B Supporting information

B.1 Cross-correlation between Outcome and Predictors

B.2 Summary of Meteorology Model


Family: quasipoisson 
Link function: log 

Formula:
count ~ s(templ0, k = 4) + s(templ1, k = 4) + s(templ2, k = 4) + 
    s(templ3, k = 4) + s(rainl0, k = 4) + s(rainl1, k = 4) + 
    s(rainl2, k = 4) + s(rainl3, k = 4)

Parametric coefficients:
            Estimate Std. Error t value Pr(>|t|)    
(Intercept)  2.47558    0.01332   185.9   <2e-16 ***
---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Approximate significance of smooth terms:
            edf Ref.df      F  p-value    
s(templ0) 2.928  2.995 25.245 5.26e-16 ***
s(templ1) 2.893  2.990  6.253 0.000261 ***
s(templ2) 2.502  2.834  5.288 0.000921 ***
s(templ3) 2.969  2.998 30.083  < 2e-16 ***
s(rainl0) 2.997  3.000 23.098 8.96e-15 ***
s(rainl1) 2.487  2.807 39.393  < 2e-16 ***
s(rainl2) 3.000  3.000 44.397  < 2e-16 ***
s(rainl3) 2.981  2.999 15.184 8.77e-10 ***
---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

R-sq.(adj) =  0.283   Deviance explained = 31.9%
GCV = 4.9062  Scale est. = 5.2576    n = 2792

B.3 Summary of Short-term Lag Model


Family: quasipoisson 
Link function: log 

Formula:
count ~ s(count1, k = 4) + s(count2, k = 4) + s(count3, k = 4) + 
    s(count4, k = 4)

Parametric coefficients:
            Estimate Std. Error t value Pr(>|t|)    
(Intercept)  2.49879    0.01271   196.7   <2e-16 ***
---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Approximate significance of smooth terms:
            edf Ref.df       F  p-value    
s(count1) 2.911  2.993 287.059  < 2e-16 ***
s(count2) 1.916  2.300   7.412 0.000409 ***
s(count3) 1.027  1.053   3.933 0.048968 *  
s(count4) 1.000  1.000   1.260 0.261669    
---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

R-sq.(adj) =  0.399   Deviance explained =   41%
GCV = 4.2141  Scale est. = 4.5806    n = 2562

B.4 Summary of Long-term Lag Model


Family: quasipoisson 
Link function: log 

Formula:
count ~ s(count1, k = 4) + s(count2, k = 4) + s(count3, k = 4) + 
    s(count4, k = 4) + s(count5, k = 4) + s(count6, k = 4) + 
    s(count7, k = 4) + s(count8, k = 4) + s(count9, k = 4) + 
    s(count10, k = 4) + s(count11, k = 4) + s(count12, k = 4) + 
    s(count13, k = 4) + s(count14, k = 4) + s(count15, k = 4) + 
    s(count16, k = 4) + s(count17, k = 4) + s(count18, k = 4) + 
    s(count19, k = 4) + s(count20, k = 4) + s(count21, k = 4) + 
    s(count22, k = 4) + s(count23, k = 4) + s(count25, k = 4) + 
    s(count26, k = 4) + s(count27, k = 4) + s(count28, k = 4) + 
    s(count29, k = 4) + s(count30, k = 4)

Parametric coefficients:
            Estimate Std. Error t value Pr(>|t|)    
(Intercept)  2.43389    0.01817   133.9   <2e-16 ***
---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Approximate significance of smooth terms:
             edf Ref.df       F  p-value    
s(count1)  2.887  2.987 162.559  < 2e-16 ***
s(count2)  1.000  1.000   0.355 0.551404    
s(count3)  1.000  1.000   0.840 0.359665    
s(count4)  1.000  1.000   1.217 0.270258    
s(count5)  1.012  1.023   5.659 0.016666 *  
s(count6)  1.000  1.000   1.509 0.219571    
s(count7)  2.678  2.920   5.216 0.003936 ** 
s(count8)  2.804  2.969   5.651 0.002368 ** 
s(count9)  1.000  1.000   3.708 0.054465 .  
s(count10) 1.510  1.836   0.712 0.563002    
s(count11) 2.840  2.978   2.330 0.085147 .  
s(count12) 1.000  1.000   2.986 0.084322 .  
s(count13) 1.560  1.886   1.501 0.164029    
s(count14) 1.000  1.000   1.644 0.200113    
s(count15) 1.425  1.706   0.642 0.375933    
s(count16) 1.000  1.000   1.641 0.200553    
s(count17) 1.970  2.325   0.560 0.483933    
s(count18) 2.143  2.489   7.704 0.000245 ***
s(count19) 1.000  1.000   1.421 0.233549    
s(count20) 1.000  1.000   0.289 0.590906    
s(count21) 1.612  1.958   5.651 0.007767 ** 
s(count22) 1.000  1.000   1.395 0.237862    
s(count23) 2.693  2.926   8.966 1.48e-05 ***
s(count25) 1.000  1.000   9.825 0.001774 ** 
s(count26) 2.137  2.531   3.156 0.056200 .  
s(count27) 2.818  2.973   4.800 0.006031 ** 
s(count28) 1.000  1.000   4.739 0.029724 *  
s(count29) 1.546  1.868   3.735 0.018887 *  
s(count30) 2.449  2.769   5.328 0.004379 ** 
---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

R-sq.(adj) =  0.623   Deviance explained = 64.4%
GCV = 3.1307  Scale est. = 3.1523    n = 978

B.5 Summary of Optimal-term Lag Model


Family: quasipoisson 
Link function: log 

Formula:
count ~ s(count1, k = 4) + s(count2, k = 4) + s(count23, k = 4)

Parametric coefficients:
            Estimate Std. Error t value Pr(>|t|)    
(Intercept)  2.47649    0.01474     168   <2e-16 ***
---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Approximate significance of smooth terms:
             edf Ref.df       F p-value    
s(count1)  2.939  2.997 278.184  <2e-16 ***
s(count2)  1.067  1.131   3.958  0.0501 .  
s(count23) 2.370  2.715  34.242  <2e-16 ***
---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

R-sq.(adj) =   0.49   Deviance explained = 48.9%
GCV = 3.7406  Scale est. = 3.9826    n = 1723

B.6 Summary of Meteorology and Optimal-term Lag Model


Family: quasipoisson 
Link function: log 

Formula:
count ~ s(templ0, k = 4) + s(templ1, k = 4) + s(templ2, k = 4) + 
    s(templ3, k = 4) + s(rainl0, k = 4) + s(rainl1, k = 4) + 
    s(rainl2, k = 4) + s(rainl3, k = 4) + s(count1, k = 4) + 
    s(count2, k = 4) + s(count23, k = 4)

Parametric coefficients:
            Estimate Std. Error t value Pr(>|t|)    
(Intercept)   2.4237     0.0126   192.4   <2e-16 ***
---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Approximate significance of smooth terms:
             edf Ref.df       F  p-value    
s(templ0)  3.000  3.000  18.473 8.50e-12 ***
s(templ1)  2.541  2.850  14.254 2.05e-08 ***
s(templ2)  2.999  3.000  19.367 2.25e-12 ***
s(templ3)  1.000  1.000 102.734  < 2e-16 ***
s(rainl0)  2.897  2.988   3.830   0.0124 *  
s(rainl1)  2.975  2.998  93.704  < 2e-16 ***
s(rainl2)  3.000  3.000  44.644  < 2e-16 ***
s(rainl3)  2.812  2.969   3.231   0.0196 *  
s(count1)  2.941  2.997 149.230  < 2e-16 ***
s(count2)  2.661  2.915  23.874 2.03e-14 ***
s(count23) 1.874  2.249  19.171 1.68e-09 ***
---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

R-sq.(adj) =  0.638   Deviance explained = 66.4%
GCV = 2.5266  Scale est. = 2.6396    n = 1723

B.7 Summary of Meteorology, Optimal-term and Short-term Surrounding Lag Model


Family: quasipoisson 
Link function: log 

Formula:
count ~ s(templ0, k = 4) + s(templ1, k = 4) + s(templ2, k = 4) + 
    s(templ3, k = 4) + s(rainl0, k = 4) + s(rainl1, k = 4) + 
    s(rainl2, k = 4) + s(rainl3, k = 4) + s(count1, k = 4) + 
    s(count2, k = 4) + s(count23, k = 4) + s(surroundingCountVec1, 
    k = 4) + s(surroundingCountVec2, k = 4)

Parametric coefficients:
            Estimate Std. Error t value Pr(>|t|)    
(Intercept)  2.42054    0.01245   194.3   <2e-16 ***
---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Approximate significance of smooth terms:
                          edf Ref.df       F  p-value    
s(templ0)               2.972  2.998  17.499 3.14e-11 ***
s(templ1)               1.956  2.314   6.408 0.000859 ***
s(templ2)               2.998  3.000  20.736 3.17e-13 ***
s(templ3)               2.056  2.462  39.326  < 2e-16 ***
s(rainl0)               1.000  1.000   0.719 0.396444    
s(rainl1)               2.982  2.999  95.402  < 2e-16 ***
s(rainl2)               2.963  2.996  29.391  < 2e-16 ***
s(rainl3)               2.987  2.999  10.072 1.15e-06 ***
s(count1)               2.920  2.994 116.329  < 2e-16 ***
s(count2)               2.679  2.923  23.681 2.55e-14 ***
s(count23)              1.924  2.302  20.554 3.15e-10 ***
s(surroundingCountVec1) 2.722  2.945  16.210 4.80e-10 ***
s(surroundingCountVec2) 1.000  1.000   8.939 0.002830 ** 
---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

R-sq.(adj) =  0.652   Deviance explained = 67.3%
GCV = 2.4659  Scale est. = 2.5611    n = 1723

B.8 Meteorology, Optimal-term and Optimal-term Surrounding Lag Model


Family: quasipoisson 
Link function: log 

Formula:
count ~ s(templ0, k = 4) + s(templ1, k = 4) + s(templ2, k = 4) + 
    s(templ3, k = 4) + s(rainl0, k = 4) + s(rainl1, k = 4) + 
    s(rainl2, k = 4) + s(rainl3, k = 4) + s(count1, k = 4) + 
    s(count2, k = 4) + s(count23, k = 4) + +s(surroundingCountVec1, 
    k = 4) + s(surroundingCountVec2, k = 4) + s(surroundingCountVec12, 
    k = 4)

Parametric coefficients:
            Estimate Std. Error t value Pr(>|t|)    
(Intercept)   2.4178     0.0123   196.6   <2e-16 ***
---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Approximate significance of smooth terms:
                           edf Ref.df       F  p-value    
s(templ0)                2.925  2.991  25.017 7.55e-16 ***
s(templ1)                2.826  2.958   7.179 0.000176 ***
s(templ2)                3.000  3.000  24.866 9.21e-16 ***
s(templ3)                2.390  2.761  27.940 6.69e-14 ***
s(rainl0)                1.574  1.865   0.656 0.536832    
s(rainl1)                3.000  3.000 107.060  < 2e-16 ***
s(rainl2)                2.872  2.969  14.988 8.42e-10 ***
s(rainl3)                3.000  3.000  17.948 1.80e-11 ***
s(count1)                2.914  2.994 114.425  < 2e-16 ***
s(count2)                2.673  2.920  20.566 1.80e-12 ***
s(count23)               1.821  2.194  24.118 1.43e-11 ***
s(surroundingCountVec1)  2.649  2.914  22.053 1.93e-13 ***
s(surroundingCountVec2)  1.000  1.000   8.352 0.003900 ** 
s(surroundingCountVec12) 2.761  2.956  13.242 1.81e-08 ***
---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

R-sq.(adj) =  0.658   Deviance explained = 68.1%
GCV = 2.4149  Scale est. = 2.4807    n = 1723

B.9 Social Data Included


Family: quasipoisson 
Link function: log 

Formula:
count ~ s(templ0, k = 4) + s(templ1, k = 4) + s(templ2, k = 4) + 
    s(templ3, k = 4) + s(rainl0, k = 4) + s(rainl1, k = 4) + 
    s(rainl2, k = 4) + s(rainl3, k = 4) + s(count1, k = 4) + 
    s(count2, k = 4) + s(count23, k = 4) + s(surroundingCountVec1, 
    k = 4) + s(surroundingCountVec2, k = 4) + s(surroundingCountVec12, 
    k = 4) + offset(log(garbage)) + s(garbage)

Parametric coefficients:
            Estimate Std. Error t value Pr(>|t|)    
(Intercept) -6.16406    0.01228  -502.1   <2e-16 ***
---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Approximate significance of smooth terms:
                           edf Ref.df       F  p-value    
s(templ0)                2.922  2.991  25.056 7.23e-16 ***
s(templ1)                2.819  2.956   7.070 0.000202 ***
s(templ2)                2.999  3.000  24.666 1.22e-15 ***
s(templ3)                2.421  2.782  28.152 5.25e-14 ***
s(rainl0)                1.575  1.866   0.666 0.533065    
s(rainl1)                3.000  3.000 107.828  < 2e-16 ***
s(rainl2)                2.867  2.967  14.760 1.18e-09 ***
s(rainl3)                3.000  3.000  17.746 2.47e-11 ***
s(count1)                2.919  2.994 110.705  < 2e-16 ***
s(count2)                2.710  2.937  19.116 1.02e-11 ***
s(count23)               1.771  2.139  23.832 2.69e-11 ***
s(surroundingCountVec1)  2.706  2.938  23.221 2.77e-14 ***
s(surroundingCountVec2)  1.000  1.001   8.100 0.004481 ** 
s(surroundingCountVec12) 2.766  2.957  13.226 1.77e-08 ***
s(garbage)               3.070  3.838 327.587  < 2e-16 ***
---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

R-sq.(adj) =  0.659   Deviance explained = 74.6%
GCV = 2.4091  Scale est. = 2.4687    n = 1723

B.10 Initial Analyses Summary