Linking hypolimnion to epilimnion in a stratified arctic lake: Machine learning-based estimation of hypolimnetic water quality from epilimnetic measurements
文献类型：期刊论文
| 作者 | Mahdian, Mehran2; Noori, Roohollah3; Saravani, Mohammad Javad4; Shahvaran, Ali Reza5; Shahmohammad, Mohsen6; Gaffney, Paul P. J.7; Salamattalab, Mohammad Milad4; Anboohi, Milad Shamsi8; Hosseinzadeh, Majid4; Xia, Fan1 |
| 刊名 | WATER RESEARCH
 |
| 出版日期 | 2026-04-15 |
| 卷号 | 294页码:125367 |
| 关键词 | Deep-water quality Hypolimnion Machine learning Deep learning Lake Inari Arctic lakes |
| ISSN号 | 0043-1354 |
| DOI | 10.1016/j.watres.2026.125367 |
| 产权排序 | 6 |
| 文献子类 | Article |
| 英文摘要 | Understanding water quality in the deeper layers of stratified lakes is critical, as these zones govern ecosystem stability and biodiversity health. The hypolimnion, the bottommost layer of a stratified lake, is characterized by limited vertical circulation and facilitates pollutant accumulation, thus serving as an indicator of long-term lake condition. While surface water quality can now be routinely monitored using in-situ sensors and satellite observations, assessing hypolimnetic conditions remains costly and logistically challenging, particularly in the Arctic regions with severe environmental conditions and logistical constraints. Using a long-term dataset spanning 1979-2022 collected at a single monitoring station where both epilimnetic and hypolimnetic profiles are measured, this study develops and compares five machine learning and deep learning models including artificial neural networks (ANN), random forest (RF), extreme gradient boosting, support vector regression, and Kolmogorov-Arnold networks to estimate hypolimnetic total nitrogen (TN), total phosphorus (TP), and dissolved oxygen (DO) in Lake Inari, Finland, based on readily available epilimnetic water quality predictors. Model performance was evaluated using five-fold cross-validation and assessed with the Nash-Sutcliffe efficiency (NSE), normalized mean absolute error (NMAE), and coefficient of determination (R2). For TN, the RF model achieved the best overall performance, with mean cross-validation values of NSE = 0.52, NMAE = 0.11, and R2 = 0.52, outperforming the other models, which yielded NSE values of 0.38-0.52, NMAE of 0.12-0.14, and R2 of NSE values of 0.17-0.46, NMAE of 0.13-0.18, and R2 of 0.32-0.51 for the remaining models. For DO, RF consistently outperformed all other approaches, achieving NSE = 0.76, NMAE = 0.09, and R2 = 0.77, whereas competing models produced NSE values of 0.62-0.72, NMAE of 0.10-0.11, and R2 of 0.68-0.73. Overall, all five models demonstrated their strongest performance for DO prediction. Permutation importance analysis revealed that surface TN, TP, and water temperature were key predictors for hypolimnetic TN, TP, and DO, respectively. This study demonstrates a practical, near-real-time, and cost-effective approach for assessing deep-water quality in stratified Arctic lakes, offering new opportunities for improved monitoring and management under growing climatic pressures. |
| URL标识 | 查看原文 |
| WOS关键词 | DISSOLVED-OXYGEN DYNAMICS ; INTERNAL PHOSPHORUS LOAD ; TOTAL NITROGEN ; SEDIMENTS ; PREDICTION ; SELECTION ; MODEL ; DIEL |
| WOS研究方向 | Engineering ; Environmental Sciences & Ecology ; Water Resources |
| 语种 | 英语 |
| WOS记录号 | WOS:001685201800001 |
| 出版者 | PERGAMON-ELSEVIER SCIENCE LTD |
| 源URL | [http://ir.igsnrr.ac.cn/handle/311030/220909]  |
| 专题 | 陆地水循环及地表过程院重点实验室_外文论文 |
| 通讯作者 | Noori, Roohollah |
| 作者单位 | 1.Yangzhou Univ, Coll Environm Sci & Engn, Yangzhou, Peoples R China; 2.Univ Eastern Finland, Dept Environm & Biol Sci, Kuopio, Finland; 3.Chinese Acad Sci, Nanjing Inst Geog & Limnol, State Key Lab Lake & Watershed Sci Water Secur, Nanjing 210008, Peoples R China; 4.Iran Univ Sci & Technol, Sch Civil Engn, Tehran, Iran; 5.Univ Waterloo, Dept Earth & Environm Sci, Ecohydrol Res Grp, Waterloo, ON N2L 3G1, Canada; 6.Univ Michigan, Sch Environm & Sustainabil, 440 Church St, Ann Arbor, MI 48109 USA; 7.Chinese Acad Sci, Inst Geog Sci & Nat Resources Res, Key Lab Water Cycle & Related Land Surface Proc, Beijing, Peoples R China; 8.Univ Oulu, Water Energy & Environm Engn Res Unit, FI-90014 Oulu, Finland; 9.Univ Chinese Acad Sci, Beijing 100049, Peoples R China; 10.Univ Chinese Acad Sci, Coll Nanjing, Nanjing 211135, Peoples R China; |
| 推荐引用方式 GB/T 7714 | Mahdian, Mehran,Noori, Roohollah,Saravani, Mohammad Javad,et al. Linking hypolimnion to epilimnion in a stratified arctic lake: Machine learning-based estimation of hypolimnetic water quality from epilimnetic measurements[J]. WATER RESEARCH,2026,294:125367. |
| APA | Mahdian, Mehran.,Noori, Roohollah.,Saravani, Mohammad Javad.,Shahvaran, Ali Reza.,Shahmohammad, Mohsen.,...&Abolfathi, Soroush.(2026).Linking hypolimnion to epilimnion in a stratified arctic lake: Machine learning-based estimation of hypolimnetic water quality from epilimnetic measurements.WATER RESEARCH,294,125367. |
| MLA | Mahdian, Mehran,et al."Linking hypolimnion to epilimnion in a stratified arctic lake: Machine learning-based estimation of hypolimnetic water quality from epilimnetic measurements".WATER RESEARCH 294(2026):125367. |
入库方式： OAI收割
来源：地理科学与资源研究所
浏览0
下载0
收藏0
其他版本
除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。