My talk on Sarasehan Geologi Populer, 16th March 2015, at Badan Geologi. This talk covers various open source tools for geological and hydrogeological analysis with focus on Cikapundung river case. Some examples of R code to extract hidden pattern in the data set, in order to explain natural phenomenon.

1. Geological analysis with open-source software:
case Cikapundung River
Event: Sarasehan Geologi Populer, Badan Geologi Indonesia
Dasapta Erwin Irawan
16th March 2015

2. Pre-talk quiz
Pls visit www.govote.at enter code 88 74 35, and answer
three yes/no questions

3. Part one: Introduction

4. Slide license
need attribution (BY)
for non-commercial purposes only (NC)
can be copied, modified (SA, share-alike)

5. A bit about me
Personal data
Name: Dasapta Erwin Irawan
Job: Lecturer/researcher at Groundwater Engineering Program,
ITB
Education (Geology, ITB): 1994-1998 (undergrad), 1999-2001
(Master), 2005-2009 (PhD)

6. A bit about me
Visiting program
Nov-Dec 2009
Center for Environmental Remote Sensing (CeRES), Chiba
University (2009)
Supervisor: Prof. Josaphat T.S. Sumantyo
Research: remote sensing for hydrological purposes
Feb 2014-Feb 2015
Faculty of Agriculture and Environment, University of Sydney
Supervisor: Dr. Willem Vervoort
Research: hydrological modeling in R

7. A bit about me
Research
hydrogeology
hydrochemistry
multivariate [statistical] analysis

8. A bit about me
Media social
Website: - R and Linux - Writing - SlideShare
Twitter: @dasaptaerwin
Email: d_erwin_irawan[at]yahoo[dot]com

9. Skills in Geology

10. Essential skills for geologist

11. Software skills (not free version)
office: Microsoft Office (Word, Excel, ppt) -> annual
subscription (from USD 10 per month)
citation and referencing: EndNote -> from USD 250
statistical: Minitab, SPSS, Statistica, Stata -> basic version
from USD 700 (2012)
spatial / GIS: ArcGIS, Mapinfo, etc -> annual subscription
(basic version from USD 100 per year)
Sources:
Openwetware
ESRI

12. Software skills (free equivalent)
office: OpenOffice or LibreOffice
citation and referencing: Zotero, Mendeley, etc
statistical: R and R Studio, Orange Data Mining, PSPP, etc
GIS: QGIS, GRASSGIS, R

13. Why open source?
free as breathing
mostly cross-platform (Linux, Mac, Win)
strong community, hence rapid development
supporting reproducibility

14. What is reproducibility in science?
Every step can be:
re-do
re-analysed and re-evaluate
re-developed

15. What is reproducibility in science?
Those principles are applied to:
data (items and locations)
software used in the analyses:
each software has distinct feature and algorithm
what would happen if not everyone could purchase the software?

16. End of part one

17. Part two: Cikapundung case

18. Slide license
need attribution (BY)
for non-commercial purposes only (NC)
can be copied, modified (SA, share-alike)

19. Background

20. Background
Cikapundung has important roles:
is one of the major water source for Bandung Basin: WTP
Dago Pakar = 40 L/sec
electrical generator (since 1923):
PLTA Bengkok = 3 MW
PLTA Dago = 0.7 MW
Drainase kota

21. Background

22. Background
Vast growth of settlements + landuse change -> declining water
quality (both river and groundwater).

23. Background

24. What do we know so far?
There types of groundwater and river water interactions (Lubis
and Puradimaja 2006)
isolated stream at Maribaya area (upstream)
effluent stream (or gaining stream) at Maribaya to Viaduct
segment (Bandung central)
influent stream (or losing stream) from Viaduct to Dayeuhkolot
some facts of springs and seepages at isolated segment
(Tanuwijaya 2014).

25. What do we know so far?

26. What do we know so far?
Conceptual model to mimic the interactions (Darul et.al
2014ab)
It confirms the Lubis Model

27. What do we know so far?

28. What do we know so far?

29. Our question
Does water quality reflect the interactions?

30. Our tools
R
Let’s do some [simple] analyses

31. Showing pairs analysis (bivariate analysis)

32. Data format
variables or measurements in columns
cases or samples in rows
no merged columns or rows
read also Data is the new soil

33. Why pairs analysis
equivalent to correlation matrix
the fastest way to see correlations between variables
pls bear in mind correlation does not always mean
causality

34. Our data

35. Our data
295 samples
From five years periode (1997, 1998, 2007, 2011, 2012)

36. Load
# load data
data <- as.data.frame(read.csv("BandungData.csv",
header = TRUE))
attach(data)
## The following object is masked from package:datasets:
##
## CO2

37. Data structure
# data structure
str(data)
## 'data.frame': 295 obs. of 33 variables:
## $ no : int 16 22 263 17 12 18 13 19 14 20 ...
## $ code : int 116 122 8 117 112 118 113 119 114 120 .
## $ year : int 1997 1997 1997 1997 1997 1997 1997 1997
## $ type : Factor w/ 2 levels "groundwater",..: 1 1 2 1
## $ x : num 785175 785168 799275 785175 785181 ...
## $ y : num 10752836 10752843 10753680 10752840 107
## $ distx : num 6897 6904 0 6897 6891 ...
## $ elv : int 1338 1336 1336 1320 1300 1247 1240 1230
## $ aq : Factor w/ 3 levels "breccias","clay",..: 3 3
## $ zone : Factor w/ 2 levels "eff","inf": 1 1 1 1 1 1
## $ ec : num 71.9 71.9 77 71.9 71.9 71.9 71.9 71.9 7
## $ ph : num 6.89 6.89 6.39 6.89 6.89 ...
## $ hard : num 11 11 26.4 11 11 11 11 11 11 11 ...
## $ tds : num 58.7 58.7 50 58.7 58.7 ...

38. pairs plot 1
pairs(data)
no
0 1.0 10740000 700 1.0 5 0 −50 0 0.0 0 0 0 0 0 20
0
0
code
year
2000
1.0
type
x
790000
y
distx
0
700
elv
aq
1.0
1.0
zone
ec
200
5
ph
hard
0
0
tds
temp
15
−50
eh
Q
0
0
Ca
Mg
0
0.0
Fe
Mn
0.00
0
K
Na
0
0
CO3
HCO3
0
0
CO2
Cl
0
0
SO4
NO2
0.0
0
NO3
SiO2
10
20
cumrain
0 2000 770000 0 1.0 200 0 15 0 0 0.00 0 0 0 0.0 10 40
40
lag1

39. pairs plot 1
ugly, too small
no legend and axis
we need to tweak it: group the variables and change plot code

40. pairs plot 1
pairs(group1,labels=colnames(group1),
main="Physical parameter",
pch=21, bg=c("red", "blue")
[unclass(data$type)],
upper.panel=NULL)
legend(x=0.6, y=0.8, levels(data$type),
pt.bg=c("red", "blue"),
pch=21, bty="n", ncol=2,
horiz=F)

41. Grouping variables
# group data
# Data group 1: Physical parameters
group1 <- data[,c("x", "y", "elv", "aq", "ec", "ph",
"hard", "tds", "temp", "eh", "Q")]
# Data group 2: Cation
group2 <- data[,c("x", "y", "elv", "Ca",
"Mg", "Fe", "Mn", "K", "Na")]
## Data group 3: Anions (unit = ppm)
group3 = data[,c("x", "y", "CO3", "HCO3",
"CO2", "Cl", "SO4", "NO2",
"NO3", "SiO2")]

42. Pairs plot group 1 (physical parameters)
770000
770000
x
10735000
y
700
elv
1.0
aq
200
ec
58
ph
080
hard
0
tds
1535
temp
−50
eh
770000
08
10735000 700 1.0 3.0 200 5 8 0 80 0 15 35 −50 0 8
08Q
Physical parameter
groundwater river

43. Pairs plot group 2 (cations )
770000
770000
x
10735000
y
700
elv
0100
Ca
030
Mg
0.01.0
Fe
0.00
Mn
040
K
770000
0150
10735000 700 0 80 0 30 0.0 1.0 0.00 0.30 0 40 0 150
0150Na
Cations
groundwater river

44. Pairs plot group 3 (anions )
770000
770000
x
10735000
y
025
CO3
0500
HCO3
0150
CO2
0150
Cl
0250
SO4
0.02.0
NO2
080
NO3
770000
1070
10735000 0 25 0 500 0 150 0 150 0 200 0.0 2.0 0 60 10 60
1070SiO2
Anions
groundwater river

45. Showing PCA analysis (multivariate analysis)

46. Why PCA (Principle Component Analysis)?
nature embeds multivariable process
has been widely used and developed since the 60’s
simple, straighforward, nearest neighbour (cluster) principles
offers nice visualisation

47. [Simple] codes
# install library
install.packages("pcaMethods") # for PCA
install.packages("gridExtra") # for plot lay out
# load library
library(pcaMethods) # for PCA
library(gridExtra) # for plot lay out
# run PCA
pca1 <- pca(group1,
method = "svdImpute",
scale = "uv",
center = T,
nPcs = 3,
evalPcs = 1:3)

48. [Simple] codes
# evaluate results
summary(pca1) # result summary
sDev(pca1) # extracting eigenvalues
plot(sDev(pca1)) # plotting eigenvalues
loadings(pca1) # plot loadings
scores(pca1) # plot scores
## Loading required package: Biobase
## Loading required package: BiocGenerics
## Loading required package: parallel
##
## Attaching package: 'BiocGenerics'
##
## The following objects are masked from 'package:parallel'
##
## clusterApply, clusterApplyLB, clusterCall, clusterEv
## clusterExport, clusterMap, parApply, parCapply, parL
## parLapplyLB, parRapply, parSapply, parSapplyLB

49. Results: summary PCA1
## svdImpute calculated PCA
## Importance of component(s):
## PC1 PC2 PC3
## R2 0.258 0.1672 0.1257
## Cumulative R2 0.258 0.4251 0.5509

50. Results: summary PCA2
## svdImpute calculated PCA
## Importance of component(s):
## PC1 PC2 PC3
## R2 0.2891 0.1392 0.1064
## Cumulative R2 0.2891 0.4283 0.5347

51. Results: summary PCA3
## svdImpute calculated PCA
## Importance of component(s):
## PC1 PC2 PC3
## R2 0.1991 0.1337 0.09922
## Cumulative R2 0.1991 0.3327 0.43194

52. Results: Extract Eigenvalues PCA1
1.0 2.0 3.0
1.21.31.41.51.6
Principal Component
Variance
1.0 2.0 3.0
1.11.21.31.41.51.61.71.8
Principal Component
Variance
1.0 2.0 3.0
1.21.31.41.51.6
Principal Component
Variance

53. Results: plot PCA Group 1
−2 0 2 4
−2024
PC 1
R^2 = 0.26
−2024
PC 2
R^2 = 0.17
−2 0 2 4
−3−2−1012
−2 0 2 4 −3 −2 −1 0 1 2
−3−2−1012
PC 3
R^2 = 0.13
o
o
groundwater
river water

54. Results: plot PCA Group 2
−2 0 2 4 6
−20246
PC 1
R^2 = 0.29
−20123
PC 2
R^2 = 0.14
−2 0 2 4 6
−4−2012
−2 0 1 2 3 −4 −2 0 1 2
−4−2012
PC 3
R^2 = 0.11
o
o
groundwater
river water

55. Results: plot PCA Group 3
−2 0 2 4 6
−20246
PC 1
R^2 = 0.2
−20246
PC 2
R^2 = 0.13
−2 0 2 4 6
−3−2−1012
−2 0 2 4 6 −3 −2 −1 0 1 2
−3−2−1012
PC 3
R^2 = 0.1
o
o
groundwater
river water

56. Results: loadings and scores Group1
−0.4 0.0 0.4
−0.4−0.20.00.20.4
Variable loadings
Group1
PC1
PC2
distx
ec
elv
ph
hard
tds
temp
eh
cumrain
lag1
−2 0 2 4
−2024
Case scores
Group1
PC1
PC2
Water type:
Groundwater
River Water

57. Results: loadings and scores Group2
−0.2 0.2
−0.6−0.4−0.20.00.20.4
Variable loadings
Group2
PC1
PC2
distx
ec
elv
Ca
Mg
Fe
Mn
K
Na
cumrain
lag1
−2 2 4 6
−2−10123
Case scores
Group2
PC1
PC2
Water type:
Groundwater
River Water

58. Results: loadings and scores Group3
−0.2 0.2
−0.4−0.20.00.20.4
Variable loadings
Group3
PC1
PC2
distx
ec
elv
CO3
HCO3
CO2
ClSO4
NO2
NO3
SiO2
cumrain
lag1
−2 0 2 4 6
−20246
Case scores
Group3
PC1
PC2
Water type:
Groundwater
River Water

59. spatial analysis (bubble plot)

60. why bubble plot?
shows spatial variation as well as values distribution
simple and straigtforward visualisation

61. [simple] codes
# load library (assuming all libraries are installed)
library(gstat)
library(sp)
library(rgdal)
library(latticeExtra)
# open and load data
df <- read.csv("BandungData.csv", header=TRUE)
# convert xy values as coordinates
coordinates(df) <- ~ x + y

62. [simple] codes
# make bubble plot
bubbleCa <- bubble(data, zcol="Ca",
xlab="X coord", ylab="Y coord",
main="Bubble plot Ca",
col = data$zone,
scales=list(tck=0.5))
print(bubbleCa)

63. Process
## rgdal: version: 0.9-1, (SVN revision 518)
## Geospatial Data Abstraction Library extensions to R succ
## Loaded GDAL runtime: GDAL 1.10.1, released 2013/08/26
## Path to GDAL shared files: /usr/share/gdal/1.10
## Loaded PROJ.4 runtime: Rel. 4.8.0, 6 March 2012, [PJ_VER
## Path to PROJ.4 shared files: (autodetected)

64. Results: groundwater end member (Ca)
Bubble plot Ca
X coord
Ycoord
10740000
10750000
10760000
770000 780000 790000 800000
0.7
6.75
17.68
31.475
138.6

65. Results: groundwater end member (Mg)
Bubble plot Mg
X coord
Ycoord
10740000
10750000
10760000
770000 780000 790000 800000
0.7
2.47
6.1
13.1
42.08

66. Results: river end member (Na)
Bubble plot Na
X coord
Ycoord
10740000
10750000
10760000
770000 780000 790000 800000
2.65
6.7
14.18
27
166

67. Results: river end member (K)
Bubble plot K
X coord
Ycoord
10740000
10750000
10760000
770000 780000 790000 800000
0.2
2.1
4.1
6.4
57.5

68. Results: contamination signature (NO3)
Bubble plot NO3
X coord
Ycoord
10740000
10750000
10760000
770000 780000 790000 800000
0
0.6
3.6
5.2
93.7

69. Results: contamination signature (NO2)
Bubble plot NO2
X coord
Ycoord
10740000
10750000
10760000
770000 780000 790000 800000
0
0
0
0.02
2.11

70. Results: contamination signature (SO4)
Bubble plot SO4
X coord
Ycoord
10740000
10750000
10760000
770000 780000 790000 800000
0
0.6
6.1
21.515
249

71. Results: contamination signature (Cl)
Bubble plot Cl
X coord
Ycoord
10740000
10750000
10760000
770000 780000 790000 800000
1
5.61
12.12
41.97
164

72. Remarks
higher mineral concentration in river water than groundwater
should have occured in effluent flow.
higher mineral concentration in groundwater than river water
should have occured in influent flow.
both natural indications are not detected, except for NO2.

73. Remarks
the anomaly is due to dilution effect.
dilution overides enrichment effect.
the opposite would happen if sampling is conducted in dry
season.
possibility of different catchment between groundwater and
river water.

74. Closing
Future research opportunities:
to add more data in different locations along river bank, taken
in both rain and dry season.
more exploratory statistical analysis, eg: multiple regression
tree to extract data pattern.

75. Main references
Lubis, RF and Puradimaja, DJ, 2006, Hydrodynamic
relationsships between groundwater and river water:
CIkapundung river stream, West Java, Indonesia
Darul, A, Irawan, DE, and Trilaksono, NJ, 2014a, Groundwater
and river water interaction on Cikapundung River: Revisited,
International Conference on Math and Natural Sciences, ITB.
Darul, A, 2014b, Model konseptual interaksi air tanah dan air
sungai di bantaran S. Cikapundung, Bandung, Jawa Barat,
Tesis S2, Supervisor: Dr. Dasapta Erwin Irawan dan
Dr. Nurjanna Joko Trilaksono.
Tanuwijaya, ZAJ, 2014, Identifikasi interaksi air sungai dan air
tanah di DAS Cikapundung, Disertasi, Geologi Universitas
Padjadjaran.

76. These slides were made using open-source tools
Ubuntu Linux (14.04)
R
Dia flowcharter
Gimp image editor

77. More resources
Me on Wordpress
Me on Blogger