Kernel Density Method Research Articles

A novel semantic recognition framework of urban functional zones supporting urban land structure analytics based on open‐source data

AbstractTo solve problems related to urban land development, protection, and management, urban planners focus on achieving high precision and currency for fine‐grained semantic recognition of urban functional zones (UFZs). The classification of discrete functional zones and qualitative recognition of UFZs have been well studied, but the precise recognition of UFZs with explicit depictions of socioeconomic urban functions has seldom been addressed. Thus, we propose a novel open‐source data‐guided semantic urban function recognition (OSD‐SFR) framework to analyze UFZs and quantitatively link open‐source data (OSD) with the Chinese national urban function classification code. The spatial relationship between the combined semantic database and the segmented urban zones data set was measured, graded, and categorized. Kernel density and statistical methods based on a single data source of points‐of interest were used for comparative experiments. The results indicate that OSD‐SFR yields more fine‐grained delineations between urban functions and achieves more accurate recognition. Critical semantic information from different OSD sources is deeply integrated. This study provides a realistic reference, including detailed descriptions of the spatial distribution and morphological structure of UFZs, to support long‐term city planning.

GIS Identifies Elderly Healthcare Centre Future Planning: A Case Study of Local Areas in Northeast of Thailand.

Spatial density measures are an important tool for future healthcare planning, particularly in Northeast Thailand where the ageing population has rapidly increased for a decade. The objective of this study was to determine the relationship of geographic and elderly population density in Chiang Yuen district, Mahasarakham province, to ascertain suitable areas for elderly healthcare centres. The data of the elderly located in the villages were collected by the Global Position System (GPS) using Kernel density method to employ the analysis of population density, and GIS for healthcare mapping, and Analytical Hierarchy Process (AHP) as multiple criteria decision-making tools. Factors were determined by using the overlay analysis method, where social, physical and economic factors were variables of interest that were used for the analysis of suitable areas for Elderly Healthcare Centres. The results revealed that the spatial density of the elderly population could be divided into four levels: Low density for the agriculture areas, medium density for agricultural areas including small villages. High density for the areas near communities and public area services and highest density for the centre of towns. The most suitable area was an area with many elderly inhabitants in the community, near transportation routes including main highways leading to other areas. In conclusion, the results confirmed that the density of elderly and proximity to both local stores and transport would make the lower part of Chiang Yuen Sub-district a suitable area to establish a healthcare centre for the elderly, therefore, the authorities and stakeholders should recognise this in planning.

Model Clusterization of WebGIS-based Mobile Internet Provider Connections

The spread of the corona virus or COVID-19 pandemic has presented its own challenges for educational institutions in Indonesia. As a result of this policy, the education sector such as universities stopped the face-to-face learning process. Ibn Khaldun Bogor University conducts an online learning process which can be carried out from each student's home. The easy teaching and learning process, cost efficiency, easy access anywhere and anytime have made many educational institutions adopt the e-learning method. Geographical Information System (GIS) is information based on a work system that enters, manages, manipulates, and analyzes data and explains descriptions. Process Clusterization Model of Cellular Internet Provider Connection Based on Webgis, the information arrangement needs to be packaged in a geographic information system using the Kernel Density method displayed in a WebGIS.

Habitat use by the Guiana dolphin (Sotalia guianensis) (Cetartiodactyla: Delphinidae) in southeastern Brazil

The coastal habit and site fidelity reported for Sotalia guianensis make this species especially vulnerable to human activities. Thus, ecological studies focusing on the habitat use of the populations are essential for S. guianensis protection, as it helps to identify critical areas for conservation. The present study aims to analyze the habitat use by Sotalia guianensis in three bays in southeastern Brazil: Guanabara Bay (GB11Guanabara Bay.), Sepetiba Bay (SB22Sepetiba Bay.) and Ilha Grande Bay (IGB33Ilha Grande Bay.). From 2017 to 2019, four surveys per month were carried out in each bay on board of outboard-powered boats. After locating a group of dolphins, researchers recorded data every five minutes, applying the focal-group method. Information collected included geographical position, time of day, group size and composition, predominant behavioral state, and local depth in meters. The ArcGIS software was selected to create distribution maps and estimate the area use using the fixed Kernel Density method (KDE). The variables collected were statistically compared within and between populations. Feeding was the predominant behavior (frequency of 84.2% in GB, 73.3% in SB and 52.2% in IGB), but behavior frequencies differed between populations (X2=57.622, df=2, p < 0.00001). Group size and composition also varied according to the study area (KWsize: H2,732=68.479, p < 0.0001; KWcomposition: H2,732=85.311, p < 0.001), IGB and SB showing large aggregations with many calves and GB, small groups. In GB, dolphin distribution was restricted to shallow waters of the northeastern region and central channel (K95% = 9.9 km2; K50% = 1.2 km2). Guiana dolphins from SB selected greater depths, concentrating in the central region (K95% = 190.6 km2; K50% = 32.6 km2). In IGB, dolphins occurred in an extensive area (K95% = 264.9 km2; K50% = 42.5 km2), though observed more frequently in shallow waters of the western region. The results described can contribute to a better knowledge of the populations and serve as a basis for conservation strategies.

The Influence of Spatial Grid Division on the Layout Analysis of Urban Functional Areas

The identification of urban functional areas is essential for urban planning and sustainable development. Spatial grids are the basic units for the implementation of urban plans and management by cities or development zones. The emergence of internet “big data” provides new ideas for the identification of urban functional areas. Based on point of interest (POI) data from Baidu Maps, the Xicheng District of Beijing was divided into grids with side lengths of 200, 500, and 1000 m in this study. The kernel density method was used to analyze the spatial structure of POI data. Two indicators, that is, the frequency density and category ratio, were then used to identify single- and mixed-functional areas. The results show that (1) commercial and financial areas are concentrated in the city center and multiple business centers have not developed; (2) scenic areas account for the largest proportion of single-functional areas in the Xicheng District of Beijing, followed by education and training, residence, and party and government organizations areas; and (3) the 200 × 200 m and 500 × 500 m grids are the most suitable for the identification of single- and mixed-functional areas, respectively.

A Comparison of Statistical Tools for Identifying Modality in Body Mass Distributions

The assessment of modality or \bumps in distributions is of in- terest to scientists in many areas. We compare the performance of four statistical methods to test for departures from unimodality in simulations, and further illustrate the four methods using well-known ecological datasets on body mass published by Holling in 1992 to illustrate their advantages and disadvantages. Silverman's kernel density method was found to be very conservative. The excess mass test and a Bayesian mixture model approach showed agreement among the data sets, whereas Hall and York's test pro- vided strong evidence for the existence of two or more modes in all data sets. The Bayesian mixture model also provided a way to quantify the un- certainty associated with the number of modes. This work demonstrates the inherent richness of animal body mass distributions but also the diculties for characterizing it, and ultimately understanding the processes underlying them.

Wildlife roadkill in the Tsavo Ecosystem, Kenya: identifying hotspots, potential drivers, and affected species

Roadkill is one of the highest causes of wildlife mortality and is of global conservation concern. Most roadkill studies have focused on wildlife in developed countries such as the United States of America and temperate biomes, but there are limited data for the impacts of roads on wildlife in the African tropics, where road infrastructure development is projected to grow rapidly in natural environments and conservation areas. The Tsavo Conservation Area is an important biodiversity hotspot in eastern Kenya and is bisected by a major highway and railways that connect the port of Mombasa to the interior. Along this infrastructure corridor, roadkill was recorded for 164 days over an 11-year period (2007–2018). In total, 1,436 roadkill were recorded from 13,008 km driven of a 164.42 km Nairobi-Mombasa road representing 0.11 collisions per kilometer. The majority of roadkill were small to medium sized mammals (<15kg) (53%; n = 756), whereas birds comprised 32% (n = 460), reptiles 10% (n = 143), with the remaining 5% (n = 77) being large mammals (>15kg). Of the 460 birds recorded, 264 were identifiable represented by 62 species. All large mammals comprising 10 species were identified, including the African elephant, Loxodonta africana and the endangered African wild dog, Lycaon pictus. Thirteen species of small mammal were also identified dominated by Kirk's dik-dik (Madoqua kirkii). Reptiles were represented by 11 species which were identified to the species level. Roadkill hotspots were identified using a kernel density method. The spatial distribution of roadkill was associated with adjacent shrub vegetation and proximity to permanent and seasonal rivers, and differences in seasonality and habitats were observed. Roadkill was lowest on road sections that traversed settled areas as opposed to roads adjacent to the protected areas. The results demonstrate that roadkill for two of the taxonomic groups - mammals and birds - appear high with numerous species detected in the Tsavo Conservation Area. These results can be used to focus efforts to reduce wildlife mortality by guiding future mitigation efforts.

Questioning gender stereotypes: A case study of adolescents walking activity space in a small Central European city

BackgroundThere has been a lack of research on the gender differences in spatial attributes of adolescents' walking activity space (WAS). Moreover, while other literature shows girls walking less through city centres than boys, there are limitations concerning the absence of its precise definition and spatial delimitation. Additionally, the morphological structure of city centres, and differences in their constituent parts have not been considered. AimsIn this study, the extent to which the area is used for walking by girls and boys, as well as gender differences in spatio-temporal walking patterns during weekdays were investigated, with a special emphasis on the city centre and its constituent parts. MethodsUsing a mapping activity, self-reported spatio-temporal walking data from 304 adolescents (168 girls) aged 13–16 years were collected. To fill the gap in existing research, the study was conducted in the central ward of the small medieval historic city of Banská Bystrica (Slovakia). Attributes of WAS were calculated using Standard deviational ellipse analysis and Kernel density method. Gender differences in walking were tested by non-parametric Mann-Whitney U test for the city centre (defined by a pedestrian zone), its constituent parts (streets and squares) and for the whole ward separately. In all the cases, time of the day was considered. ResultsDespite the distinction between streets and squares in the city centre, no significant gender differences in walking were found, which is in contrast to the prevailing body of literature. Yet, a slight predominance of girls in the squares (regardless of the time of day) and boys in adjacent streets and alleys (after dark) was present. The WAS of girls was more compact compared to that of boys, with the biggest difference after dark. ConclusionsThe compact WAS of girls is in line with existing research on adults, despite the absence of traditional gender roles of wife and mother during adolescence. In addition, there is a subtle indication that a pedestrianized historic centre with a high degree of open space in squares and visually enclosed adjacent alleys may create different conditions for gender differences compared to other types of urban cores. In the future, a more in-depth examination of the degree of enclosure in public space and its influence on walking patterns is necessary.

Diel Activity Patterns of Three Sympatric Medium-Sized Carnivores during Winter and Spring in a Heavy Snowfall Area in Northeastern Japan

In a heavy snowfall area in northeastern Japan, we clarified the diel activity patterns of and seasonal differences in three sympatric carnivores (red foxes, raccoon dogs, and Japanese martens) during winter and spring, and the interspecific differences among the activity patterns in each season. The activity time of the target species was obtained by camera-trap surveys at seven sites around the Experimental Forest of Yamagata University from January to June 2019. The diel activity patterns of each species, and seasonal and interspecific differences in the activity patterns were evaluated using the kernel density method. Red foxes showed a unimodal nocturnal pattern in spring, whereas they were also observed during the daytime in winter. Raccoon dogs showed a unimodal nocturnal pattern, and peak activity after sunset in both seasons. Japanese martens were clearly nocturnal during both seasons. Because there were a few interspecific differences in diel activity patterns among three species, species interactions had little effect on diel activity patterns. Our findings indicate that the diel activity of medium-sized carnivores responding to winter conditions may vary depending on species.

Map simulation of dogs' behaviour using population density of probabilistic model

This paper proposes a simulator to demonstrate dogs' behaviours considering individual and group habits, which is designed to be purposefully expandable for disease control. The proposed system is developed using Unity and Mapbox SDK. The normal distribution, kernel density method and probabilistic model are applied to simulate the movement behaviour, world interaction and behaviour rates, respectively. The simulation is validated on an area of Saibai, located in the north-western of Torres Strait islands, Australia. This reports a median tie-strength of 0.0106 which is slightly different from the value calculated from the GPS information of 0.0113. It thus contains the relative error of 6.19%. Then, the simulation is applied to three cities in Thailand. They are all reported with higher tie-strengths, when compared to Saibai. This is because of the significantly higher average numbers of dogs and group distances, with the larger connections between dogs and their communities.

Seasonal Variation in the Home Ranges of Black Curassow,Crax alector, in French Guiana

Cracidae is the most threatened avian family in the Neotropics, mainly because of habitat destruction, heavy hunting pressure and poaching. In French Guiana, Black Curassows are heavily hunted, although basic knowledge of the ecological and demographical traits of the species remains limited. Such a gap prevents any attempt to assess the impact of hunting and to help stakeholders to develop proposals ensuring hunting sustainability. The spatial relationship between animals and their habitat is important for conservation management, being related to population densities through complex patterns. Here, we report on a radio-tracking study of Black Curassows in tropical primary rainforest, in Nouragues National Reserve, French Guiana. The aims of the study were to estimate home range size and its variation across seasons, and to quantify movement patterns of the birds. We captured and fitted VHF tags to four adults, and tracked them for 10 to 21.5 months. Daily movements were recorded, and home ranges estimated using the Kernel Density method, for two consecutive wet seasons and one dry season. Using 95% and 50% Kernel densities, the average annual home range and core area were 96.3± 32.6 ha (SE) and 22.8 ± 2.8 ha respectively. Home ranges appeared spatially stable over the two years, and overlapped between neighbouring groups. During the dry season, Black Curassows did not migrate but tended to enlarge their home range, with greater daily movements and higher home range overlap. Although additional data are still needed, our results can help to improve the knowledge and management of this poorly studied species.

Epidemic analysis of COVID-19 in Italy based on spatiotemporal geographic information and Google Trends.

Since the first two novel coronavirus cases appeared in January of 2020, the outbreak of the COVID-19 epidemic seriously threatens the public health of Italy. In this article, the distribution characteristics and spreading of COVID-19 in various regions of Italy were analysed by heat maps. Meanwhile, spatial autocorrelation, spatiotemporal clustering analysis and kernel density method were also applied to analyse the spatial clustering of COVID-19. The results showed that the Italian epidemic has a temporal trend and spatial aggregation. The epidemic was concentrated in northern Italy and gradually spread to other regions. Finally, the Google Trends index of the COVID-19 epidemic was further employed to build a prediction model combined with machine learning algorithms. By using Adaboost algorithm for single-factor modelling,the results show that the AUC of these six features (mask, pneumonia, thermometer, ISS, disinfection and disposable gloves) are all >0.9, indicating that these features have a large contribution to the prediction model. It is also implied that the public's attention to the epidemic is increasing as well as the awareness of the need for protective measures. This increased awareness of the epidemic will prompt the public to pay more attention to protective measures, thereby reducing the risk of coronavirus infection.

Using movement to inform conservation corridor design for Mojave desert tortoise

BackgroundPreserving corridors for movement and gene flow among populations can assist in the recovery of threatened and endangered species. As human activity continues to fragment habitats, characterizing natural corridors is important in establishing and maintaining connectivity corridors within the anthropogenic development matrix. The Mojave desert tortoise (Gopherus agassizii) is a threatened species occupying a variety of habitats in the Mojave and Colorado Deserts. Desert tortoises have been referred to as corridor-dwellers, and understanding how they move within suitable habitat can be crucial to defining corridors that will sustain sufficient gene flow to maintain connections among populations amidst the increases in human development.MethodsTo elucidate how tortoises traverse available habitat and interact with potentially inhospitable terrain and human infrastructure, we used GPS dataloggers to document fine-scale movement of individuals and estimate home ranges at ten study sites along the California/Nevada border. Our sites encompass a variety of habitats, including mountain passes that serve as important natural corridors connecting neighboring valleys, and are impacted by a variety of linear anthropogenic features. We used path selection functions to quantify tortoise movements and develop resistance surfaces based on landscape characteristics including natural features, anthropogenic alterations, and estimated home ranges with autocorrelated kernel density methods. Using the best supported path selection models and estimated home ranges, we determined characteristics of known natural corridors and compared them to mitigation corridors (remnant habitat patches) that have been integrated into land management decisions in the Ivanpah Valley.ResultsTortoises avoided areas of high slope and low perennial vegetation cover, avoided moving near low-density roads, and traveled along linear barriers (fences and flood control berms).ConclusionsWe found that mitigation corridors designated between solar facilities should be wide enough to retain home ranges and maintain function. Differences in home range size and movement resistance between our two natural mountain pass corridors align with differences in genetic connectivity, suggesting that not all natural corridors provide the same functionality. Furthermore, creation of mitigation corridors with fences may have unintended consequences and may function differently than natural corridors. Understanding characteristics of corridors with different functionality will help future managers ensure that connectivity is maintained among Mojave desert tortoise populations.

The Spatio-Temporal Characteristics and Influencing Factors of Covid-19 Spread in Shenzhen, China-An Analysis Based on 417 Cases.

The global pandemic of COVID-19 has made it the focus of current attention. At present, the law of COVID-19 spread in cities is not clear. Cities have long been difficult areas for epidemic prevention and control because of the high population density, high mobility of people, and high frequency of contacts. This paper analyzed case information for 417 patients with COVID-19 in Shenzhen, China. The nearest neighbor index method, kernel density method, and the standard deviation ellipse method were used to analyze the spatio-temporal characteristics of the COVID-19 spread in Shenzhen. The factors influencing that spread were then explored using the multiple linear regression method. The results show that: (1) The development of COVID-19 epidemic situation in Shenzhen occurred in three stages. The patients showed significant hysteresis from the onset of symptoms to hospitalization and then to diagnosis. Prior to 27 January, there was a relatively long time interval between the onset of symptoms and hospitalization for COVID-19; the interval decreased thereafter. (2) The epidemic site (the place where the patient stays during the onset of the disease) showed an agglomeration in space. The degree of agglomeration constantly increased across the three time nodes of 31 January, 14 February, and 22 February. The epidemic sites formed a “core area” in terms of spatial distribution and spread along the “northwest–southeast” direction of the city. (3) Economic and social factors significantly impacted the spread of COVID-19, while environmental factors have not played a significant role.

Changing River Flood Timing in the Northeastern and Upper Midwest United States: Weakening of Seasonality over Time?

Climate change is likely to impact precipitation as well as snow accumulation and melt in the Northeastern and Upper Midwest United States, ultimately affecting the quantity and seasonal distribution of streamflow. The objective of this study is to analyze seasonality of long-term daily annual maximum streamflow (AMF) records and its changes for 158 sites in Northeastern and Upper Midwest Unites States. A comprehensive circular statistical approach comprising a kernel density method was used to assess the seasonality of AMF. Temporal changes were analyzed by separating the AMF records into two 30-year sub-periods (1951–1980 and 1981–2010). Results for temporal change in seasonality showed mixed pattern/trend across the stations. While for majority of stations, the distribution of AMF timing is strongly unimodal (concentrated around spring season) for the period 1951–1980, the seasonal modes have weakened during the period 1981–2010 for several stations along the coastal region with simultaneous emergence of multiple modes indicating changes of seasonality therein. The fresh statistical approach based on non-parametric circular density estimates reduces some of the limitations of previous studies to detect and model event timing distributions with multiple seasons and addresses issues of non-stationarity in the data records of extreme events.

Spatio-temporal dynamics of landscape use by the bumblebee Bombus pauloensis (Hymenoptera: Apidae) and its relationship with pollen provisioning.

Understanding how bees use resources at a landscape scale is essential for developing meaningful management plans that sustain populations and the pollination services they provide. Bumblebees are important pollinators for many wild and cultivated plants, and have experienced steep population declines worldwide. Bee foraging behavior can be influenced by resource availability and bees’ lifecycle stage. To better understand these relationships, we studied the habitat selection of Bombus pauloensis by tracking 17 queen bumblebees with radio telemetry in blueberry fields in Entre Ríos province, Argentina. To evaluate land use and floral resources used by bumblebees, we tracked bees before and after nest establishment and estimated home ranges using minimum convex polygons and kernel density methods. We also classified the pollen on their bodies to identify the floral resources they used from the floral species available at that time. We characterized land use for each bee as the relative proportion of GPS points inside of each land use. Bumblebees differed markedly in their movement behavior in relation to pre and post nest establishment. Bees moved over larger areas, and mostly within blueberry fields, before nest establishment. In contrast, after establishing the nest, the bees preferred the edges near forest plantations and they changed the nutritional resources to prefer wild floral species. Our study is the first to track queen bumblebee movements in an agricultural setting and relate movement changes across time and space with pollen resource availability. This study provides insight into the way bumblebee queens use different habitat elements at crucial periods in their lifecycle, showing the importance of mass flowering crops like blueberry in the first stages of queen’s lifecycle, and how diversified landscapes help support bee populations as their needs changes during different phases of their lifecycle.

DISTRIBUTION OF WATER POLLUTION AREAS BASED ON THE KERNEL DENSITY METHOD FOR LAPINDO MUDFLOW DISASTER OF SIDOARJO

The Lapindo mudflow disaster in Sidoarjo Regency which occurred on May 29, 2006 is a natural event that is classified as high risk. This disaster has a significant impact, especially for the water pollution caused by various conditions due to the Lapindo mudflow disaster. The purpose of the study is to mapping the location and analyze the distribution of water pollution based on the kernel density method. This research uses spatial approachment through descriptive-quantitative, qualitative and explorative methods. Primary and secondary data are both used in the research. The research results showed that water pollution is found on the north, west, south and east sides of the Lapindo mudflow disaster area in villages in three sub-districts namely Porong, Tanggulangin, and Jabon with the highest radius of pollution risk maximum of 1 Km from the center of the mudflow. The results of the distribution of water pollution areas based on the method of kernel density showed that the distribution of the highest water pollution leads to the north and west sides of the center of the mudflow.

Mobile home residence as a risk factor for adverse events among children in a mixed rural-urban community: A case for geospatial analysis.

Given the significant health effects, we assessed geospatial patterns of adverse events (AEs), defined as physical or sexual abuse and accidents or poisonings at home, among children in a mixed rural-urban community. We conducted a population-based cohort study of children (<18 years) living in Olmsted County, Minnesota, to assess geographic patterns of AEs between April 2004 and March 2009 using International Classification of Diseases, Ninth Revision codes. We identified hotspots by calculating the relative difference between observed and expected case densities accounting for population characteristics (; hotspot ≥ 0.33) using kernel density methods. A Bayesian geospatial logistic regression model was used to test for association of subject characteristics (including residential features) with AEs, adjusting for age, sex, and socioeconomic status (SES). Of the 30,227 eligible children (<18 years), 974 (3.2%) experienced at least one AE. Of the nine total hotspots identified, five were mobile home communities (MHCs). Among non-Hispanic White children (85% of total children), those living in MHCs had higher AE prevalence compared to those outside MHCs, independent of SES (mean posterior odds ratio: 1.80; 95% credible interval: 1.22-2.54). MHC residency in minority children was not associated with higher prevalence of AEs. Of addresses requiring manual correction, 85.5% belonged to mobile homes. MHC residence is a significant unrecognized risk factor for AEs among non-Hispanic, White children in a mixed rural-urban community. Given plausible outreach difficulty due to address discrepancies, MHC residents might be a geographically underserved population for clinical care and research.

Building a Series of Consistent Night-Time Light Data (1992–2018) in Southeast Asia by Integrating DMSP-OLS and NPP-VIIRS

Satellite-derived nighttime light (NTL) data from the Defense Meteorological Satellite Program’s Operational Linescan System (DMSP-OLS) and the Suomi National Polar-orbiting Partnership Visible Infrared Imaging Radiometer Suite (NPP-VIIRS) have been extensively used for monitoring human activities and urbanization processes. Differences of these two datasets in their spatial and radiometric properties make it difficult for a temporally consistent analysis using these two datasets together. In this article, we developed a new approach to integrate these two datasets and generated a temporally consistent NTL dataset from 1992 to 2018. First, we performed the pixel-level spatial resampling of VIIRS data using a kernel density method after preprocessing the raw VIIRS data. Second, we conducted a logarithmic transformation of the aggregated VIIRS data. Third, we proposed a sigmoid function between DMSP and processed VIIRS data to characterize their relationship. Using the proposed method, we generated a series of consistent DMSP NTL data in Southeast Asia from 1992 to 2018 and analyzed the dynamic of resulted NTL at different scales. The evaluations based on profile curves, spatial patterns, scatter correlations, and histograms, of NTLs, indicate that our approach can achieve a good agreement between DMSP and simulated DMSP data in the same year. Our approach offers the potential for generating a time series of global DMSP NTL data from 1992 to present, which can contribute a more continuous and consistent monitoring of human activities and a better understanding of the urbanization process.

Estimating load spectra probability distributions of train bogie frames by the diffusion-based kernel density method

The load spectra of a train bogie frame are used for bogie frame design, simulated fatigue testing and reliability evaluation and fatigue life prediction, however, modelling the bogie load spectrum is one of the most challenging tasks in this field. In this paper, diffusion-based kernel density estimation (DKDE) is used to obtain the load spectra distributions of train bogie frames. A full-scale B-type subway motor bogie and a full-scale high-speed EMU trailer bogie are analysed. DKDE, Gaussian kernel density estimation (GKDE) and parametric estimation are used to fit the distributions of the loads. Goodness-of-fit tests suggest that the distributions obtained by DKDE are close to the actual distributions. Fatigue damage are calculated to evaluate the distributions estimated by DKDE and GKDE. From the distributions obtained by DKDE, reasonable maximum load values and reasonable extrapolated load spectra are deduced.