"10.1016\/j.nlp.2023.100032"},{"issue":"1","key":"key-10.2166\/wst.2025.156-11","first-page":"1","article-title":"Semantic relations, dynamicity, and terminological knowledge bases","volume":"1","author":"Faber","year":"2009","journal-title":"Current Issues in Language Studies"},{"key":"key-10.2166\/wst.2025.156-12","unstructured":"Gong X., Holmes J., Li Y., Liu Z., Gan Q., Wu Z., Zhang J., Zou Y., Teng Y., Jiang T., Zhu H., Liu W., Liu T. & Yan Y. (2023) Evaluating the potential of leading large language models in reasoning biology questions. arXiv (Cornell University). https:\/\/doi.org\/10.48550\/arxiv.2311.07582."},{"key":"key-10.2166\/wst.2025.156-13","volume-title":"New and Improved Embedding Model","author":"Greene","year":"2022"},{"key":"key-10.2166\/wst.2025.156-14","unstructured":"Hatakeyama-Sato K., Igarashi Y., Katakami S., Nabae Y. & Hayakawa T. (2023) Teaching Specific Scientific Knowledge into Large Language Models through Additional Training. arXiv (Cornell University). https:\/\/doi.org\/10.48550\/arxiv.2312.03360."},{"key":"key-10.2166\/wst.2025.156-15","doi-asserted-by":"crossref","unstructured":"Hua W., Zhang Y., Chen Z., Li J. & Weber M. (2022) LegalRelectra: Mixed-domain language modeling for long-range legal text comprehension. arXiv (Cornell University). https:\/\/doi.org\/10.48550\/arxiv.2212.08204.","DOI":"10.18653\/v1\/2023.nllp-1.7"},{"key":"key-10.2166\/wst.2025.156-16","doi-asserted-by":"publisher","DOI":"10.1038\/s42003-023-05476-9"},{"key":"key-10.2166\/wst.2025.156-17","unstructured":"Jia N., Yuan C., Wu Y. & Zheng Z. (2025) Improving LLM Interpretability and Performance via Guided Embedding Refinement for Sequential Recommendation. ArXiv, abs\/2504.11658. https:\/\/doi.org\/10.48550\/arXiv.2504.11658."},{"key":"key-10.2166\/wst.2025.156-18","doi-asserted-by":"publisher","DOI":"10.3390\/hydrology11090148"},{"key":"key-10.2166\/wst.2025.156-19","doi-asserted-by":"publisher","DOI":"10.2166\/hydro.2025.042"},{"key":"key-10.2166\/wst.2025.156-20","doi-asserted-by":"publisher","DOI":"10.1017\/eds.2025.10006"},{"key":"key-10.2166\/wst.2025.156-21","doi-asserted-by":"publisher","DOI":"10.1609\/aies.v7i1.31679"},{"key":"key-10.2166\/wst.2025.156-22","doi-asserted-by":"publisher","DOI":"10.1093\/bib\/bbaa199"},{"key":"key-10.2166\/wst.2025.156-23","unstructured":"Leivada E., Dentella V. & Murphy E. (2023) The Quo Vadis of the Relationship between Language and Large Language Models. arXiv (Cornell University). https:\/\/doi.org\/10.48550\/arxiv.2310.11146."},{"key":"key-10.2166\/wst.2025.156-24","unstructured":"Manchanda J., Boettcher L., Westphalen M. & Jasser J. (2024) The open source advantage in Large Language models (LLMs). arXiv (Cornell University). https:\/\/doi.org\/10.48550\/arxiv.2412.12004."},{"key":"key-10.2166\/wst.2025.156-25","unstructured":"Matarazzo A. & Torlone R. (2025) A Survey on Large Language Models with some Insights on their Capabilities and Limitations. arXiv (Cornell University). https:\/\/doi.org\/10.48550\/arxiv.2501.04040."},{"key":"key-10.2166\/wst.2025.156-26","unstructured":"Mukherjee M. & Hellendoorn V. J. (2023) Stack Over-Flowing with Results: The Case for Domain-Specific Pre-Training Over One-Size-Fits-All Models. arXiv (Cornell University). https:\/\/doi.org\/10.48550\/arxiv.2306.03268."},{"key":"key-10.2166\/wst.2025.156-27","volume-title":"New Embedding Models and API Updates","author":"OpenAI","year":"2024"},{"key":"key-10.2166\/wst.2025.156-28","doi-asserted-by":"crossref","unstructured":"Plaat A., Wong A., Verberne S., Broekens J., Niki V. S. & Back T. (2024) Reasoning with Large Language Models, a Survey. arXiv (Cornell University). https:\/\/doi.org\/10.48550\/arxiv.2407.11511.","DOI":"10.1145\/3774896"},{"key":"key-10.2166\/wst.2025.156-29","first-page":"100183","article-title":"Performance of ChatGPT on the US fundamentals of engineering exam: comprehensive assessment of proficiency and potential implications for professional environmental engineering practice","volume":"5","author":"Pursnani","year":"2023","journal-title":"Computers and Education: Artificial Intelligence"},{"key":"key-10.2166\/wst.2025.156-30","doi-asserted-by":"publisher","DOI":"10.1016\/j.ijdrr.2025.105422"},{"key":"key-10.2166\/wst.2025.156-31","doi-asserted-by":"crossref","unstructured":"Reimers N. & Gurevych I. (2019) Sentence-bert: Sentence embeddings using siamese bert-networks. arXiv preprint arXiv:1908.10084.","DOI":"10.18653\/v1\/D19-1410"},{"key":"key-10.2166\/wst.2025.156-32","unstructured":"Reimers N. & Gurevych I. (2020) The curse of dense low-dimensional information retrieval for large index sizes. arXiv preprint arXiv:2012.14210."},{"key":"key-10.2166\/wst.2025.156-33","unstructured":"Rongali S., Jagannatha A., Rawat B. P. S. & Yu H. (2020) Continual domain-tuning for pretrained language models. arXiv preprint arXiv:2004.02288."},{"key":"key-10.2166\/wst.2025.156-34","unstructured":"Ronzano F. & Nanavati J. (2024) Towards Ontology-Enhanced Representation Learning for Large Language Models. ArXiv, abs\/2405.20527. https:\/\/doi.org\/10.48550\/arXiv.2405.20527."},{"key":"key-10.2166\/wst.2025.156-35","first-page":"66","article-title":"Rhetorical structure of research article abstracts in environmental science","volume":"1","author":"Saeeaw","year":"2014","journal-title":"Thoughts"},{"key":"key-10.2166\/wst.2025.156-36","doi-asserted-by":"publisher","DOI":"10.1109\/ACCESS.2025.3554222"},{"key":"key-10.2166\/wst.2025.156-37","doi-asserted-by":"publisher","DOI":"10.1109\/ACCESS.2025.3548591"},{"key":"key-10.2166\/wst.2025.156-38","unstructured":"Sajja R., Sermet Y. & Demir I. (2025c) An Open-Source Dual-Loss Embedding Model for Semantic Retrieval in Higher Education. arXiv preprint arXiv:2505.04916."},{"key":"key-10.2166\/wst.2025.156-39","doi-asserted-by":"crossref","unstructured":"Sajja R., Xiong S., Mermer O., Sermet M. & Demir I. (2025d) A comprehensive bibliometric analysis of large language models in hydrology and environmental sciences. EarthArXiv (California Digital Library). https:\/\/doi.org\/10.31223\/x5sm61.","DOI":"10.31223\/X5SM61"},{"key":"key-10.2166\/wst.2025.156-40","first-page":"1","article-title":"Introduction to hydrology","author":"Salas","year":"2013"},{"key":"key-10.2166\/wst.2025.156-41","doi-asserted-by":"crossref","unstructured":"Samuel D. J., Sermet M. Y., Mount J., Vald G., Cwiertny D. & Demir I. (2024) Application of Large Language Models in Developing Conversational Agents for Water Quality Education, Communication and Operations. EarthArxiv, 7056. https:\/\/doi.org\/10.31223\/X5XT4K.","DOI":"10.31223\/X5XT4K"},{"key":"key-10.2166\/wst.2025.156-42","doi-asserted-by":"publisher","DOI":"10.1007\/s12145-019-00398-9"},{"key":"key-10.2166\/wst.2025.156-43","doi-asserted-by":"publisher","DOI":"10.3390\/bdcc5040057"},{"key":"key-10.2166\/wst.2025.156-44","doi-asserted-by":"publisher","DOI":"10.1111\/gwat.13195"},{"key":"key-10.2166\/wst.2025.156-45","unstructured":"Shen J., Tenenholtz N., Hall J. B., Alvarez-Melis D. & Fusi N. (2024) Tag-LLM: Repurposing General-Purpose LLMs for specialized domains. arXiv (Cornell University). https:\/\/doi.org\/10.48550\/arxiv.2402.05140."},{"key":"key-10.2166\/wst.2025.156-46","doi-asserted-by":"publisher","DOI":"10.1016\/j.envsci.2025.104055"},{"key":"key-10.2166\/wst.2025.156-47","unstructured":"Tang Y. & Yang Y. (2024) Do we need Domain-Specific embedding Models? an empirical investigation. arXiv (Cornell University). https:\/\/doi.org\/10.48550\/arxiv.2409.18511."},{"key":"key-10.2166\/wst.2025.156-48","unstructured":"Thakur N., Reimers N., R\u00fcckl\u00e9 A., Srivastava A. & Gurevych I. (2021) Beir: A heterogenous benchmark for zero-shot evaluation of information retrieval models. arXiv preprint arXiv:2104.08663."},{"key":"key-10.2166\/wst.2025.156-49","doi-asserted-by":"publisher","DOI":"10.1002\/2015WR017049"},{"key":"key-10.2166\/wst.2025.156-50","doi-asserted-by":"publisher","DOI":"10.1075\/term.21053.wan"},{"key":"key-10.2166\/wst.2025.156-51","doi-asserted-by":"publisher","DOI":"10.1016\/j.heliyon.2023.e21502"},{"key":"key-10.2166\/wst.2025.156-52","doi-asserted-by":"publisher","DOI":"10.1038\/s41586-024-08141-1"},{"key":"key-10.2166\/wst.2025.156-53","doi-asserted-by":"publisher","DOI":"10.3791\/66796"},{"key":"key-10.2166\/wst.2025.156-54","author":"Yadav","year":"2024"},{"key":"key-10.2166\/wst.2025.156-55","first-page":"127","article-title":"Mine water hydrology","volume-title":"Environmental Pollution","author":"Younger","year":"2002"},{"key":"key-10.2166\/wst.2025.156-56","first-page":"162","article-title":"Domain-specific contextualized embedding: a systematic literature review","author":"Yunianto","year":"2020"},{"key":"key-10.2166\/wst.2025.156-57","unstructured":"Zarlenga M. E., Barbiero P., Ciravegna G., Marra G., Giannini F., Diligenti M., Shams Z., Precioso F., Melacci S., Weller A., Lio P. & Jamnik M. (2022) Concept embedding models: Beyond the Accuracy-Explainability Trade-Off. arXiv (Cornell University). https:\/\/doi.org\/10.48550\/arxiv.2209.09056."},{"key":"key-10.2166\/wst.2025.156-58","doi-asserted-by":"crossref","unstructured":"Zhang Y., Feng Z., Teng Z., Liu Z. & Li H. (2023) How well do text embedding models understand syntax? arXiv (Cornell University). https:\/\/doi.org\/10.48550\/arxiv.2311.07996.","DOI":"10.18653\/v1\/2023.findings-emnlp.650"},{"key":"key-10.2166\/wst.2025.156-59","doi-asserted-by":"crossref","DOI":"10.1145\/3616855.3635688","article-title":"Domain level interpretability: interpreting Black-box model with domain-specific embedding","author":"Zhang","year":"2024"}],"container-title":["Water Science &amp; Technology"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/iwaponline.com\/wst\/article-pdf\/92\/9\/1328\/1598384\/wst2025156.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"syndication"},{"URL":"https:\/\/iwaponline.com\/wst\/article-pdf\/92\/9\/1328\/1598384\/wst2025156.pdf","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,11,14]],"date-time":"2025-11-14T13:48:03Z","timestamp":1763128083000},"score":1,"resource":{"primary":{"URL":"https:\/\/iwaponline.com\/wst\/article\/92\/9\/1328\/109925\/Domain-specific-embedding-models-for-hydrology-and"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,10,25]]},"references-count":59,"journal-issue":{"issue":"9","published-online":{"date-parts":[[2025,10,25]]},"published-print":{"date-parts":[[2025,11,1]]}},"URL":"https:\/\/doi.org\/10.2166\/wst.2025.156","relation":{},"ISSN":["0273-1223","1996-9732"],"issn-type":[{"value":"0273-1223","type":"print"},{"value":"1996-9732","type":"electronic"}],"subject":[],"published":{"date-parts":[[2025,10,25]]}}}