Methods, data and sources

Loss and AAL

AAL is the frequency-weighted sum of event losses. For park p: AAL_p = Σ_e f_e × L_p(e), where f_e is the annual frequency of event e and L_p(e) is the loss at park p. Each park uses a USD 100m standard asset (A01), an upper-bound assumption, not client book value.

Typhoon loss uses the Eberenz et al. (2021) South East Asia calibrated impact function (A06). Flood loss uses the report's five-point curve: 0/0.5/1/2/3 m → 0/15/35/65/100% (A07). The flood curve is an upper-bound screening assumption; it is not calibrated to site elevation, drainage or flood defences.

Event frequencies come from CLIMADA event tables (A08). The typhoon set has 3,890 base storms × 11 variants; the frequency sum is 93.37/year for RCP8.5 and 96.01/year for RCP4.5. These are treated as annual event probabilities for screening. The flood set has 480 events with frequency sum 1.0.

Stop-work, restriction and power rules

Thresholds combine a public meteorological definition with a report-specific operating mapping. 32.7 m/s is the lower bound of Beaufort 12 and 24.5 m/s of Beaufort 10 (NOAA WPC). Using them as park/land stop-work and port-closure lines is a report assumption, not a Vietnamese statutory rule. 0.3 m refers to Hanoi's road closure/ diversion practice above 30 cm ponding; applying it to all parks, ports and roads is a screening extrapolation.

The power score is a 1–5 relative rubric: 0.40 grid, 0.25 hydro dependence, 0.20 backup power, 0.15 ENSO pressure. It is not an outage probability. Scores are: VSIP Bac Ninh 5, Quang Chau 5, DEEP C 4, An Duong 4, VSIP III 3, Amata 3.

Data freeze and scenarios

The data freeze uses CLIMADA 6.1 Vietnam tropical-cyclone tracks (10synth, 150 arcsec) and river-flood hazard (150 arcsec). Typhoon scenarios: RCP8.5/2040 main and RCP4.5/2040 control. Flood scenarios: RCP8.5/2030–2050 main and RCP2.6/2030–2050 low-case envelope; the CLIMADA Vietnam flood set has no RCP4.5.

The 13 nodes are sampled at the nearest 150 arcsec flood grid point and the nearest 0.05° typhoon grid point. Match distances are 0.39–2.81 km. This is point screening, not a park-boundary maximum. The event tables contain 42,790 typhoon events (RCP8.5), 43,560 (RCP4.5) and 480 flood events per flood scenario.

10-year cost-benefit

ROI screens three packages: conservative (8% capex, 28% AAL reduction, 10% discount), base (6.5%, 40%, 6%) and active (5%, 52%, 6%). Annual opex is 1.5% of capex; horizon is 10 years. NPV = −CAPEX + Σ_t=1..10 (annual net benefit / (1+r)^t). Payback is undiscounted; over 10 years is shown as “no payback”.

The calculation uses AAL from the screening model and a USD 100m standard asset. It excludes downtime, brand, financing, residual value and insurance. A positive NPV means “worth diligence”, not “invest now”; it is not a quote or engineering study.

Assumptions A01–A16

Six-park AAL summary

Main: typhoon RCP8.5/2040 + flood RCP8.5/2030–2050. Low case: typhoon RCP4.5/2040 + flood RCP2.6/2030–2050. AAL is an upper-bound screen on a USD 100m standard asset.

ROI by park and scheme

NPV is not a quote. Negative NPV under the standard package does not mean “no investment”; smaller power, water and drainage measures may still be justified.

13-node annual disruption

Annual days = event frequency × assumed closure days. Figures are planning metrics, not calendar commitments. Wind and flood days use different definitions and should not be added beyond the total shown for screening.

ENSO watch indicators

Near-term watch only. ENSO is not included in AAL, not added to RCP8.5/2040 and not a loss forecast.

13-node peak wind

IBTrACS observed track + CLIMADA H08 model wind. Values are model output at the nearest 0.05° grid point, not node observations or return-period wind.

Source log

Internal project notes are excluded. Full assumption IDs and data nature are in the table above. Links are provided for traceability; source terms and licences remain with the publishers.