Table 1 Performance of binomial generalised linear models predicting the probability of occurrence of flood anomalies in each region using portions of flood generation processes and different precipitation indices as explanatory variables.

From: Shifts in flood generation processes exacerbate regional flood anomalies in Europe

 

NORTH (n = 124)e

ATL (n = 477)

MED (n = 182)

CENTR (n = 570)

FP

FR

FP

FR

FP

FR

FP

FR

7-days precipitation max

 Likelihood ratio testa

p < 0.01

p < 0.01

p < 0.01

p < 0.01

p < 0.01

p < 0.01

p < 0.01

p = 0.18

 Pseudo R2 (McFadden)b

0.08

0.05

0.06

0.15

0.04

0.03

0.06

0.02

 Pseudo R2 (Lave/Efron)c

0.14

0.11

0.09

0.32

0.09

0.04

0.09

0.04

 AICbaseline-AICfulld

29.54

11.65

60.52

68.43

19.39

14.75

57.05

−1.15

1-day precipitation max

 Likelihood ratio test

p < 0.01

p < 0.01

p < 0.01

p < 0.01

p < 0.01

p < 0.01

p < 0.01

p = 0.40

 Pseudo R2 (McFadden)

0.08

0.05

0.06

0.12

0.05

0.03

0.01

0.02

 Pseudo R2 (Lave/Efron)

0.14

0.11

0.09

0.27

0.09

0.05

0.02

0.05

 AICbaseline-AICfull

29.54

11.65

68.19

54.97

22.10

17.23

12.65

0.03

  1. aSignificance of the covariates (i.e., frequencies of flood generation processes and precipitation maxima for a given year in each region).
  2. bAn analogy of R2 of linear regression models for generalised linear models based on log-likelihood72.
  3. cA correlation-based analogy of R2 of linear regression models for generalised linear models72.
  4. dAIC describes the trade-off between the goodness of fit and the simplicity of the model. Higher AIC values indicate higher information losses by the model74. Differences in Akaike Information Criterion (AICbaseline-AICfull) that are >2 indicate considerable improvement of the model76 accounting for flood generation processes and precipitation maxima compared to the baseline model (i.e., only uses intercept and assumes that the probability of occurrence of a flood anomaly in each given year equals the mean observed probability of the corresponding flood anomaly in the whole study period).
  5. eIn the Northern region model coefficients of both precipitation covariates are equal to zero, resulting in identical performance of the models including 7-days and 1-day precipitation maxima.