Melissa Martin and Julia Jones, PhD. Oregon State University
Jan 19, 2016
Melissa Martin and Julia Jones, PhD.Oregon State University
O R E G O N
study basinsstreams
44°N
5 km
LookoutCreek
climate stationAndrews stream gage
Primet CS-2
WS 2
WS 8
WS 9
Vanmet
Hi-15
boundaries ofLookout Creek
WS 2 WS 8 WS 9
Gradient > 50% 25% > 60%
Slope length Median 150 m Median 90 m Median 60 m
Size 60 ha 15 ha 9 ha
Snow presence 6 mo. at high elev 6 mo. 1-2 weeks
Order Second order basin First order basin First order basin
Storm events 67 WR, 32 DR, 113 WRS (212)
14 WR, 7 DR, 23 WRS (44)
226 WR, 109 DR, 34 WRS (369)
peak discharge
precipitationbeforehydrographrise
event precipitation
centroid lag
start lag time to peak
event duration
maximum 15-minute precipitation intensity
baseflowat beginning
of hydrographrise
peak dischargedepth
time
Fig. 4
Compare peak discharge and precipitation for WR, DR, WRS events for each watershed
Examine relationship of hydrograph variables to intensity/timing of precipitation, under the three soil conditions listed above for each watershed
How are these different across the watersheds?
Why consider PCA? What is PCA?
- Explains variance-covariance structure
- Good for data reduction The ith PC is given by
Proportion of total population variance due to the kth PC is equal to
1 1 2 2' ...i i i i ip pY e X e X e X e X
( ) ' 1, 2,...,
( , ) ' 0i i i i
i k i k
Var Y i p
Cov Y Y i k
e Σe
e Σe
1 2 ...k
p
Examine the 10 hydrograph variables for normality
Chi-square plot for WS2 (untransformed variables)
sort(d)
Qch
i
0 20 40 60
51
01
52
02
5
Chi-square plot for WS2 (log transformed variables)
sort(d2)
Qch
i2
0 20 40 60 80 100
51
01
52
02
5
Obtain the coefficients of the PC’sVariable
ltotalprecip 0.400 0.028 -0.436 0.144
lmaxprecip -0.058 0.162 -0.465 -0.833
lprerunoff 0.172 -0.623 -0.166 -0.190
lstartlag 0.135 -0.652 -0.037 0.023
lcentroidlag 0.287 -0.221 0.330 -0.185
ltimetopeak 0.372 0.054 0.015 0.081
leventdur 0.445 0.124 -0.055 0.070
lqbase 0.215 0.062 0.599 -0.440
lqpeak 0.400 0.173 -0.244 0.077
lrunoffratio 0.406 0.243 0.183 -0.051
Variance 4.198 1.868 1.217 0.941
cum % of total var.
42.0 61.2 73.4 82.8
1e 2e 3e 4e
]734.1)[ln(406.0...]644.2))[ln(max058.0(]616.4)[ln(400.0ˆ1 orunoffratiprecipptotalpreciy
]734.1)[ln(243.0...]644.2)[ln(max162.0]616.4)[ln(028.0ˆ 2 orunoffratiprecipptotalpreciy
How many PC’s to choose?
Four PC’s are included in all principal component analyses
i
pro
p
2 4 6 8 10
0.0
0.2
0.4
0.6
0.8
1.0
Screeplot for WS2
Biplots: A 2-dimensional representation
Comp.1
Co
mp
.2
-0.2 -0.1 0.0 0.1 0.2
-0.2
-0.1
0.0
0.1
0.2
1
2
3
4
5
6
7
8
910
11
1213
14 1516
17
18
19
20
21
22
23
24 25 26
27
2829
3031
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
5051 52
5354
55
56
57
58
59
60
61
62
63
6465
66
6768 69
70
71
72
73
74 75 76
77
78
79
80
81
82
83
84
8586
8788
89
90
91
92
93
94
95
9697
98
99
100
101
102
103
104
105
106
107
108109
110
111
112
113114 115
116
117
118
119
120
121
122123
124
125
126
127
128
129
130 131
132
133
134135
136
137
138
139
140
141
142
143
144
145
146
147
148
149150
151
152153
154
155
156
157
158
159 160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175176
177
178
179
180
181
182183
184
185
186
187
188189
190
191192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
-10 -5 0 5 10
-10
-50
51
0
ltotalprecip
lmaxprecip
lprerunofflstartlag
lcentroidlag
ltimetopeakleventdur
lqbase
lqpeaklrunoffratio
Biplot for storm events for WS2
-4 -2 0 2 4
y2hat
-20
24
y1h
atScatterplot of (y2hat,y1hat) for WS2
wrdrwrs
Storm event type for WS2
WS2 (60 ha) WS8 (15 ha) WS9 (9 ha)Gradient > 50% 25% > 60%
Slope length Median 150m Median 90m Median 60m
Snow presence 6 mo. at high elev 6 mo. 1-2 weeks
PC1 Magnitude of storm (no maxprecip)
Magnitude with maxprecip, no qbase
Magnitude, no qbase
PC2 Timing Antecedent vs. timing “Full” timing
PC3 Antecedent wetness vs. totalprecip and maxprecip
Centroidlag, qbase vs. maxprecip
Antecedent wetness
Event type PCA Fairly consistent-for DR events, PC3 is measure of maxprecip
Not consistent, but sample sizes are small
Fairly consistent-for WR events, PC3 is measure of peak discharge (instead of qbase)
Not good for WS9 Not good for WS8
Chi-square plot for WS9 (log transformed variables)
sort(d2)
Qch
i2
10 20 30 40
51
01
52
02
5
Chi-square plot for WR events in WS8
sort(d3)
Qch
i3
7 8 9 10 11 12
51
01
52
0
Statistical models: Factor analysis, two-way MANOVA, blocking, restricted date ranges, PCA regression
Hydrograph properties: Getting more storm events for WS8 (GetPQ), creating software for HJA
Slide 1: Photograph courtesy of Julia Jones. Slide 3: HJ Andrews Experimental Forest figure
from Perkins and Jones, Snow and physiographic controls, 1997.
- Information about the watersheds and event classification: Perkins and Jones, Snow and physiographic controls, 1997.
Slide 4: Hydrograph figure from Perkins and Jones, Snow and physiographic controls, 1997.
Slide 6: The result quoted is Result 8.1, found in Johnson, Richard A. and Wichern, Dean W. Applied Multivariate Statistical Analysis, 2002. Fifth edition, Prentice Hall, Upper Saddle River, New Jersey. Page 428.
Desiree Tullos for creating the EISI program Julia Jones and Tom Dietterich for
mentoring help on this research project Reed Perkins for running GetPQ to create
the hydrograph data set The entire HJA staff Fellow EISI participants