TTHHO-LSSVM-ABKDE Model-driven Collaborative Interval Prediction of Multi-source Drought Indices in Yellow River Basin: Ordos Reach in Middle Yellow River
Institute of Pastoral Water Resources Science, Ministry of Water Resources, Hohhot 010020, China
Show less
文章历史+
Received
Published
Revised
2025-10-09
2026-08-18
2025-12-17
Online Date
2026-09-04
PDF (23854K)
摘要
植被作为全球碳循环的关键组分,准确预测植被状况对于理解全球碳平衡机制具有重要科学意义。然而,传统植被指数预测存在明显局限性,无法量化预测结果的不确定性。以黄河中游鄂尔多斯段为研究区域,通过瞬时三角函数哈里斯鹰优化器(transient trigonometric-function Harris hawks optimization, TTHHO)实现参数空间的智能搜索与自适应优化,结合最小二乘支持向量机(least squares support vector machine, LSSVM)构建非线性回归模型TLA(TTHHO-LSSVM-ABKDE),并引入自适应带宽核密度估计(adaptive bandwidth kernel density estimation, ABKDE)进行概率分布建模,最终实现植被干旱指数的概率区间预测。研究结果如下:①TLA模型展现出显著的性能优势,其预测精度较长期短期记忆网络(long short-term memory, LSTM)和Transformer等传统模型提高5.10%~16.87%(p<0.05);②对5%~95%置信区间的评估指标分析显示,预测区间归一化平均宽度(prediction interval normalized average width, PINAW)值显著降低,预测区间覆盖率(prediction interval coverage probability, PICP)有所下降,覆盖宽度系数(coverage width criterion, CWC)随置信度提升而增加,连续排序概率评分(continuous ranked probability score, CRPS)保持稳定;③尽管多目标TLA模型的整体预测精度较单目标模型出现5.32%~7.44%的适度下降,但其仍能保持较高的预测准确度(R2>0.84)。通过提出TLA多目标区间预测模型,为区域尺度植被干旱监测提供了科学方法,进一步推动全球生态系统管理预测系统的开发。
Abstract
Vegetation is established as an essential element of the global carbon cycle, playing a significant role in regulating the Earth’s carbon balance. Therefore, the accurate forecasting of vegetation changes is recognized as crucial for understanding global carbon balancing mechanisms. However, traditional point forecasting methods are limited by their failure to account for the uncertainty inherent in prediction outcomes. To address this issue, the Ordos section of the middle reaches of the Yellow River was selected as the study area. A trigonometric-function Harris hawks optimization (TTHHO) method was proposed to facilitate intelligent parameter-space exploration and adaptive optimization. Subsequently, the optimized parameters were integrated into a least squares support vector machine (LSSVM) to construct a nonlinear regression model, denoted as TLA. Additionally, probabilistic interval prediction of the vegetation drought index (VDI) was achieved through the employment of an adaptive bandwidth kernel density estimation (ABKDE) method. Experimental results demonstrate follows.Substantial performance improvements are achieved by the proposed TLA model, with prediction accuracies found to be 5.10%~16.87% higher than comparative models such as LSTM and Transformer (p < 0.05).An increase in the coverage width criterion (CWC) is observed as the confi-dence level decreases (from 95% to 5%).And high predictive accuracy (R2 > 0.84) is maintained by the multi-objective TLA model, despite a moderate decrease in overall accuracy compared to the single-objective model. Consequently, a multi-objective TLA interval prediction framework is proposed, providing a rigorous scientific method for monitoring regional-scale vegetation dryness and contrib-uting to the development of predictive systems for global ecosystem management.
ZHENG He-xiang, WANG Ying, HOU Hong-fei, QIN Zi-yuan.
TTHHO-LSSVM-ABKDE Model-driven Collaborative Interval Prediction of Multi-source Drought Indices in Yellow River Basin: Ordos Reach in Middle Yellow River[J].
Science Technology and Engineering, 2026, 26(23): 9808-9820 DOI:10.12404/j.issn.1671-1815.2507085
Yang等[15]指出,在处理非平稳性或高维数据时,传统点预测方法会忽视预测结果的不确定性,单一预测值难以充分体现数据的真实波动。区间概率预测的提出,能以置信区间的形式有效展示预测结果的不确定性。与传统点预测不同,区间概率预测不仅输出单一预测值,还能量化预测结果的不确定性,提供预测值所在范围及其对应概率,从而更全面地反映数据可能的变化区间[16]。Liu等[17]通过多种机器学习模型的对比研究表明,区间预测模型在低频监测场景下具有更优的适用性。该模型通过引入不确定性量化机制,克服了传统点预测的局限,能够动态捕捉并反映数据的波动特征。自适应带宽核密度估计(adaptive bandwidth kernel density estimation, ABKDE)为估计预测误差的概率密度分布提供了新方法,其优势在于能够精确捕捉数据中的不稳定性[18]。当数据中存在异常值时,区间预测通过容忍数据波动、提供较宽泛的预测范围,减少了对单一值的过度依赖,从而增强了模型的鲁棒性与稳定性。Liu等[19]构建了基于ABKDE与深度学习的耦合模型,不仅实现了高精度的区间预测,且在应对异常数据干扰时,展现出比现有基线模型更强的鲁棒性与稳定性。但ABKDE在面对高维空间时,由于数据特征复杂且相互关联,带宽参数的选择变得尤为重要。然而超参数优化过程通常复杂且计算量大,人工调参还容易陷入局部最优解。因此Abdulrab等[20]基于经典哈里斯鹰优化算法(Harris hawks optimization,HHO)提出了一种新的算法——瞬时三角函数哈里斯鹰优化器(transient trigonometric-function Harris hawks optimization,TTHHO),为高维数据的优化问题提供了解决方案。该算法能够自适应调整搜索策略,避免局部最优解,并逐步引导模型寻找全局最优解,从而提升优化效率。虽然Zhang等[21]提出的融合梯度信息的自适应矩估计算法通过引入一阶动量和二阶自适应步长,在强噪声环境下显著优于传统算法,但该方法仍受限于基于梯度的迭代优化器的缺陷,如在多频信号高度非线性的多峰误差曲面上,依然面临陷入局部最优的风险。而群体智能算法TTHHO相比梯度优化器,能够在非凸优化、高维和复杂约束环境下有效避免局部最优,较好地探索全局最优解;其次针对多峰值问题,TTHHO在搜索过程中能同时探测到多个候选最优点,避免传统梯度方法仅停靠在单个局部最优解的状况[22];再次,TTHHO应用于多维预测时能有效精细调控最小二乘支持向量机(least squares support vector machine,LSSVM)和ABKDE的核心参数,从而缓解和避免高维预测数据分布复杂化引起的“维数灾难”,确保模型能把握潜在数据规律和关联。此外,文献[23]表明,ABKDE与机器学习模型LSSVM的结合,通过平滑分布估计,能够更好地处理数据中的异常值和噪声,减少数据干扰,并进一步提升模型的鲁棒性。TTHHO的结合则能够有效降低异常值对训练和预测过程的影响,进而提升模型的预测准确性。在植被干旱多因子预测任务中,ABKDE基于概率密度估计的区间预测方法,能够有效反映数据变化的范围及其不确定性[24]。尽管ABKDE在多目标区间概率预测中展现了显著潜力,但在面对高维数据和异常值预测任务时仍需进一步优化。
在全球气候变暖和干旱加剧的背景下,植被干旱指数可用于政府开展干旱压力缓解监测以及水资源的合理规划与利用。然而,传统点预测模式在处理非线性数据和应对极端气象状况时,其预测准确性仍存在局限。为此,提出一种基于TTHHO算法、LSSVM与ABKDE的耦合方法,通过多源数据长时间序列植被干旱参数,包括TVDI、TVPDI、ESI、DSI、植被供水指数(vegetation supply water index,VSWI)和TCI,构建具备多目标特性的深度学习模型——TLA(TTHHO-LSSVM-ABKDE),以捕捉非线性指数的变化及其置信区间的动态特征,进而弥补传统机器学习方法在区间概率估算方面的不足。
LiuYihao, DingYuanyuan, ZhaoJianyun, et al. Impact of extreme climate change on ecosystem of the Yellow River Source Area[J]. Science Technology and Engineering, 2024, 24(26): 11167-11179.
YinBensu, LiZhenfa, YueRong, et al. Monitoring drought in Guanzhong areas using temperature-vegetation drought index[J]. Transactions of the Chinese Society of Agricultural Engineering, 2024, 40(17): 111-119.
LiWenqing, ZhaoYong, LiuZhao, et al. Characteristics of agricultural drought and its influencing factors in the loess tableland of the north of the Wei River from 2001—2020[J]. Journal of Water Resources and Water Engineering, 2024, 35(5): 1-10, 19.
ZhaoYanyan, ZhangXuan, XuYang, et al. Study on the propagation thresholds from meteorological drought to agricultural drought in arid and semi-arid region[J]. Research of Soil and Water Conservation, 2025, 32(2): 234-243.
FanHu, DangXinghai, ZhaoJianyun, et al. Temporal and spatial variation of drought in northwest arid region based on temperature vegetation drought index[J]. Science Technology and Engineering, 2024, 24(27): 11537-11546.
LiGuoqing, ZhangChunkang, ZhangXianyun, et al. Response of kernel temperature vegetation drought index to urbanization in Northeast China[J]. Chinese Journal of Applied Ecology, 2025, 36(1): 208-218.
LiHuawei, ZhuXiaochun, ZhangXudong, et al. Retrieval of surface soil moisture at field scale in oasis irrigation area based on multi-source data[J]. Water Saving Irrigation, 2024(6): 19-26.
ZhangWeijie, WangWenjun, YinHang, et al. Relationship between hydrologic drought and meteorological drought in arid desert grassland basins[J]. Science Technology and Engineering, 2023, 23(24): 10568-10574.
JiangHao, ZhangHanbo, LiuChengbin, et al. Spatiotemporal dynamics monitoring of drought in yangtze river basin using temperature-precipitation-vegetation integrated dryness index (TPVDI)[J]. Transactions of the Chinese Society for Agricultural Machinery, 2025, 56(8): 95-106, 119.
WangJinsong, YaoYubi, YuanShujie, et al. Research advances and prospects in wheat drought monitoring techniques[J]. Agricultural Research in the Arid Areas, 2024, 42(4): 273-285.
XuXinyao, WangXufeng, ZhangSonglin, et al. Characteristics of spatial and temporal variations of global terrestrial droughts and analysis of their future trends[J]. Plateau Meteorology, 2025, 44(4): 923-942.
MaHongzhang, MengQingtao, LiuSumei. Improvement and parameterization of the TAU-OMEGA radiation model[J]. Transactions of the Chinese Society of Agricultural Engineering, 2024, 40(11): 292-298.
TongChangfu, HouHongfei, ZhengHexiang, et al. Investigation of sunflower soil moisture and development in the Kubuqi Desert using extensive time series remote sensing data[J]. China Rural Water and Hydropower, 2025(6): 157-165, 172.
GuoYuxing, SunCongjian, ChenWei, et al. Spatio-temporal characteristics and driving factors of vegetation drought in the middle Yellow River under different geomorphic conditions[J]. Scientia Geographica Sinica, 2024, 44(9): 1676-1683.
[29]
YangM, JiangY, ZhangW, et al. Short-term interval prediction strategy of photovoltaic power based on meteorological reconstruction with spatiotemporal correlation and multi-factor interval constraints[J]. Renewable Energy, 2024, 237: 121834.
XieWeiming, ZhangZhonglili, TaoJianping, et al. Multi-step hourly reference crop transpiration prediction method based on EA-BiLSTM-SCSO[J]. Water Saving Irrigation, 2025(3): 57-63, 70.
[32]
LiuX, YueF J, GuoT L, et al. High-frequency data significantly enhances the prediction ability of point and interval estimation[J]. Science of the Total Environment, 2024, 912: 169289.
ZhaoBingnan, SuXiaoling, ChuJingdong, et al. Domestic water use in Shaanxi Province based on KDE-IGMS model interval forecast[J]. Water Saving Irrigation, 2025(8): 7-14.
[35]
LiuS, WangZ, LiY. A novel approach for multivariate time series interval prediction of water quality at wastewater treatment plants[J]. Water Science & Technology, 2024, 90(10): 2813-2841.
[36]
AbdulrabH, HussinF A, IsmailI, et al. Energy efficient optimal deployment of industrial wireless mesh networks using transient trigonometric Harris Hawks optimizer[J]. Heliyon, 2024, 10(7): e28719.
[37]
ZhangK, JiY, ZhengW. The gradient descent adaptive moment parameter estimation for multi-frequency sine signal systems[J]. Optimal Control Applications and Methods, 2025, 46(6): 2799-2811.
[38]
ChenH, DengT, DuT, et al. An RF and LSSVM-NSGA-II method for the multi-objective optimization of high-performance concrete durability[J]. Cement and Concrete Composites, 2022, 129: 104446.
[39]
ChenH, ZhengY, HuangH, et al. A point-interval prediction framework for minimum miscibility pressure of CO2-crude oil systems[J]. Fuel, 2025, 381: 133573.
[40]
WangY, ZhengH, WangH, et al. Research on the rapid detection of formaldehyde emission from wood-based panels based on the AMSHKELM[J]. IEEE Access, 2025, 13: 5112-5128.
ChangLimao, ZhengHexiang, HouHongfei. Comprehensive impact assessment of Yellow River water security based on the analytic hierarchy process[J]. Water Saving Irrigation, 2025(8): 80-89.
ZhangQiqi, SunWenyi, MuXingmin, et al. Comprehensive climate-ecological-hydrological zoning of the Yellow River Basin using K-means clustering[J]. Water Saving Irrigation, 2025(7): 59-65.
[45]
SunY D, LiC, BiQ Y, et al. Predicted uniaxial compressive strength of fractured weakly cemented rock by LSSVM-NCV-ABKDE-based data-driven method[J]. Geomechanics and Geophysics for Geo-Energy and Geo-Resources, 2025, 11(1): 1-20.
SuYingying, LuXiaoping, XiaoFeng, et al. Soil drought monitoring with TVDI and ARIMA in Henan Province[J]. Transactions of the Chinese Society for Agricultural Machinery, 2024, 55(11): 391-401, 522.
[48]
KhodadadiN, TowfekS K, ZakiA M, et al. Predicting normalized difference vegetation index using a deep attention network with bidirectional GRU: a hybrid parametric optimization approach[J]. International Journal of Data Science and Analytics, 2025, 20(4): 3119-3146.
ZhouZhengdao, HuangBin. Several variants of support vector machine for monthly precipitation prediction based on data decomposition and hyperparameter optimization[J]. Water Saving Irrigation, 2025(9): 36-43.
[51]
WangY, ZhangN, ChenM, et al. Prediction and spatiotemporal dynamics of vegetation index based on deep learning and environmental factors in the Yangtze River Basin[J]. Forests, 2025, 16(3): 460.
[52]
GharakhanlouN M, PerezL. From data to harvest: leveraging ensemble machine learning for enhanced crop yield predictions across Canada amidst climate change[J]. Science of the Total Environment, 2024, 951: 175764.
[53]
ZhangD, ZhangX, HuH, et al. A hybrid model for point and interval forecasting of agricultural price based on the decomposition-ensemble and KDE[J]. Soft Computing, 2024, 28(17): 10153-10176.
[54]
LiJ D, SunY Z, WangX C, et al. Explainable machine learning for predicting mechanical properties of hot-rolled steel pipe[J]. Journal of Iron and Steel Research International, 2025, 32(8): 2475-2490.
[55]
MehrA D, JabarnejadM, SafariM J S. MOGGP: a novel multi objective geometric genetic programming model for drought forecasting[J]. Physics and Chemistry of the Earth, 2025, 138: 103879.
[56]
TongC, HouH, ZhengH, et al. A coupled model for forecasting spatiotemporal variability of regional drought in the Mu Us Sandy Land using a meta-heuristic algorithm[J]. Land, 2024, 13(11): 1731.
[57]
HouH, YinJ, QiuX, et al. Water heavy metal prediction using a CBM model based on hybrid metaheuristic algorithms[J]. IEEE Transactions on Geoscience and Remote Sensing, 2025, 63: 4211231.
[58]
YangC, XiaY. Interval Pareto front-based multi-objective robust optimization for sensor placement in structural modal identification[J]. Reliability Engineering & System Safety, 2024, 242: 109703.