Clustering Analysis

Diday, Edwin; Simon, Jean Claude

doi:10.1007/978-3-642-96303-2_3

Cited by 178 publications

(48 citation statements)

References 21 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Considering more PCs to retain a larger fraction of the variance yields very similar weather types but slows the computations used in Part II. K-means is an iterative clustering procedure (Diday and Simon 1976) that consists in partitioning the data into k clusters so as to minimize the sum of variance within-cluster. The first step concerns choosing the appropriate number of clusters.…”

Section: A Definition Of the Number Of Weather Typesmentioning

confidence: 99%

Weather Types and Rainfall over Senegal. Part I: Observational Analysis

Moron

Robertson²,

Ward³

et al. 2008

Journal of Climate

View full text Add to dashboard Cite

A k-means cluster analysis is used to summarize unfiltered daily atmospheric variability at regional scale over the western Sahel and eastern tropical North Atlantic during the boreal summer season [JulySeptember (JAS)] 1961-98. The analysis employs zonal and meridional regional wind fields at 925, 700, and 200 hPa from the European Centre for Medium-Range Weather Forecasts reanalyses. An eight-cluster solution is shown to yield an integrated view of the complex regional circulation variability, without the need for explicit time filtering. Five of the weather types identified characterize mostly typical phases of westward-moving wave disturbances, such as African easterly waves (AEWs), and persistent monsoon surges, while the three others describe mostly different stages of the seasonal cycle. Their temporal sequencing describes a systematic monsoonal evolution, together with considerable variability at subseasonal and interannual time scales.Daily rainfall occurrence at 13 gauge stations in Senegal is found to be moderately well conditioned by the eight weather types, with positive rainfall anomalies usually associated with southerly wind anomalies at 925 hPa. Interannual variability of daily rainfall frequency is shown to depend substantially on the frequency of occurrence of weather types specific to the beginning and end of the JAS season, together with the number of persistent monsoon surges over the western Sahel. In contrast, year-to-year changes in the frequency of the weather types mostly associated with westward-moving waves such as AEWs are not found to influence seasonal frequency of occurrence of daily rainfall substantially.The fraction of seasonal rainfall variability related to weather-type frequency is found to have a strong relationship with tropical Pacific sea surface temperatures (SSTs): an El Niño (La Niña) event tends to be associated with a higher (lower) frequency of dry weather types during early and late JAS season with enhanced trade winds over the western Sahel, together with lower (higher) prevalence of persistent monsoon surges. The component of seasonal rainfall variability not related to weather-type frequency is characterized by changes in rainfall probability within each weather type, especially those occurring in the core of the JAS season; it exhibits a larger decadal component that is associated with an SST pattern previously identified with recent observed trends in Sahel rainfall.

show abstract

Section: A Definition Of the Number Of Weather Typesmentioning

confidence: 99%

Weather Types and Rainfall over Senegal. Part I: Observational Analysis

Moron

Robertson²,

Ward³

et al. 2008

Journal of Climate

View full text Add to dashboard Cite

show abstract

“…Some methods were developed to measure the proximity for heterogeneous type patterns: [14] proposes a combination of a modified Minkowski metric for continuous features and a distance for nominal attributes. A variety of other metrics have been reported in [15,16] for computing the similarity between patterns represented using quantitative as well as qualitative features.…”

Section: B Similarity Measuresmentioning

confidence: 99%

Pattern Recognition Approach in Multidimensional Databases: Application to the Global Terrorism Database

Salem¹,

Naouali²

2016

ijacsa

View full text Add to dashboard Cite

Abstract-This paper presents a pattern recognition approach in multidimensional databases. The approach is based on a clustering method using the distance measurement between a reference profile and the database observations. Two distance measurements will be proposed: an adaptation of the Khi 2 formula to the multidimensional context, extracted from the Multiple Correspondence Analysis (MCA), and the Euclidean distance. A comparison between the two distances will be provided to retain the most efficient one for the multidimensional clustering context. The proposed approach will be applied to a real case study representing armed attacks worldwide stored in the Global Terrorism Database (GTD).

show abstract

“…The NCAR SLP standardized anomalies of the 93.9% of days without missing values are weighted by the cosine of latitude and compressed onto the 11 leading empirical orthogonal functions (EOFs) that account for 90% of the total variance. The standard k-means algorithm (Diday and Simon, 1976) is applied to the 11 principal components to extract five clusters, in accordance with previous studies (Vautard, 1990;Michelangeli et al, 1995;Plaut and Simonnet, 2002;Moron and Plaut, 2003;. Two hundred cluster analyses were performed with random seeds, and the classifiability index (Michelangeli et al, 1995;Plaut and Simonnet, 2002) that measures the average similarity within the 200 sets of clusters is used to select the best partition.…”

Section: Weather Classificationmentioning

confidence: 99%

Changes in atmospheric circulation over the North Atlantic and sea‐surge variations along the Belgian coast during the twentieth century

Ullmann

Monbaliu

2009

Intl Journal of Climatology

View full text Add to dashboard Cite

Wintertime (October-to-March period) 99th percentile of sea level (sea surges) at Ostend has increased at a rate of +3 mm/year (+1 mm/year) from 1925 to 2000. Relationships between daily sea surge at Ostend and five weather regimes -Zonal (ZO), East Atlantic (EA), Greenland Above (GA), Blocking (BL), and Atlantic Ridge (AR) -over the northeast Atlantic and Europe (40°W-40°E, 30°-70°N) are analysed during the period of 1925-2000. More than 70% of sea surges ≥65 cm occur during the AR weather regime, ahead of low pressure travelling on a northern track from Iceland to Scandinavia. The relationships between monthly/wintertime frequency of the AR weather regime and 99th percentile of sea surge at Ostend tend to strengthen during the twentieth century: for example, correlation between wintertime frequencies of AR and 99th percentile of sea surge increases from 0.21 in 1925-1950 to 0.74 in 1975-2000. This increase is associated with the pressure rise over the near Atlantic, between (30°W and 15°W), (30°N and 50°N), leading to an increase in the frequency of strong surge-related pressure gradient during AR days.

show abstract

Clustering Analysis

Cited by 178 publications

References 21 publications

Weather Types and Rainfall over Senegal. Part I: Observational Analysis

Weather Types and Rainfall over Senegal. Part I: Observational Analysis

Pattern Recognition Approach in Multidimensional Databases: Application to the Global Terrorism Database

Changes in atmospheric circulation over the North Atlantic and sea‐surge variations along the Belgian coast during the twentieth century

Contact Info

Product

Resources

About