landslide warning: development of the hong kong system
DESCRIPTION
Landslide Warning: Development of the Hong Kong System. Major Landslide at Maracas, Trinidad - Late December 2002. Why Do Landslides Occur?. Slope Profile Geology Material Properties Water Ground Water Surface Water. How can we predict whether a Landslide will occur?. Survey the Profile - PowerPoint PPT PresentationTRANSCRIPT
Landslide Warning:
Development of the Hong Kong System
Major Landslide at Maracas, Trinidad - Late December 2002.
Why Do Landslides Occur?
•Slope Profile
•Geology
•Material Properties
•Water
•Ground Water
•Surface Water
•Survey the Profile
•Test Material Properties
•Estimate / measure where water table is.
•Analyse Stability using
•Infinite Slope
•Method of Slices
How can we predict whether a Landslide will occur?
> Factor of Safety
if < 1 slope is likely to fail
if > 1 slope should be stable
•Is it realistic to do a full analysis on all Landslides to this detail?
•In some critical cases …….YES
•Impossible to analyse all slopes under prevailing climatic conditions to give adequate warning
Some Questions
•Injury•Death•Economic Loss•Disruption to Transport Links
Consequences of Landslides
•Landslide Warnings.
•Categorise slopes
•Landslide Preventative Measures (LPM)
•Historically: Reactive Approach to Landslides•From 1977 became pro-active
Hong Kong Approach
Date 24 hrRainfall
StormType
Land-slides
Casualties Evacuation Category
25/08/76 416.2 STSEllen
314 57 2400 D
12/06/66 401.2 Trough 100 35 8500 D29/05/82 394.3 Trough 498 48 8000 D17/10/78 379.8 STS
Nina15 1 No data S
16/08/82 362.4 STS Dot 62 9 1500 D17/06/83 346.7 Trough 114 2 600 D27/09/65 338.2 TS
Agnes9 4 200 S
17/08/71 327.8 T Rose 10 7 No data S13/10/64 303.8 T Dot 8 39 8000 S13/06/68 287.2 Trough 10 27 200 S16/06/72 280.1 Trough >15 21 D18/06/72 275.1 Trough dozens 229
7800D
11/05/72 271.1 Trough dozens 0 6000 S30/04/75 150.0 Trough 21 3 No data S
Serious/Disaster Landslide Events 1963 - 1983
Number of Landslides as reported in South China Morning Post
Significant Landslide Events while NKT was in Hong Kong
Number ofReported Landslips
Rainfall (mm)DAY
Total AffectingSquatters
Antecedent15 days
24 hrrainfall
LandslipRED
Warning Issued
198223/4 6 5 71 59.928/4 6 5 180.8 36.87/5 9 5 321.6 55.9
29/5 223 107 188.8 258.4 *31/5 91 40 456.9 205.5 *2/6 28 12 663.7 22.1 *3/7 3 1 125.2 54.31/8 4 4 89.9 114.93/8 9 5 207.6 204.2 *
16/8 98 32 375.6 334.2 *18/8 8 2 669.9 47.6 *16/9 3 3 199.1 73.8 *Total 488 221 7 issued
Significant Landslide Events in 1982 in Hong Kong.
N. K. Tovey arrived in Hong Kong on 26th March 1982
Disaster
Disaster
Landslides in Kowloon East 28th - 31st May 1982
0
2
4
6
8
10
121930
1935
1940
1945
1950
1955
1960
1965
1970
1975
1980
Fre
qu
en
cy
Minor
Severe
Disaster
War Years - No Data
Missing Data 1974 - 1977
GCO Established
NKT arrives in Hong Kong
Frequency of Landslide Incidents
Requirements of A Landslide Warning System
It should:
1) provide sufficient warning of an event
•to alert general public
•to mobilise Emergency Services
•to open temporary Shelters
2) predict IN ADVANCE all serious EVENTS
3) minimise number of false alarms
Three criteria can be in conflict:
•How long should warning be?
•Longer the time, the less accurate will be prediction
Two Approaches
1. Detailed Warning - e.g. 1. Conduit Road
2. Warning based solely on Rainfall
automatic piezometer gives warning when ground water level gets above a critical level as determined by Slope Stability Analysis
Aim to give warning when a significant number of landslides are likely to occur.
(>10)
Background to Warning System
0
100
200
300
400
500
0 100 200 300 400 500 600 700 800Antecedent 15 day Rainfall (mm)
Rai
nfa
ll o
n D
ay o
f In
cid
ent (
mm
)
Minor Incident
Severe Incident
DisastrousIncident
Disaster
Minor Severe
Categorisation of Landslide Events by Lumb (1975)
Data points coloured RED and GREEN occurred after Lumb’s Paper
20 hours4 hours
Lan
dsl
ipP
red
icti
on
Cri
teri
a(L
PC
)
Warning Time (WT) (Rainfall predicted to reach LPC in 4 hours)
Cumulative Rainfall
Actual Cumulative Rainfall Predicted Cumulative Rainfall
Landslip Time (LT) (The time when first landslip is reported to FSD).
Criteria Time (CT)
The time when LPC are actually reached.
Rainfall Profile and Onset of Landslides
•Are Slopes more susceptible to failure if there has been prolonged rainfall?
•How should Antecedent rainfall Conditions be incorporated.
•Lumb (1975) - 15-day antecedent conditions.
•charts for Warning Purposes based both on Rainfall on Day AND Antecedent conditions.
•Most simple model uses simple cumulative 15-day antecedent rainfall.
•Could use a weighted system with days more distant weighted less.
•Lumb favoured simple approach.
ANTECEDENT CONDITIONS.
2 3 4 5 6 7 8 9 10 11 12 13 14 15
Day
24 – hour criteria
Cumulative Rainfall over previous 15 days
Cu
mm
ula
tiv e
Ra i
nfa
ll
Basis of Lumb’s Predictor
0 5 10
km
N
Semi-Automatic Gauges (Royal Observatory) Automatic Gauges (GCO)
Automatic Gauges (Royal Observatory) Built up areas
Distribution of Automatic Rain Gauges in Hong Kong (mid 1984)
0
50
100
150
200
250
300
350
400
450
0 100 200 300 400 500 600 700 800
Antecedent Rainfall (mm)
Rain
fall
on
Da
y (
mm
)
Minor Incident Severe Incident
Disastrous Incident RED Warning (1977 - 1979)
Amber Warning (1977 - 1979)
First Landslide Warning System (1977 - 1979)
AMBER and RED Warnings issued when predicted 24 hour rainfall would plot above relevant line.
A Problem: Difficult to use without direct access to Chart.
0
50
100
150
200
250
300
350
400
450
0 100 200 300 400 500 600 700 800
Minor Incident Severe Incident
Disastrous Incident RED Warning
AMBER Warning
Landslide Warning System 2: (1980 - mid 1983)
Advantage: Much easier to identify whether WARNING should be called - even when chart is not to hand.
Landslide Warning: 1/82Issued at 09:00 on 29/05/82Landslides reported:Total: 223Squatters: 107
0 100 200 300 400 500 600 700 800
Antecedent Rainfall in previous 15 days (mm)
Rainfall on Landslip Day (mm)
400
300
200
100
0
0020
16
12
09
04
Landslide Event 28 - 29th May 1982
Landslide Warning: 1/82Issued at 09:00 on 29/05/82Landslides reported:Total: 223Squatters: 107
0 100 200 300 400 500 600 700 800
Antecedent Rainfall in previous 15 days (mm)
Rainfall on Landslip Day (mm)
400
300
200
100
0
0020
16
12
09
04
Landslide Event 28 - 29th May 1982
Even with 24hr day plotting, the plot for 29th May should have been as follows
Landslide Warning: 1/82Issued at 09:00 on 29/05/82Landslides reported:Total: 223Squatters: 107
0 100 200 300 400 500 600 700 800
Antecedent Rainfall in previous 15 days (mm)
Rainfall on Landslip Day (mm)
400
300
200
100
0
09
04
0020
16
12
Situation with running 24 hr criterion
Landslide Event 28 - 29th May 1982
Criterion was reached at approx 03:00
BUT
1st Landslide was reported at 02:00 when rainfall was about 220mm
Even if Warning procedure has been operated correctly, warning would have been 1 hour too late!
09
04
0020
16
12
Landslide Warning: 1/82Issued at 09:00 on 29/05/82Landslides reported:Total: 223Squatters: 107
20
16
12
20
16
12
08
0400
00
16
12
08
04
16
0616
LW 2/8206:15 – 31/05/82Total: 91/ Sq: 40LW 4/82
11:00 – 03/08/82Total: 9Sq: 5
LW 6/8206:35 – 18/08/82*Total: 8Sq: 2
LW 3/8211:00 – 02/06/82*Total: 28/Sq: 12
LW 5/8205:50 – 16/08/82Total: 98Sq: 32
LW 7/8223:52 – 16/09/82Total: 3Sq: 3
0 100 200 300 400 500 600 700 800Antecedent Rainfall in previous 15 days (mm)
400
300
200
100
0
All Landslide Warning Incidents in 1982
Warning Criteria
Time
Warning Forecast
Gain
First
Landslip
No. Date (CT) Time (WT) (N) Time
1/82 29.05.82 0300 - 0400 0900 (-)b 0123
2/82 31.05.82a 0600 - 0700 0615 0 1351
3/82 02.06.82a not reached 1140 NA NR
4/82 03.08.82 1300 — 1400 1100 2 NR
5/82 16.08.82 0500 - 0600 0550 0 1009
6/82 18.08.82a not reached 0635 NA NR
7/82 16.09.82 not reached 2352 NA NR
1/83 27.03.83 2300 — 2400 2355 0 0011
2/83 08.04.83 not reached 1102 NA NR
3/83 17. 06.83 0800 - 0900 0745 1 0840
Performance of All LandSlip Warnings 1982 - 1983
Red Landslides with No Warning!
Green Landslide Warnings with Several Hours Warning
Blue Landslide Warnings with 1 Hour Warning
0
50
100
150
200
250
300
350
400
450
0 100 200 300 400 500 600 700 800
Antecedent 15-day Rainfall (mm)
24-h
ou
r R
ain
fall
du
rin
g E
vent
(m
m) Disastrous
Severe
Minor
Null Event
New 1983 Criteria
All Rainstorm Events: Daily Rainfall vs Antecedent Rainfall
Disastrous > 50 reported Landslides: Severe 10 - 50 LandslidesMinor < 10 Landslides : Null Event: No reported Landslides
Criteria for low antecedent rainfall reduced to conform to actual 1st landslide in Event 1/82
Landslide Warnings: The Problems
1. Antecedent Condition leads to confusion - (Incident 1/82)
2. Must use rolling 24 hour scheme
3. Previous Analysis (e.g. Lumb) has been based on 24 hr day basis
4. Total Rainfall in day will not generally be a good correlator as final cumulative 24 hr rainfall (whether day or rolling) will occur AFTER Landslides have occurred.
5. Some Landslides Events will occur after very low Antecedent Rainfall
6. Some Landslides Events occur after short periods of very intense rainfall.
7. It is difficult to predict with accuracy future rainfall.
Is it sensible to continue with Antecedent Rainfall Condition??
Landslide Warnings: The Final Approach
1. Abolish Antecedent Criteria - base solely on Rolling 24hr approach.
2. When Rainfall exceeds 100 mm in a period of 24 hours and is expected to exceed 175 mm (total) within 4 hours: CONSIDER issuing a LANDSLIDE WARNING.
If weather conditions suggest that Rainfall will cease shortly then issue could be delayed.
3. If Rainfall exceeds 175 mm then Landslides are likely and Warning should now be issued regardless of whether rain is likely to cease shortly
4. Landslide Warning should be issued regardless of above if rainfall in any one hour exceeds 70 mm in any one hour in Urban Area.
0
50
100
150
200
250
300
350
400
450
0 100 200 300 400 500 600 700 800
Antecedent 15-day Rainfall (mm)
24-h
ou
r R
ain
fall
du
rin
g E
ven
t (m
m)
Disastrous: > 50 LandslidesSevere: 10 - 50 LandslidesExisting CriteriaWarning LineLandslide Line
Existing Criteria Line - in use mid 1982 - mid 1984
Warning and Landslide Lines in use from mid 1984
Severe and Disastrous Landslide Events: with 1984 Scheme
The 1984 Warning Scheme
•Is simple to understand
•On average there would be 0 - 7 Warnings in a Year
•up to one third would be false alarms
•two-thirds would identify all serious/disastrous events correctly although half of the events would have less than 10 Landslides and thus be classified as minor.
•Has the option of Selected Warnings.
i.e. Warning Issued locally if 10 or more Automatic Gauges in Urban Area meet criteria
or a single gauge in New Territories.
BUT - first use of New Scheme……………..
Improvements to Selected Warning Schemes were introduced in 1999