Introduction to the principle of geographical detector

Spatial differentiation Is one of the basic characteristics of geographical phenomena , Geo detector It is to detect the spatial differentiation of geographical elements , And a way to reveal its driving factors , Proposed by Wang Jinfeng and others .

Geographical detector is a tool for detecting and utilizing spatial differentiation , It mainly includes 4 part ： Differentiation and factor detection 、 Interaction detection 、 Risk area detection 、 Ecological Exploration . According to the principle and application fields of geographical detectors , This paper uses the first detector to analyze the impact of landform on the development of river network and the correlation between the two .

（1） Differentiation and factor detector ： Probe the research object Y Spatial differentiation of , And detect various factors X To what extent does it explain Y Spatial differentiation of . The degree of explanation is q Value to measure , Expression for ：

In style ,h=1,2,⋯,L As a variable Y Or factor X The layered , That is, classification or partition ;N_h and N They are layers h And the number of units in the region ;σ_h^{2 and σ}2 They are layers h And the whole region Y The variance of the value ;SSW and SST They are the sum of the intra layer variance and the total variance of the whole region .q The value range of is [0,1],q The greater the value, the more Y The more obvious the spatial differentiation of ; If layering is made up of variables X Generated , q A larger value indicates that the argument X Pair attribute Y The stronger the explanatory power of , On the contrary, the weaker . In extreme cases , When q The value is 1 when , Indicates the variable X Be able to completely control Y The spatial distribution of ; When q The value is 0 when , It indicates that the variable X And Y It doesn't matter ,q Values represent variables X Explained 100×q% Of Y.

q A transformation of values satisfies noncentral F Distribution ：

In style ,λ Is a non central parameter ,Y ̅_h For layer h The average of . According to the type 2.20 It can be used to test q Whether the value is significant .

（2） Interaction detection ： Identify different risk factors Xs Interaction between , That is, the evaluation factor X1 and X2 Whether it will increase or decrease the dependent variable when acting together Y The explanatory power of , Or these factors are right Y The influence of is independent . The method of evaluation is to calculate two factors respectively X1 and X2 Yes Y Of q value ： q(X1) and q(X2), And calculate their interaction （ Superimposed variable X1 and X2 The new polygon distribution formed by the tangency of two layers , chart 2） At the time of the q value ： q(X1 ∩ X2) , Also on q(X1)、q(X2) And q(X1 ∩ X2) Compare .

（3） Risk area detection ： It is used to judge whether there is a significant difference between the attribute mean values of the two sub regions , use t Statistics to test ：

（4） Ecological Exploration ： Used to compare two factors X1 and X2 Pair attribute Y Whether there is a significant difference in the impact of spatial distribution , With F Statistics to measure ：

See the literature for details ： Wang Jinfeng , Xuchengdong . Geo detector : Principle and prospect [J]. Journal of Geography ,2017,72(01):116-134.

There is also a software package for geographical detectors , Is in excel Use in , You can click the link to download it directly , Sample data is attached ： Geo detector software package