Spatial Downscaling of the Tropical Rainfall Measuring Mission Precipitation Using Geographically Weighted Regression Kriging over the Lancang River Basin, China

Satellite-based precipitation products have been widely used to estimate precipitation, especially over regions with sparse rain gauge networks. However, the low spatial resolution of these products has limited their application in localized regions and watersheds.This study investigated a spatial downscaling approach, Geographically Weighted Regression Kriging(GWRK), to downscale the Tropical Rainfall Measuring Mission(TRMM) 3 B43 Version 7 over the Lancang River Basin(LRB) for 2001–2015. Downscaling was performed based on the relationships between the TRMM precipitation and the Normalized Difference Vegetation Index(NDVI), the Land Surface Temperature(LST), and the Digital Elevation Model(DEM). Geographical ratio analysis(GRA) was used to calibrate the annual downscaled precipitation data, and the monthly fractions derived from the original TRMM data were used to disaggregate annual downscaled and calibrated precipitation to monthly precipitation at 1 km resolution. The final downscaled precipitation datasets were validated against station-based observed precipitation in 2001–2015. Results showed that: 1) The TRMM 3 B43 precipitation was highly accurate with slight overestimation at the basin scale(i.e., CC(correlation coefficient) = 0.91, Bias = 13.3%). Spatially, the accuracies of the upstream and downstream regions were higher than that of the midstream region. 2) The annual downscaled TRMM precipitation data at 1 km spatial resolution obtained by GWRK effectively captured the high spatial variability of precipitation over the LRB. 3) The annual downscaled TRMM precipitation with GRA calibration gave better accuracy compared with the original TRMM dataset. 4) The final downscaled and calibrated precipitation had significantly improved spatial resolution, and agreed well with data from the validated rain gauge stations, i.e., CC = 0.75, RMSE(root mean square error) = 182 mm, MAE(mean absolute error) = 142 mm, and Bias = 0.78%for annual precipitation and CC = 0.95, RMSE = 25 mm, MAE = 16 mm, and Bias = 0.67% for monthly precipitation.

Spatial distribution of snow depth based on geographically weighted regression kriging in the Bayanbulak Basin of the Tianshan Mountains, China

Snow depth is a general input variable in many models of agriculture,hydrology,climate and ecology.This study makes use of observational data of snow depth and explanatory variables to compare the accuracy and effect of geographically weighted regression kriging(GWRK)and regression kriging(RK)in a spatial interpolation of regional snow depth.The auxiliary variables are analyzed using correlation coefficients and the variance inflation factor(VIF).Three variables,Height,topographic ruggedness index(TRI),and land surface temperature(LST),are used as explanatory variables to establish a regression model for snow depth.The estimated spatial distribution of snow depth in the Bayanbulak Basin of the Tianshan Mountains in China with a spatial resolution of 1 km is obtained.The results indicate that 1)the result of GWRK's accuracy is slightly higher than that of RK(R^2=0.55 vs.R^2=0.50,RMSE(root mean square error)=0.102 m vs.RMSE=0.077 m);2)for the subareas,GWRK and RK exhibit similar estimation results of snow depth.Areas in the Bayanbulak Basin with a snow depth greater than 0.15m are mainly distributed in an elevation range of 2632.00–3269.00 m and the snow in this area comprises 45.00–46.00% of the total amount of snow in this basin.However,the GWRK resulted in more detailed information on snow depth distribution than the RK.The final conclusion is that GWRK is better suited for estimating regional snow depth distribution.

Modeling of Spatial Distributions of Farmland Density and Its Temporal Change Using Geographically Weighted Regression Model

This study used spatial autoregression(SAR)model and geographically weighted regression(GWR)model to model the spatial patterns of farmland density and its temporal change in Gucheng County,Hubei Province,China in 1999 and 2009,and discussed the difference between global and local spatial autocorrelations in terms of spatial heterogeneity and non-stationarity.Results showed that strong spatial positive correlations existed in the spatial distributions of farmland density,its temporal change and the driving factors,and the coefficients of spatial autocorrelations decreased as the spatial lag distance increased.SAR models revealed the global spatial relations between dependent and independent variables,while the GWR model showed the spatially varying fitting degree and local weighting coefficients of driving factors and farmland indices(i.e.,farmland density and temporal change).The GWR model has smooth process when constructing the farmland spatial model.The coefficients of GWR model can show the accurate influence degrees of different driving factors on the farmland at different geographical locations.The performance indices of GWR model showed that GWR model produced more accurate simulation results than other models at different times,and the improvement precision of GWR model was obvious.The global and local farmland models used in this study showed different characteristics in the spatial distributions of farmland indices at different scales,which may provide the theoretical basis for farmland protection from the influence of different driving factors.

Comparison of Geographically Weighted Regression of Benthic Substrate Modeling Accuracy on Large and Small Wadeable Streams

Aquatic habitat assessments encompass large and small wadeable streams which vary from many meters wide to ephemeral. Differences in stream sizes within or across watersheds, however, may lead to incompatibility of data at varying spatial scales. Specifically, issues caused by moving between scales on large and small streams are not typically addressed by many forms of statistical analysis, making the comparison of large (>30 m wetted width) and small stream (<10 m wetted width) habitat assessments difficult. Geographically weighted regression (GWR) may provide avenues for efficiency and needed insight into stream habitat data by addressing issues caused by moving between scales. This study examined the ability of GWR to consistently model stream substrate on both large and small wadeable streams at an equivalent resolution. We performed GWR on two groups of 60 randomly selected substrate patches from large and small streams and used depth measurements to model substrate. Our large and small stream substrate models responded equally well to GWR. Results showed no statistically significant difference between GWR R2 values of large and small stream streams. Results also provided a much needed method for comparison of large and small wadeable streams. Our results have merit for aquatic resource managers, because they demonstrate ability to spatially model and compare substrate on large and small streams. Using depth to guide substrate modeling by geographically weighted regression has a variety of applications which may help manage, monitor stream health, and interpret substrate change over time.

Estimation of crop water requirement based on principal component analysis and geographically weighted regression

In this study the principal component analysis (PCA) and geographically weighted regression (GWR) are combined to estimate the spatial distribution of water requirement of the winter wheat in North China while the effect of the macroand micro-topographic as well as the meteorological factors on the crop water requirement is taking into account. The spatial distribution characteristic of the water requirement of the winter wheat in North China and its formation are analyzed based on the spatial variation of the main affecting factors and the regression coefficients. The findings reveal that the collinearity can be effectively removed when PCA is applied to process all of the affecting factors. The regression coefficients of GWR displayed a strong variability in space, which can better explain the spatial differences of the effect of the affecting factors on the crop water requirement. The evaluation index of the proposed method in this study is more efficient than the widely used Kriging method. Besides, it could clearly show the effect of those affecting factors in different spatial locations on the crop water requirement and provide more detailed information on the region where those factors suddenly change. To sum up, it is of great reference significance for the estimation of the regional crop water requirement.

Determination of the effective utilization coefficient of irrigation water based on geographically weighted regression

This study uses geographically weighted regression to determine the spatial distribution of the effective utilization coefficient of irrigation water in Zhejiang Province,China,owing to the influences of spatial attributes on the irrigation efficiency.The sample set of this study comprised 165 agricultural test sites.A multivariate linear regression model and a geographically weighted regression model were established using the effective utilization coefficient of agricultural irrigation water as the dependent variable in addition to a suite of independent variables,including the actual irrigation area,the percentage of farmland using water-saving irrigation,the type of irrigation area,the net water consumption per mu,the water intake method,the terrain slope,and the soil field capacity.Results revealed a positive spatial correlation and noticeable agglomeration features in the effective utilization coefficient of irrigation water in Zhejiang Province.The geographically weighted regression model performed better in terms of fit and prediction accuracy than the multivariate linear regression model.The obtained findings confirm the suitability of the geographically weighted regression model for determining the spatial distribution of the effective utilization coefficient of irrigation water in Zhejiang,and offer a new approach on a regional scale.

Covid-19 in West &East Africa, a Geographical Weighted Regression Exploration with http://mygeoffice.org/

Understanding the dynamics that affect the spread of Covid-19 is critical for the development of government measures to stop and reverse this nowadays disease propagation. Like in any epidemiological study, it is essential to analyze the spatial data to account for the inherent spatial heterogeneity within the data (spatial autocorrelation). This paper uses Geographically Weighted Regression (GWR) to identify the factors that influence the outbreak of Covid-19 in Western and Eastern countries of Africa. The analyses include traditional linear regression (including descriptive statistics, hierarchical clustering and correlations were not forgotten either) to reveal the importance of eight risk factors (population density, median age, aged over 65 years, GDP per capita, cardiovascular death rates, diabetes prevalence, female and male smokers) regarding Covid-19 dissemination. It is believed that this is the first attempt to explore possible causes associated with the spread of the Covid-19 pandemic in these disadvantage countries, where some intriguing clues are presented for further research such as the positive relationship between the financial purchase power of nations and the total number of infected people or the smoker’s gender impact on Covid-19.

Spatiotemporal Differentiation of Urban Spatial Form and Carbon Emissions in Poyang Lake City Group

In response to the inherent requirements of low-carbon land spatial planning in Jiangxi Province and the lack of existing research,this paper explored the mechanism of spatial form elements of Poyang Lake urban agglomeration on urban carbon emissions.Based on generalized linear regression and geographically weighted regression models,this paper analyzed the spatiotemporal distribution characteristics of carbon emissions,the spatiotemporal relationship between urban form index and carbon emissions,and the spatial differentiation of the intensity of dominant factors from 63 county-level administrative units in the Poyang Lake city group from 2005 to 2020.The results showed that:①The carbon emissions of urban agglomerations around Poyang Lake are generally increasing,and the spatial distribution of carbon emissions is characterized by high-value concentration in the middle and low-value agglomeration in pieces;②The main driving factor for the spatial heterogeneity of carbon emissions was the expansion of built-up area;③Improving urban compactness and optimizing urban form could effectively reduce urban carbon emissions.The results showcased the correlation between urban spatial landscape pattern and the spatiotemporal distribution of carbon emissions,which could make the low-carbon land spatial planning in the Poyang Lake city group more reasonable and practical.

A spatial regression modeling framework for examining relationships between the built environment and pedestrian crash occurrences at macroscopic level:A study in a developing country context

Researchers have been trying to identify the contributory factors behind pedestrian crash occurrences through studies at both microscopic and macroscopic levels.However,built environment-related factors have primarily been examined in developed countries,resulting in a limited understanding of the phenomenon in the context of developing countries.Methodologically,these studies mostly used global regression models,which failed to incorporate spatial autocorrelation and spatial heterogeneity.Additionally,some of these studies applied spatial regression models randomly without following a comprehensive logical framework behind their selections.Our study aimed to develop a comprehensive spatial regression modeling framework to examine the relationships between pedestrian crash occurrences and the built environment at the macroscopic level in a megacity,Dhaka,the capital of a developing country:Bangladesh.Using secondary pedestrian crash data,the study applied one global non-spatial model,two global spatial regression models,and two local spatial regression models following a comprehensive spatial regression modeling framework.The factors which significantly contributed to pedestrian crash occurrences in Dhaka were employed person density,mixed and recreational land use density,primary road density,major intersection density,and share of non-motorized modes.Except for the last factor,all the other ones were positively related to pedestrian crash density.Among the five models used in this study,the multiscale geographically weighted regression(MGWR)performed the best as it calibrated each local relationship with a distant spatial scale parameter.The findings and recommendations presented in this study would be useful for reducing pedestrian crashes and choosing the appropriate modeling technique for crash analysis.

Spatial Regression Analysis of Pedestrian Crashes Based on Point-of-Interest Data

Pedestrian safety has recently been considered as one of the most serious issues in the research of traffic safety. This study aims at analyzing the spatial correlation between the frequency of pedestrian crashes and various predictor variables based on open source point-of-interest (POI) data which can provide specific land use features and user characteristics. Spatial regression models were developed at Traffic Analysis Zone (TAZ) level using 10,333 pedestrian crash records within the Fifth Ring of Beijing in 2015. Several spatial econometrics approaches were used to examine the spatial autocorrelation in crash count per TAZ, and the spatial heterogeneity was investigated by a geographically weighted regression model. The results showed that spatial error model performed better than other two spatial models and a traditional ordinary least squares model. Specifically, bus stops, hospitals, pharmacies, restaurants, and office buildings had positive impacts on pedestrian crashes, while hotels were negatively associated with the occurrence of pedestrian crashes. In addition, it was proven that there was a significant sign of localization effects for different POIs. Depending on these findings, lots of recommendations and countermeasures can be proposed to better improve the traffic safety for pedestrians.

Spatiotemporal dynamics of population density in China using nighttime light and geographic weighted regression method

The distribution and dynamic changes of regional or national population data with long time series are very important for regional planning,resource allocation,government decision-making,disaster assessment,ecological protection,and other sustainability research.However,the existing population datasets such as LandScan and WorldPop all provide data from 2000 with limited time series,while GHS-POP only utilizes land use data with limited accuracy.In view of the limited remote sensing images of long time series,it is necessary to combine existing multi-source remote sensing data for population spatialization research.In this research,we developed a nighttime light desaturation index(NTLDI).Through the cross-sensor calibration model based on an autoencoder convolutional neural network,the NTLDl was calibrated with the same period Visible Infrared Imaging Radiometer Suite Day/Night Band(VIRS-DNB)data.Then,the geographically weighted regression method is used to determine the population density of China from 1990 to 2020 based on the long time series NTL.Furthermore,the change characteristics and the driving factors of China's population spatial distribution are analyzed.The large-scale,long-term population spatialization results obtained in this study are of great significance in government planning and decision-making,disaster assessment,resource allocation,and other aspects.

Spatial distribution and potential poverty-returning factors of former poverty-stricken villages in the Liangshan Mountains,China

Remote mountainous villages are at risk of falling back into poverty,despite having been lifted out of extreme poverty.However,there has been a lack of focus on the factors contributing to povertyreturn in these villages,which making it difficult to understand the risks and their underlying causes.This study investigates the spatial distribution of 546 key assistance villages(KAVs)in the Liangshan mountainous region,a former poverty-stricken area,using the average nearest neighbor(ANN)and kernel density estimation(KDE)methods.Linear regression and geographically weighted regression(GWR)models are then employed to analyze the relationship between the KAVs'economy and potential povertyreturning factors.The results show that KAVs are primarily located in elevation ranges of 1800-2500 m(31.87%),with slopes of 6°-15°(42.67%)and 2-3 km from the township(28.94%).The distribution of KAVs exhibits distinct spatial clustering,forming four gathering areas.Several factors impact the KAVs'economy positively,including the normalized difference vegetation index(NDVI),built-up area,grassland,and education facilities,while elevation has a negative effect.The built-up area has the most critical impact on the rural economy,followed by NDVI and elevation.Additionally,education facilities and grassland areas also have significant effects.The study suggests promoting the Ex-situ Poverty Alleviation Relocation Program(ESPARP)and increasing rural built-up areas,grasslands,and educational facilities as practical measures for preventing poverty return and promoting economic development promotion in remote mountain villages.

Temporal and spatial responses of ecological resilience to climate change and human activities in the economic belt on the northern slope of the Tianshan Mountains, China

In the Anthropocene era,human activities have become increasingly complex and diversified.The natural ecosystems need higher ecological resilience to ensure regional sustainable development due to rapid urbanization and industrialization as well as other intensified human activities,especially in arid and semi-arid areas.In the study,we chose the economic belt on the northern slope of the Tianshan Mountains(EBNSTM)in Xinjiang Uygur Autonomous Region of China as a case study.By collecting geographic data and statistical data from 2010 and 2020,we constructed an ecological resilience assessment model based on the ecosystem habitat quality(EHQ),ecosystem landscape stability(ELS),and ecosystem service value(ESV).Further,we analyzed the temporal and spatial variation characteristics of ecological resilience in the EBNSTM from 2010 to 2020 by spatial autocorrelation analysis,and explored its responses to climate change and human activities using the geographically weighted regression(GWR)model.The results showed that the ecological resilience of the EBNSTM was at a low level and increased from 0.2732 to 0.2773 during 2010–2020.The spatial autocorrelation analysis of ecological resilience exhibited a spatial heterogeneity characteristic of"high in the western region and low in the eastern region",and the spatial clustering trend was enhanced during the study period.Desert,Gobi and rapidly urbanized areas showed low level of ecological resilience,and oasis and mountain areas exhibited high level of ecological resilience.Climate factors had an important impact on ecological resilience.Specifically,average annual temperature and annual precipitation were the key climate factors that improved ecological resilience,while average annual evapotranspiration was the main factor that blocked ecological resilience.Among the human activity factors,the distance from the main road showed a negative correlation with ecological resilience.Both night light index and PM2.5 concentration were negatively correlated with ecological resilience in the areas with better ecological conditions,whereas in the areas with poorer ecological conditions,the correlations were positive.The research findings could provide a scientific reference for protecting the ecological environment and promoting the harmony and stability of the human-land relationship in arid and semi-arid areas.

可达性对城市群多模式交通碳排放的空间异质性影响

当前,在交通行业碳减排存在巨大挑战以及未来中国城市群交通发展长远规划双重背景下,如何通过改善交通可达性提高居民出行效率和减少碳排放是亟待解决的关键问题之一。本文基于城际出行手机信令数据,从居民出行的角度提出城际多模式交通客运碳排放量方法,并采用梯度提升决策树(GBDT)模型及多尺度地理加权回归(MGWR)模型探讨可达性对区域碳排放量的空间异质性影响。以关中平原城市群为例进行验证,结果表明:城际公路客运碳排放量远大于铁路,呈现沿交通基础设施线路分布的特征;在整体区域范围内,可达性指标对碳排放水平具有一定的正向边际效应;MGWR能够刻画碳排放与可达性指标关系的空间异质性及尺度差异;经济潜能可达性、介数中心性及接近中心性对城际碳排放具有显著的正向空间异质性影响,但影响尺度不同;公路客运碳排放对介数中心性及接近中心性要素较为敏感,经济潜能对碳排放的影响较为平稳;铁路出行可达性的提升对中心城市的影响效应低于周边区县城市。

Industrial Carbon Emission Distribution and Regional Joint Emission Reduction:A Case Study of Cities in the Pearl River Basin,China

China’s low-carbon development path will make significant contributions to achieving global sustainable development goals.Due to the diverse natural and economic conditions across different regions in China,there exists an imbalance in the distribution of car-bon emissions.Therefore,regional cooperation serves as an effective means to attain low-carbon development.This study examined the pattern of carbon emissions and proposed a potential joint emission reduction strategy by utilizing the industrial carbon emission intens-ity(ICEI)as a crucial factor.We utilized social network analysis and Local Indicators of Spatial Association(LISA)space-time trans-ition matrix to investigate the spatiotemporal connections and discrepancies of ICEI in the cities of the Pearl River Basin(PRB),China from 2010 to 2020.The primary drivers of the ICEI were determined through geographical detectors and multi-scale geographically weighted regression.The results were as follows:1)the overall ICEI in the Pearl River Basin is showing a downward trend,and there is a significant spatial imbalance.2)There are numerous network connections between cities regarding the ICEI,but the network structure is relatively fragile and unstable.3)Economically developed cities such as Guangzhou,Foshan,and Dongguan are in the center of the network while playing an intermediary role.4)Energy consumption,industrialization,per capita GDP,urbanization,science and techno-logy,and productivity are found to be the most influential variables in the spatial differentiation of ICEI,and their combination in-creased the explanatory power of the geographic variation of ICEI.Finally,through the analysis of differences and connections in urban carbon emissions under different economic levels and ICEI,the study suggests joint carbon reduction strategies,which are centered on carbon transfer,financial support,and technological assistance among cities.

一类变系数空间滞后的混合地理加权回归模型

为解决因变量空间滞后存在的局域性问题,对现有常系数空间滞后的混合地理加权回归模型作了具有更广泛形式的拓展,提出一类变系数空间滞后的混合地理加权回归(MGWR-VSLR)模型。MGWR-VSLR模型实现了空间相关性与空间异质性融合,涵盖了绝大多数地理加权回归的模型形式,基于重构参数化方法和似然比检验分别给出模型的系数估计方法与显著性检验以及选取变系数的判别检验。在蒙特卡罗模拟与实际应用中,MGWR-VSLR模型均表现出优异的因变量拟合与预测能力。MGWR-VSLR模型的提出为定量化研究空间效应问题设定适宜的模型形式提供了支撑依据。

1990-2020年长三角地区生境质量与夜间灯光的空间关系

基于1990-2020年长三角地区的土地利用数据和夜间灯光数据,运用InVEST生境质量评估模型、双变量空间自相关、相关系数对生境质量与夜间灯光的时空关系进行研究,并应用地理加权回归和归纳法分析两者的响应规律,结果表明:(1)浙江省总体上生境最优且城镇发展最快,安徽省在城镇快速发展的同时生境相对下降最少,生境质量和夜间灯光空间分布格局呈现出与区位、地形等因素的空间耦合性。(2)长三角地区生境质量与夜间灯光指数随着时间的发展空间上的依赖性越来越强,聚类模式主要为高-低、低-高及低-低3种。高-低聚类区域主要分布在浙江省和安徽省的南部、西部的山地丘陵,研究期内呈现缩小的趋势;低-高聚类区域在长三角东部平原区呈集聚分布、北部平原丘陵区呈零散分布,呈现扩张趋势。(3)不同年份各省份生境质量与夜间灯光指数呈显著负相关,且经济越发达的省份,生境质量与夜间灯光指数的负相关关系越强;生境质量与夜间灯光指数呈高度负相关的区县主要分布在长江下游沿岸和浙江省东部,相关性不显著的区县主要分布在安徽省、江苏省的北部,变化类型以负相关关系持续增强型占比最高。(4)生境质量与夜间灯光指数的相关关系随着城市发展呈U型变化,即随着城市发展,生境质量与夜间灯光指数的负相关关系先增强后下降。研究结果可为长三角城市可持续发展、生境调控提供前期基础。

湖南省城镇土地利用多样性对人口密度的影响研究

城镇人口密度降低已经成为制约部分城镇发展的重要因素之一。选取湖南省为案例区,在定量刻画湖南省城镇人口密度与土地利用多样性格局基础上,运用地理加权回归模型探索两者的关系,并构建压力-状态-响应模型探索湖南省城镇土地利用多样性对人口密度的影响路径。结果表明:①湖南省城镇人口密度、城镇建设用地土地利用多样性均存在明显的空间正相关性,但聚集趋势相反;②单一化的城镇建设用地利用方式既不利于湖南省城镇土地利用多样性的提升,也不利于湖南省城镇人口密度的提升,工业布局可以引导湖南省城镇人口密度的提升,注重经济发展、提高城镇公共服务能力是提高城镇土地利用多样性进而提高城镇人口密度的有效方法;③湖南省二、三产业增加值与城镇人口密度呈现弱负相关且呈现“南高北低”的格局;④湖南省城镇居民可支配收入与城镇人口密度的相关性呈现出“西北高、东南低”的格局,并在长沙市中心城区一带出现正相关与负相关的分界。最后,从优化城镇职能体系对应对历史遗留问题、促进人口密度有序提升方面提出了政策建议。

轨道站网络中心性、客流与空间热力耦合分析

城市空间热力反映了人口聚集与街道活力。为探究城市轨道交通与空间热力分布的互动关系,从微观层面的轨道站点切入,采用百度热力图和轨道站点客流数据,以上海为例,对轨道站点的网络中心性、客流与站域空间热力进行耦合分析。首先采用Pearson双变量相关性研究两类轨道站点属性与空间热力的总体耦合关系,然后引入双变量空间自相关和地理加权回归分析方法分别挖掘网络中心性与站域热力、站域热力与站点客流的空间关联模式,并对比两类耦合性的空间差异。结果表明:轨道站点的网络中心性与空间热力的耦合性明显优于轨道客流与空间热力的耦合性,交通区位优势通常能够形成较高的空间热力,客流水平的影响因素则更为复杂;空间热力更适合量化核心区以外区域的轨道交通与城市空间互动关系,轨道交通网络化对空间热力提升具有乘数效应,而在开发密度低的区域提升空间热力更有助于刺激轨道客流;利用空间热力数据评估城市核心区以外区域的新建站点客流潜力具有可行性,但仅用热力预测客流具有局限性;轨道站点周边城市更新可参考不同空间区位站点的两类耦合性差异进行优化。该研究探索了结合城市空间热力分布完善轨道交通线网布局、针对不耦合因素优化轨道站点公共交通导向型开发(TOD)的分析框架,为微观层面衡量城市轨道交通“人-地”关系提供了新视角。

土地利用与城市轨道交通客流的非线性关系

城市轨道交通站点影响范围内土地利用对客流影响具有时空分异特征且存在类型差异,为针对性探讨不同站点两者的复杂非线性关系,提出一种基于土地利用空间分布规律、对站点实际影响范围进行差异化识别的方法;并通过分时段多尺度地理加权回归,获取能够表征土地利用对客流影响时空变化特征的站点聚类指标,采用K-means++算法将研究区域内的站点划分为4类;进而基于改进的梯度提升决策树模型分类定量探讨不同类别下土地利用与轨道交通客流的复杂非线性关系。研究表明:通过捕捉不同站点土地利用与客流的时空分异特征对站点进行分类识别,可有效提升两者非线性关系模型的解释度。根据模型输出结果,发现不同类别站点影响轨道交通客流的关键土地利用要素不同,第1类中关键变量为相对重要性分别为61.35%和30.08%的公交站点数量和慢行密度;第4类的情况类似但相对数值有所变化,公交站点数量的相对重要性由61.35%下降至30.31%;建筑密度在第2类中以66.57%的相对重要度占据最大比例;但在第3类中仅占5.59%。此外,不同类别站点影响范围内土地利用与轨道交通客流的关系存在较为显著且各异的阈值效应。研究表明,对于不同类别站点的用地开发应各有侧重,且应结合实际将土地利用设计指标控制在相应的合理范围内。研究为差异化的站点周边土地利用开发策略的制定提供了理论支持和量化指导。