05 August 2014

Preliminary "mgcv" result

Dear friends,

I managed to add some other predictors: anions, cumulative monthly rain, and lagged-1 monthly rain.

The following table shows the updated result.
Based on the table, I made a conceptual model about the water flow and the interaction between predictors in the system. As I've mentioned before, the surficial processes have stronger control to water chemistry than lithology. Hopefully the following sketch can give more spatial sense of the area. The strongest process detected by mgcv is NO3 enrichment in the river water as it gains water from groundwater flow. The other process is dilution  effect by river water as shown by decreasing pattern for elements like Cl, SO4, and Mn towards river.



New results

GAM Predictant Predictors Significance Family Residuals AIC Deviance GCV Rsq Pattern TrendRiver
10 logEC te(x,y), elv, lithology, cumRain, lagRain Y, Y, N, N, Y gamma good 3794.627 16.200 0.448 0.140 clear decreasing
11 logCO3 te(x,y), elv, lithology, cumRain, lagRain Y, Y, N, N, N gaussian poor 886.045 25.800 1.171 0.229 clear decreasing
12 logHCO3 te(x,y), elv, lithology, cumRain, lagRain N, Y, N, N, Y gaussian good 778.806 40.300 0.819 0.362 not clear
13 logCO2 te(x,y), elv, lithology, cumRain, lagRain Y, Y, N, Y, N gaussian poor 1035.771 14.200 1.950 0.111 clear decreasing
14 logCl te(x,y), elv, lithology, cumRain, lagRain Y, Y, N, N, Y gaussian good 814.054 31.000 0.922 0.267 not clear
15 logSO4 te(x,y), elv, lithology, cumRain, lagRain Y, Y, N, N, N gaussian good 1129.933 32.700 2.686 0.296 clear decreasing
16 logNO2 te(x,y), elv, lithology, cumRain, lagRain Y, Y, N, N, Y gaussian good 778.314 62.200 0.819 0.589 clear decreasing
17 logNO3 te(x,y), elv, lithology, cumRain, lagRain Y, N, N, Y, N gaussian good 1168.619 42.500 3.062 0.399 clear increasing
18 logFe te(x,y), elv, lithology, cumRain, lagRain Y, N, N, N, N gaussian good 649.973 12.300 0.528 0.083 clear decreasing
19 logCa te(x,y), elv, lithology, cumRain, lagRain N, Y, N, N, Y gaussian good 785.075 30.300 0.833 0.281 not clear
20 logMg te(x,y), elv, lithology, cumRain, lagRain Y, Y, N, N, Y gaussian good 730.692 41.600 0.696 0.371 not clear
21 logMn te(x,y), elv, lithology, cumRain, lagRain Y, Y, N, N, N gaussian poor -72.300 79.100 0.046 0.774 clear decreasing
22 logNa te(x,y), elv, lithology, cumRain, lagRain Y, Y, N, N, Y gaussian good 641.144 39.500 0.514 0.353 clear decreasing
23 logK te(x,y), elv, lithology, cumRain, lagRain N, N, N, N, N gaussian good 592.524 18.400 0.434 0.166 not clear






----
Previous post

This post would be the continuation of the previous serial posts on GAM using mgcv package. As I have posted before, I am looking for predictors that explain the interaction between groundwater and river water in Cikapundung watershed (Bandung, West Java, Indonesia).

The following table is the preliminary result. I'll make further explanation.




GAM Predictant Predictors Significance Family Residuals AIC Deviance (%) Notes Trend
10 EC (x,y), elv, lithology yes, yes, no gamma good 3805.445 10.6 clear pattern, decreasing trend towards river E-W
11 logCO3 (x,y), elv, lithology yes, yes, no gaussian poor 889.724 24.5 clear pattern, decreasing trend towards river E-W
13 logCO2 (x,y), elv, lithology yes, yes, no gaussian poor 1046.908 11.8 clear pattern, decreasing trend towards river NE-SW
15 logSO4 (x,y), elv, lithology yes, yes, no gaussian poor 1132.996 32.9 clear pattern, decreasing trend towards river NW-SE
17 NO3 (x,y), elv, lithology yes, no, no gaussian poor 1388.982 8.38 clear pattern, decreasing trend towards river NW-SE
20 logMg (x,y), elv, lithology no, yes, no gaussian poor 780.82 26.4 clear pattern, decreasing trend towards river E-W
21 Mn (x,y), elv, lithology yes, yes, no gaussian poor -971.85 74.4 clear pattern, decreasing trend towards river N-S, NE-SW
16 logNO2 (x,y), elv, lithology yes, no, no gaussian poor 1183.261 43.3 clear pattern, increasing trend towards river NE, SE
18 Fe (x,y), elv, lithology no, no, no gaussian poor -70.72 2.44 clear pattern, increasing trend towards river NW-SE
22 logNa (x,y), elv, lithology yes, yes, no gamma good 649.918 33.3 clear pattern, increasing trend towards river E-W
12 logHCO3 (x,y), elv, lithology no, yes, no gaussian good 847.1502 22.8 no clear pattern
14 logCl (x,y), elv, lithology yes, yes, no gaussian good 828.088 27.4 no clear pattern
19 logCa (x,y), elv, lithology no, yes, no gaussian poor 838 17.6 no clear pattern
23 logK (x,y), elv, lithology no, yes, no gaussian good 594.923 18.3 no clear pattern
Post a Comment