26
P. Muñoz et al.
Table 2.2 Input data composition of the RF models and their parsimonious versions (4, 8, 12 and
24-h lead time) for the Tomebamba and Yanuncay catchments
Tomebamba catchment
Lead time
[hours]
Discharge
lags
Toreadora
lags
Chirimachay
lags
Virgen
lags
Total
features
4
8
24
15
10
60
4 a
8
9
9
9
38
8
8
32
23
19
85
8 a
8
15
15
15
56
12
8
36
27
23
98
12 a
8
18
18
18
65
24
15
48
39
35
140
24 a
15
21
21
21
81
Yanuncay catchment
Lead time
[hours]
Discharge
lags
Huizhil
lags
Izhcayrrumi
lags
Ventanas
lags
Total
features
4
21
8
21
20
73
4 a
21
6
6
6
42
8
21
16
29
28
97
8 a
21
10
10
10
54
12
21
20
33
32
109
12 a
21
12
12
12
60
24
21
32
45
44
145
24 a
21
13
13
13
63
a Parsimonious version
Table 2.3 Model performance of the RF models and their parsimonious versions (4, 8, 12 and 24-h
lead time) for the Tomebamba and Yanuncay catchments
Tomebamba
Yanuncay
Lead time
[hours]
Total features
[#]
NSE
Total features
[#]
NSE
Training
Test
Training
Test
4
60
0.9193
0.8604
73
0.9014
0.7572
4 a
38
0.9211
0.8682
42
0.898
0.7675
8
85
0.8486
0.7494
97
0.8558
0.6260
8 a
56
0.8441
0.7523
54
0.8572
0.6425
12
98
0.8131
0.6759
109
0.8131
0.4122
12 a
65
0.8074
0.6799
60
0.8161
0.4049
24
140
0.7541
0.538
145
0.7577
0.1159
24 a
81
0.7483
0.5454
63
0.7626
0.1189
a Parsimonious version
Précédent

- 32/353

Suivant