Forecasting Climate Change Effects on Post Fire Hydrology and Geomorphology in the Western US

Few studies attempt to model the range of possible hydrologic and geomorphic responses following rainfall on burned basins because of the sparseness of data, and the coupled, nonlinear, spatial, and temporal relationships among post-fire landscape variables. This study used an unsupervised artificial neural network (ANN) to project data from 540 burned basins in the western United States [12] onto a SOM. The sparsely populated data set included independent numerical landscape categories (climate, land surface form, geologic texture, and post-fire condition), independent landscape classes (bedrock geology and state), and dependent initiation processes (runoff, landslide, and runoff-and-landslide combination) and responses (debris flows, floods, and no events). Clustering of the SOM neurons identified eight conceptual models of regional post-fire hydrologic and geomorphic

Climate-Change Forecast

Response

Initiation Process

Wet

Dry

Wet

Dry

Wet

Dry

Wet

Dry

Wet Dry

Wet Dry

Runoff +

Basin

State

Debris flow

Flooding

None

Landslide

Runoff

Landslide

1

MT

Yes

Yes

Yes

2

CA

Yes

Yes

Yes Yes

3

MT

Yes

Yes

4

CA

Yes

Yes

Yes

S

CO

Yes

Yes

6

CA

Yes

Yes

Yes

7

MT

Yes

Yes

Yes

8

MT

Yes

Yes

9

CO

Yes

Yes

10

CA

Yes

Yes

Yes

11

CO

Yes

Yes

12

CO

Yes

Yes

13

UT

Yes

Yes

Yes Yes

14

CA

Yes

Yes

Yes

1S

CO

Yes

Yes

20

ID

Yes

Yes

Yes

Yes

21

CO

Yes

Yes

22

NM

Yes

Yes

Yes

23

MT

Yes

Yes

Yes

24

CO

Yes

Yes

2S

CA

Yes

Yes

Yes Yes

26

MT

Yes

Yes

Yes

27

CA

Yes

Yes

28

CO

Yes

Yes

29

CO

Yes

Yes

30

CO

Yes

Yes

Yes

S7

CO

Yes

Yes

S8

MT

Yes

Yes

Yes

S9

ID

Yes

Yes

60

ID

Yes

Yes

Yes

Yes

Number of

events =

26

15

4

2

30

41

2

2

24 14

0 0

Fig. 3.3 The forecast effects of short-term climate variability on post-fire initiation processes and associated responses were evaluated using the self-organizing map. Wet and dry conditions were characterized by El Niño and La Niña associated precipitation events recorded in last 100 years

Fig. 3.3 The forecast effects of short-term climate variability on post-fire initiation processes and associated responses were evaluated using the self-organizing map. Wet and dry conditions were characterized by El Niño and La Niña associated precipitation events recorded in last 100 years landscape interaction. Stochastic cross-validation of the SOM demonstrated that initiation process and response predictions were globally unbiased.

A split-sample validation on 60 basins (not included in the training set) revealed that the simultaneous predictions of initiation process and response events were 78% accurate. Using this model, forecasts across post-fire landscapes revealed a decrease in the total number of debris flow, flood, and runoff events as climate shifted from wet (El Niño) to dry (La Niña) conditions (Fig. 3.3). Insight on individual basin changes and variability with respect to initiation process and response events also was revealed.

0 0

Post a comment