基于GBDT-LSTM的闽江流域水质预测

doi:10.16258/j.cnki.1674-5906.2024.04.010

生态环境学报 ›› 2024, Vol. 33 ›› Issue (4): 597-606.DOI: 10.16258/j.cnki.1674-5906.2024.04.010

• 研究论文【环境科学】 • 上一篇下一篇

基于GBDT-LSTM的闽江流域水质预测

肖扬岚¹^,²(), 沈惠柔¹^,², 许一涵¹^,², 尤添革¹^,²^,^*(), 郑艺婧¹^,², 谢候展¹^,², 宁静¹^,²

1.福建农林大学计算机与信息学院，福建福州 350002
2.福建省统计信息研究中心，福建福州 350002

收稿日期:2024-01-19 出版日期:2024-04-18 发布日期:2024-05-31
通讯作者: *尤添革。E-mail: 000q101026@fafu.edu.cn
作者简介:肖扬岚（2000年生），女，硕士研究生，主要从事水生态与环境研究。E-mail: 1221153030@fafu.edu.cn
基金资助:
福建省社会科学规划项目(FJ2018B063)

Water Quality Prediction of Min River Basin Based on GBDT-LSTM

XIAO Yanglan¹^,²(), SHEN Huirou¹^,², XU Yihan¹^,², YOU Tiange¹^,²^,^*(), ZHENG Yijing¹^,², XIE Houzhan¹^,², NING Jing¹^,²

1. College of Computer and Information, University of Fujian Agriculture and Forestry, Fuzhou 350002, P. R. China
2. Fujian Statistical Information Research Center, Fuzhou 350002, P. R. China

Received:2024-01-19 Online:2024-04-18 Published:2024-05-31

摘要/Abstract

摘要：

闽江作为福建省最大的河流，其生态保护对维护全省水生态环境而言意义重大。为进一步利用水质评价和预测方法对水质状况进行有效分析，选取闽江流域2017年1月－2023年8月共20处水质监测站数据作为研究对象，采用水质综合指数法对各站点水质状态进行判断；比较LSTM模型和GBDT-LSTM模型的拟合结果，并对各指标数据进行预测。结果表明，1）闽江流域各监测点除总氮外的水质指标均呈现较好趋势，各监测点的总氮浓度存在差异，整体情况较差，其中位于三明和南平市的河段由于当地重工业发达，导致水体中的硝态氮不断增加，进而致使河段内总氮浓度过高。2）流域水质的WQI值呈现逐年上升的趋势，水质状况普遍处于中等及以上水平，仅有少数监测点的水质状况处于很差状态，从中上游携带的泥沙等固体的堆积导致了连江琯头的水质情况较差；沙县斑竹溪渡口由于位于三明和南平的交界处，且沿岸分布较多重工业城市，故水质状态略差；相较于南平和三明的大型重工业企业，位于古田县的监测点以农业、轻工业为主，对水质指标的影响相对较小，水质状况因此较好。3）采用GBDT对变异程度较高的各水质指标在预测模型中的重要性进行排序，发现GBDT-LSTM混合模型的拟合效果相较于LSTM模型更好，更有利于对水质状况进行精确地预测。4）水体中总氮、高锰酸盐指数等含量的不断增加主要源于大量的工业废水，建议加强对临近闽江流域高污染高排放企业的控制，科学合理地实现工业污染排放和污染物容量在时空上的合理分配。

关键词: 闽江流域, 水质预测, 水质综合指数, 长短期记忆神经网络, 梯度提升树, 水质评价

Abstract:

As the largest river in Fujian Province, the ecological protection of the Min River is of great significance to the maintenance of the province's water ecosystem. In order to further utilize the water quality evaluation and prediction methods to effectively analyze the water quality status, data from a total of 20 water quality monitoring stations in the Min River Basin from January 2017 to August 2023 were selected as the research object. The water quality composite index method was used to judge the status of water quality at each station. The fitting results of the LSTM model and GBDT-LSTM model were compared, and the prediction of the data of each index was carried out. The results showed that: 1) all water quality indexes, except for total nitrogen, showed improvement at each monitoring site in the Min River Basin. However, there were differences in total nitrogen concentrations across monitoring sites. The overall situation was poor, particularly in river sections located in Sanming and Nanping cities, where the development of local heavy industry had led to a continuous increase in nitrate nitrogen levels and high concentrations of total nitrogen. 2) The WQI value of water quality in the basin showed an increasing trend year by year, with most sites maintaining medium or better water quality. Only a few monitoring points exhibited very poor water quality. The accumulation of solids, such as sediment carried from the middle and upper reaches of the river, contributed to the poor water quality of Guantou in Lianjiang. The Banzhuxi Ferry in Shaxian also showed slightly poor water quality due to the influence of nearby heavy industrial cities. In contrast, monitoring sites in Gutian, which were influenced more by agriculture and light industries, displayed better water quality conditions. 3) The GBDT was used to rank the importance of various water quality indicators with high degree of variation in the prediction model. The fitting effect of the hybrid model of GBDT-LSTM was better than that of the LSTM model, aiding in more accurate water quality predictions. 4) The continuous increase in total nitrogen and permanganate index in water was mainly due to a large amount of industrial wastewater. It is recommended to strengthen the control of high-pollution and high-emission enterprises near the Min River Basin. Additionally, there should be a scientifically and reasonably timed and spatially distributed approach to industrial pollution emission and pollutant capacity management.

Key words: Min River Basin, Water quality prediction, water quality composite index, LSTM neural network, gradient boosting tree, water quality assessment

中图分类号:

X52
X832

肖扬岚, 沈惠柔, 许一涵, 尤添革, 郑艺婧, 谢候展, 宁静. 基于GBDT-LSTM的闽江流域水质预测[J]. 生态环境学报, 2024, 33(4): 597-606.

XIAO Yanglan, SHEN Huirou, XU Yihan, YOU Tiange, ZHENG Yijing, XIE Houzhan, NING Jing. Water Quality Prediction of Min River Basin Based on GBDT-LSTM[J]. Ecology and Environment, 2024, 33(4): 597-606.

图/表 11

参考文献 31

[1]	GUAN G L, WANG Y G, YANG L, et al., 2022. Water-Quality assessment and pollution-risk early-warning system based on Web Crawler Technology and LSTM[J]. International Journal of Environmental Research and Public Health, 19(18): 11818.
[2]	HE T, ZHANG L G, ZENG Y, et al., 2012. Water quality comprehensive index method of Eltrix River in Xin Jiang Province using SPSS[J]. Procedia Earth and Planetary Science, 5: 314-321.
[3]	LI X, PENG L, YAO X J, et al., 2017. Long short-termmemory neural network for air pollutant concentration predictions: Method development and evaluation[J]. Environmental Pollution, 231(Part 1): 997-1004.
[4]	SONG Q X, ZOU J, XU M, et al., 2023. Air quality prediction for Chengdu based on long short-term memory neural network with improved jellyfish search optimizer[J]. Environmental Science and Pollution Research International, 30(23): 64416-64442.
[5]	程学宁, 2018. 基于时空数据分析的闽江流域水环境质量评价研究[D]. 福州: 福州大学.
	CHENG X N, 2018. Study on water environment quality evaluation of Minjiang River based on spatial and temporal data analysis[D]. Fuzhou: Fuzhou University.
[6]	褚飞飞, 2022. 基于注意力机制和长短期记忆神经网络的水质预测方法研究[D]. 南京: 南京邮电大学.
	CHU F F, 2022. Research of water quality prediction method based on attention mechanism and long short term memory neural network[D]. Nanjing: Nanjing University of Posts and Telecommunications.
[7]	程诺, 高文晶, 杨鸿飞, 等, 2023. 闽江流域生态环境质量动态变化及其驱动力[J]. 水土保持通报, 43(2): 310-318, 329.
	CHENG N, GAO W J, YANG H F, et al., 2023. Dynamic change and driving factors of ecological environment quality in Minjiang River Basin[J]. Bulletin of Soil and Water Conservation, 43(2): 310-318, 329.
[8]	付博超, 2022. 基于内梅罗综合污染指数法的秦皇岛水质评价[J]. 水利科学与寒区工程, 5(11): 32-34.
	FU B C, 2022. Water quality evaluation of Qinhuangdao based on the Nemero comprehensive pollution index method[J]. Hydro Science and Cold Zone Engineering, 5(11): 32-34.
[9]	谷宇峰, 张道勇, 鲍志东, 等, 2021. 利用梯度提升决策树 (GBDT) 预测渗透率——以姬塬油田西部长4+5段致密砂岩储层为例[J]. 地球物理学进展, 36(2): 585-594.
	GU Y F, ZHANG D Y, BAO Z D, et al., 2021. Permeability prediction using gradient boosting decision tree (GBDT): A case study of tight sandstone reservoirs of member of chang 4+5 in western Jiyuan Oilfield[J]. Progress in Geophysics, 36(2): 585-594.
[10]	郭利进, 许瑞伟, 2022. 基于改进果蝇算法的LSTM在水质预测中的应用[J]. 长江科学院院报, 40(8): 57-63.
	GUO L J, XU R W, 2022. Application of LSTM model combining improved Fruit-Fly Algorithmafter Seasonal-Trend Decomposition using LOESS to water quality prediction[J]. Journal of the Yangtze River Academy of Sciences, 40(8): 57-63.
[11]	耿凯利, 陈兴伟, 郑美玲, 等, 2024. 闽江流域多时间尺度降雨—径流关系变化与成因[J]. 中国水土保持科学, 22(1): 35-41.
	GENG K L, CHEN X W, ZHENG M L, et al., 2024. Variations and causes of rainfall-runoff relationship at different time scales in Minjiang River Basin[J]. Science of Soil and Water Conservation, 22(1): 35-41.
[12]	郝玉莹, 赵林, 孙同, 等, 2021. 基于RF-LSTM的地表水体水质预测[J]. 水资源与水工程学报, 32(6): 41-48.
	HAO Y Y, ZHAO L, SUN T, et al., 2021. Surface water quality prediction based on RF-LSTM[J]. Journal of Water Resources and Water Engineering, 32(6): 41-48.
[13]	宦娟, 李慧, 李明宝, 等, 2021. 基于交叉验证网格寻优的GBDT-LSTM水产养殖溶解氧预测[J]. 常州大学学报(自然科学版), 33(4): 63-71.
	HUAN J, LI H, LI M B, et al., 2021. Prediction of dissolved Oxygen in GBDT-LSTM aquaculture based on Cross-Validation Grid Optimization[J]. Journal of Changzhou University: Natural Science, 33(4): 63-71.
[14]	郝萌, 闫祯, 黄一恒, 等, 2023. 延安市集中式饮用水水源地水质评价[J]. 农业现代化研究, 44(4): 736-744.
	HAO M, YAN Z, HUANG Y H, et al., 2023. Water quality evaluation of centralized drinking water sources in Yan’an City[J]. Research on Agricultural Modernization, 44(4): 736-744.
[15]	靳聪聪, 区舵样, 周晗宇, 等, 2021. 基于水质指数法的城市河流水质综合评价研究[J]. 水利与建筑工程学报, 19(2): 240-245.
	JIN C C, OU T Y, ZHOU H Y, et al., 2021. Comprehensive city river water quality assessment based on water quality identification index method[J]. Journal of Water Resources and Architectural Engineering, 19(2): 240-245.
[16]	陆建忠, 崔肖林, 陈晓玲, 2015. 基于综合指数法的鄱阳湖流域水资源安全评价研究[J]. 长江流域资源与环境, 24(2): 212-218.
	LU J Z, CUI X L, CHEN X X, 2015. Assessment of water resource security in Poyang Lake watershed based on composite index method[J]. Yangtze River Basin Resources and Environment, 24(2): 212-218.
[17]	李余隆, 张兰, 李立, 2023. 基于GCN-LSTM的钱塘江南源水质预测研究[J]. 人民黄河, 45(12): 83-87, 95.
	LI Y L, ZHANG L, LI L, 2023. Study on water quality prediction of South Source of Qiantang River based on GCN-LSTM[J]. Yellow River, 45(12): 83-87, 95.
[18]	秦文虎, 陈溪莹, 2020. 基于长短时记忆网络的水质预测模型研究[J]. 安全与环境学报, 20(1): 328-334.
	QIN W H, CHEN X Y, 2020. Water quality forecast and prediction model based on long- and short-term memory network[J]. Journal of Safety and Environment, 20(1): 328-334.
[19]	宋治岑, 张顺平, 卢敏, 2023. 基于HHO-SVM的水质预测模型及应用[J]. 水电能源科学, 41(8): 70-72, 47.
	SONG Z C, ZHANG S P, LU M, 2023. Water quality prediction model based on HHO-SVM and its application[J]. Water Resources and Power, 41(8): 70-72, 47.
[20]	汤云, 2020. 闽江流域水质时空分布特征分析及水质预测[D]. 福州: 福州大学.
	TANG Y, 2020. Spatio-temporal distribution characteristics analysis and water quality prediction in Minjiang River Basin[D]. Fuzhou: Fuzhou University.
[21]	陶伟, 王乃亮, 魏婧, 2021. 改进的综合水质标识指数法对湟水河红古段的水质时空特征分析[J]. 甘肃科学学报, 33(6): 97-102.
	TAO Wei, WANG N L, WEI J, 2021. Analysis of temporal and spatial characterization of improved comprehensive water quality identification index method in water quality assessment of Huangshui River in Honggu Section[J]. Gansu Science Journal, 33(6): 97-102.
[22]	魏天宇, 2016. 我国水污染的体制成因及对策[J]. 黑龙江水利科技 (3): 59-63.
	WEI T Y, 2016. Institutional causes of water pollution in China and countermeasures[J]. Heilongjiang Water Science and Technology (3): 59-63.
[23]	王守相, 刘天宇, 2017. 计及用电模式的居民负荷梯度提升树分类识别方法[J]. 电力系统及其自动化学报, 29(9): 27-33.
	WANG S X, LIU T Y, 2017. Gradient Boosting Decision Tree method for power consumption modes[J]. Journal of Power Residential Load Classification Considering Typical Systems and Automation, 29(9): 27-33.
[24]	王春晓, 卢毅敏, 2020. 闽江流域水质时间序列变化趋势识别及特征分析[J]. 水资源与水工程学报, 31(4): 63-69.
	WANG C X, LU Y M, 2020. Trend recognition and characteristics analysis of water quality indicator time series in Minjiang River Basin[J]. Journal of Power Systems and Automation, 31(4): 63-69.
[25]	王志勃, 姜仲秋, 张天舒, 2022. 基于CNN-LSTM的白马湖水质预测模型研究[J]. 电脑知识与技术, 18(26): 11-13.
	WANG Z B, JIANG Z Q, ZHANG T S, 2022. Research on CNN-LSTM based water quality prediction model of Lake Baima[J]. Computer Knowledge and Technology, 18(26): 11-13.
[26]	王渤权, 金传鑫, 周论, 等, 2023. 基于长短期记忆神经网络的西丽水库水质预测研究[J/OL]. 长江科学院院报, 40(6): 64-70.
	WANG B Q, JIN C X, ZHOU R, et al., 2023. Water quality prediction for Xili Reservoir based on Long-Short Term Memory[J/OL]. Journal of the Yangtze River Academy of Sciences, 40(6): 64-70.
[27]	吴月芳, 胡明昕, 孙培莉, 2023. 基于注意力机制及长短期记忆神经网络的慢性阻塞性肺疾病氧减状态辨识[J]. 南京理工大学学报, 47(5): 629-635.
	WU Y F, HU M X, SUN P L, 2023. Oxygen depletion status identification in COPD by using attention-based LSTM[J]. Journal of Nanjing University of Science and Technology, 47(5): 629-635.
[28]	叶焰中, 陈凡, 黄廷林, 2022. 基于水质综合指数法的亚热带水源型水库水质评价[J]. 水资源保护, 38(6): 116-124.
	YE Y Z, CHEN F, HUANG T L, 2022. Water quality evaluation of subtropical water source reservoir using water quality index method[J]. Water Resource Protection, 38(6): 116-124.
[29]	闫晓, 2023. 小流域水污染与水环境问题成因及对策分析[J]. 资源节约与环保 (9): 13-16.
	YAN X, 2023. Analysis of the causes and countermeasures of water pollution and water environment problems in small watersheds[J]. Resource Conservation and Environmental Protection (9): 13-16.
[30]	赵天旸, 石成春, 谢蓉蓉, 等, 2022. 闽江流域近年主要水质指标变化趋势及污染识别[J]. 渔业研究, 44(4): 324-335. DOI
	ZHAO T Y, SHI C C, XIE R R, et al., 2022. Change trend of main water quality indices and pollution identification in Minjiang River Basin in recent years[J]. Journal of Power Systems and Automation, 44(4): 324-335.
[31]	张云琳, 2023. 2002-2021年闽江水质变化及其影响因素分析[J]. 绿色科技, 25(22): 123-126.
	ZHANG Y L, 2023. Analysis of water quality changes and its influencing factors in the Minjiang River Basin from 2002 to 2021[J]. Journal of Green Science and Technology, 25(22): 123-126.

监测点	序号	所属河流	监测点	序号	所属河流	监测点	序号	所属河流	监测点	序号	所属河流
福州文山里	A	闽江下游	建瓯蓬墩	F	崇阳溪	延平浪石	K	富屯溪	古田水库	P	古田溪
连江琯头	B	闽江下游	坪洲大桥	G	南浦溪	延平南溪	L	尤溪	大田高才	Q	均溪
闽侯竹岐	C	闽江下游	将乐樟应	H	金溪	延平洋坑	M	闽江上游	建宁袁庄	R	濉溪
闽清雄江	D	闽江下游	南平水汾桥	I	沙溪	政和西津	N	松溪	斑竹渡溪口	S	沙溪
建瓯房村	E	建溪	武夷山兴田	J	崇阳溪	古田黄田	O	闽江中游	安砂水库下游	T	九龙溪

监测点	序号	所属河流	监测点	序号	所属河流	监测点	序号	所属河流	监测点	序号	所属河流
福州文山里	A	闽江下游	建瓯蓬墩	F	崇阳溪	延平浪石	K	富屯溪	古田水库	P	古田溪
连江琯头	B	闽江下游	坪洲大桥	G	南浦溪	延平南溪	L	尤溪	大田高才	Q	均溪
闽侯竹岐	C	闽江下游	将乐樟应	H	金溪	延平洋坑	M	闽江上游	建宁袁庄	R	濉溪
闽清雄江	D	闽江下游	南平水汾桥	I	沙溪	政和西津	N	松溪	斑竹渡溪口	S	沙溪
建瓯房村	E	建溪	武夷山兴田	J	崇阳溪	古田黄田	O	闽江中游	安砂水库下游	T	九龙溪

指标		单位	指标代码	指标		单位	指标代码
自然	水温	℃	x₁	社会	人均GDP	Yuan (RMB)	x₁₉
	pH值	－	x₂		常住人口	10⁴person	x₂₀
	溶解氧	mg·L^-1	x₃		人口密度	10⁴person ·km^-2	x₂₁
	电导率	µS·cm^-1	x₄		城市污水排放量	10⁴m³	x₂₂
	浊度	NTU	x₅		城镇生活废水中氨氮排放量	t	x₂₃
	高锰酸盐指数	mg·L^-1	x₆		工业废水中氨氮排放量	t	x₂₄
	氨氮	mg·L^-1	x₇		工业废水排放量	t	x₂₅
	总磷	mg·L^-1	x₈		农药使用量	t	x₂₆
	总氮	mg·L^-1	x₉		农用化肥施用量	t	x₂₇
	降水量	mg·L^-1	x₁₀		农用氮肥施用量	t	x₂₈
	平均气温	℃	x₁₁		农用磷肥施用量	t	x₂₉
	平均湿度	%	x₁₂		第一产业比重	%	x₃₀
	PM2.5	µg·m^-3	x₁₃		第二产业比重	%	x₃₁
	PM10	µg·m^-3	x₁₄		第三产业比重	%	x₃₂
	NO₂	µg·m^-3	x₁₅		城市污水处理厂集中处理率	%	x₃₃
	CO	µg·m^-3	x₁₆		生活垃圾无害化处理率	%	x₃₄
	SO₂	µg·m^-3	x₁₇		水土保持总投资	10⁴ Yuan (RMB)	x₃₅
	O₃	µg·m^-3	x₁₈		减少水土流失量	10⁴ t	x₃₆

指标		单位	指标代码	指标		单位	指标代码
自然	水温	℃	x₁	社会	人均GDP	Yuan (RMB)	x₁₉
	pH值	－	x₂		常住人口	10⁴person	x₂₀
	溶解氧	mg·L^-1	x₃		人口密度	10⁴person ·km^-2	x₂₁
	电导率	µS·cm^-1	x₄		城市污水排放量	10⁴m³	x₂₂
	浊度	NTU	x₅		城镇生活废水中氨氮排放量	t	x₂₃
	高锰酸盐指数	mg·L^-1	x₆		工业废水中氨氮排放量	t	x₂₄
	氨氮	mg·L^-1	x₇		工业废水排放量	t	x₂₅
	总磷	mg·L^-1	x₈		农药使用量	t	x₂₆
	总氮	mg·L^-1	x₉		农用化肥施用量	t	x₂₇
	降水量	mg·L^-1	x₁₀		农用氮肥施用量	t	x₂₈
	平均气温	℃	x₁₁		农用磷肥施用量	t	x₂₉
	平均湿度	%	x₁₂		第一产业比重	%	x₃₀
	PM2.5	µg·m^-3	x₁₃		第二产业比重	%	x₃₁
	PM10	µg·m^-3	x₁₄		第三产业比重	%	x₃₂
	NO₂	µg·m^-3	x₁₅		城市污水处理厂集中处理率	%	x₃₃
	CO	µg·m^-3	x₁₆		生活垃圾无害化处理率	%	x₃₄
	SO₂	µg·m^-3	x₁₇		水土保持总投资	10⁴ Yuan (RMB)	x₃₅
	O₃	µg·m^-3	x₁₈		减少水土流失量	10⁴ t	x₃₆

指标代码	x₃	x₄	x₅	x₆	x₇	x₈	x₉	x₁₀	x₁₇	x₁₈	WQI
x₃	1	2×10^-3	6.5×10^-2	7.2×10^-2	-5.2×10^-2	9.8×10^-2	-0.110	0.268^**	3×10^-2	-8.7×10^-2	0.198^*
x₄	2×10^-3	1	0.862^**	0.522^**	0.105	0.335^**	0.703^**	-0.281^**	-0.234^**	0.275^**	-0.647^**
x₅	6.5×10^-2	0.862^**	1	0.582^**	0.145	0.434^**	0.593^**	-0.147	-0.231^**	0.239^**	-0.689^**
x₆	7.2×10^-2	0.522^**	0.582^**	1	0.136	0.259^**	0.343^**	-6.4×10^-2	-1.5×10^-2	0.141	-0.530^**
x₇	-5.2×10^-2	0.105	0.145	0.136	1	0.428^**	0.375^**	0.149	0.248^**	-0.198^*	-0.465^**
x₈	9.8×10^-2	0.335^**	0.434^**	0.259^**	0.428^**	1	0.625^**	5×10^-2	-0.170^*	0.108	-0.661^**
x₉	-0.110	0.703^**	0.593^**	0.343^**	0.375^**	0.625^**	1	-0.332^**	-0.198^*	0.208^*	-0.808^**
x₁₀	0.268^{** 2)}	-0.281^**	-0.147	-6.4×10^-2	0.149	5×10^-2	-0.332^**	1	0.220^**	-0.242^**	0.215^*
x₁₇	3×10^-2	-0.234^**	-0.231^**	-1.5×10^-2	0.248^**	-0.170^*	-0.198^*	0.220^**	1	-0.393^**	0.184^*
x₁₈	-8.7×10^-2	0.275^**	0.239^**	0.141	-0.198^*	0.108	0.208^*	-0.242^**	-0.393^**	1	-0.196^*
WQI	0.198^{* 1)}	-0.647^**	-0.689^**	-0.530^**	-0.465^**	-0.661^**	-0.808^**	0.215^*	0.184^*	-0.196^*	1

基于GBDT-LSTM的闽江流域水质预测

Water Quality Prediction of Min River Basin Based on GBDT-LSTM

RichHTML

PDF

可视化

摘要/Abstract

引用本文

使用本文

图/表 11

参考文献 31

相关文章 3

编辑推荐

Metrics

本文评价

指标代码	x₂₀	x₂₁	x₂₂	x₂₃	x₂₅	x₃₀	x₃₁	x₃₂	x₃₃	x₃₅	WQI
x₂₀	1	0.995^**	0.993^**	0.683^**	-0.656^**	-0.856^**	-0.268^**	0.734^**	0.276^**	0.717^**	-0.391^**
x₂₁	0.995^**	1	0.982^**	0.693^**	0.615^**	-0.883^**	-0.195^*	0.681^**	0.225^**	0.696^**	-0.365^**
x₂₂	0.993^**	0.982^**	1	0.621^**	0.716^**	-0.860^**	-0.283^**	0.752^**	0.280^**	0.742^**	-0.390^**
x₂₃	0.683^**	0.693^**	0.621^**	1	0.274^**	-0.626^**	7.2×10^-2	0.277^**	-7.5×10^-2	0.159	-0.254^**
x₂₅	-0.656^**	0.615^**	0.716^**	0.274^**	1	-0.678^**	-0.183^*	0.566^**	2.4×10^-2	0.549^**	-0.268^**
x₃₀	-0.856^**	-0.883^**	-0.860^**	-0.626^**	0.678^**	1	-0.179^**	-0.397^**	0.186^*	-0.537^**	0.236^**
x₃₁	-0.268^**	-0.195^*	-0.283^**	7.2×10^-2	-0.183^*	-0.179^**	1	-0.830^**	-0.903^**	-0.501^**	0.305^**
x₃₂	0.734^**	0.681^**	0.752^**	0.277^**	0.566^**	-0.397^**	-0.830^**	1	0.741^**	0.770^**	-0.417^**
x₃₃	0.276^**	0.225^**	0.280^**	-7.5×10^-2	2.4×10^-2	0.186^*	-0.903^**	0.741^**	1	0.504^**	-0.280^**
x₃₅	0.717^**	0.696^**	0.742^**	0.159	0.549^**	-0.537^**	-0.501^**	0.770^**	0.504^**	1	-0.354^**
WQI	-0.391^**	-0.365^**	-0.390^**	-0.254^**	-0.268^**	0.236^**	0.305^**	-0.417^**	-0.280^**	-0.354^**	1

水质指标	LSTM			GBDT-LSTM
水质指标	MAE	RMSE	r²	MAE	RMSE	r²
高锰酸盐指数	5.4×10^-2	0.121	0.869	3.8×10^-2	9.1×10^-2	0.899
氨氮	7×10^-3	1.7×10^-2	0.887	4×10^-3	1×10^-2	0.917
总磷	3×10^-3	7×10^-3	0.879	2×10^-3	5×10^-3	0.912
总氮	5.39×10^-2	0.128	0.867	3.2×10^-2	7.86×10^-2	0.914

[1]	鲁言波, 陈湛峰, 李晓芳. 基于粒子群优化的GRU广东省跨境断面水质预测模型研究[J]. 生态环境学报, 2023, 32(9): 1673-1681.
[2]	王源哲, 华春林, 赵丽, 樊敏, 梁晓盈, 周乐乐, 蔡璨, 姚婧. 山地城市主要河流水质评价及预测研究——以四川省绵阳市为例[J]. 生态环境学报, 2023, 32(8): 1465-1477.
[3]	程鹏, 孙明东, 郝韶楠. 基于最简水质综合评价指数的官厅水库上游河流水质评价[J]. 生态环境学报, 2023, 32(2): 372-380.