Skip to content

climate.md

Latest commit

 

History

History
139 lines (107 loc) · 5.12 KB

climate.md

File metadata and controls

139 lines (107 loc) · 5.12 KB
Raw

Climate Scenario Analysis

Since 0.3.0, OTEX can project an OTEC site's performance under future climate conditions using CMIP6 ocean temperature scenarios. The implementation follows the IPCC-standard delta-method downscaling: historical CMEMS reanalysis is used as the high-resolution baseline, and a multi-model ensemble of CMIP6 GCMs supplies the time-mean warming (Δ) at each grid cell, which is then added to the observed time series before the techno-economic analysis runs.

Why delta-method?

T_future(lon, lat, t) = T_CMEMS_observed(lon, lat, t) + Δ(lon, lat)

           Δ(lon, lat) = ⟨T_GCM_future⟩ − ⟨T_GCM_baseline⟩

Two important properties:

  1. The spatial detail of CMEMS (1/12°) and the observed seasonality are preserved — only the climate state is shifted.
  2. GCM biases cancel in the difference, making Δ much more robust between models than absolute GCM temperatures.

The same mechanism is applied independently at the OTEC warm-water intake depth (~20 m) and cold-water intake depth (~1062 m). Surface waters typically warm faster than the deep ocean, so under climate change the ΔT thermal gradient widens, which generally reduces LCOE for OTEC.

Available scenarios

Scenario Description
historical (default) No delta applied — identical to the standard CMEMS pipeline.
ssp126 Strong mitigation, ~1.5 °C global warming target.
ssp245 Middle-of-the-road emissions.
ssp370 High emissions, no mitigation.
ssp585 Worst-case fossil-fuel-intensive trajectory.

Default ensemble

Three CMIP6 GCMs spanning the range of equilibrium climate sensitivity:

Model Institution ECS (°C)
MPI-ESM1-2-LR Max Planck Institute ~3.0
EC-Earth3 EC-Earth Consortium ~4.3
CanESM5 CCCma ~5.6

Override via --climate-models (CLI) or climate.models=(...) (Python).

Quick start

# Project Jamaica's 2020-2023 OTEC performance under SSP2-4.5 by 2050.
otex-regional Jamaica \
    --year-start 2020 --year-end 2023 \
    --climate-scenario ssp245 --climate-year 2050
from otex.regional import run_regional_analysis

otec_plants, sites = run_regional_analysis(
    studied_region='Jamaica',
    year_start=2020,
    year_end=2023,
    climate_scenario='ssp245',
    climate_year=2050,
)

Output files include a _<scenario>_<target_year> suffix:

Data_Results/Jamaica/Jamaica_2020-2023_ssp245_2050_100.0_MW_low_cost/
├── OTEC_sites_Jamaica_2020-2023_ssp245_2050_100.0_MW_low_cost.csv
├── OTEC_sites_yearly_Jamaica_2020-2023_ssp245_2050_100.0_MW_low_cost.csv
└── net_power_profiles_per_day_Jamaica_2020-2023_ssp245_2050_100.0_MW_low_cost.csv

The schema is identical to the baseline run — same LCOE, AEP, AEP_min/p50/max/std columns — but the values reflect the projected future ocean state.

Comparing baseline vs scenario

The simplest workflow is two runs (baseline and future) side by side:

import pandas as pd

base = pd.read_csv("Data_Results/Jamaica/Jamaica_2020-2023_100.0_MW_low_cost/"
                    "OTEC_sites_Jamaica_2020-2023_100.0_MW_low_cost.csv", sep=';')
fut  = pd.read_csv("Data_Results/Jamaica/Jamaica_2020-2023_ssp245_2050_100.0_MW_low_cost/"
                    "OTEC_sites_Jamaica_2020-2023_ssp245_2050_100.0_MW_low_cost.csv", sep=';')

merged = base.merge(fut[['longitude', 'latitude', 'LCOE', 'AEP']],
                    on=['longitude', 'latitude'], suffixes=('_base', '_2050'))
merged['ΔLCOE_pct'] = 100 * (merged['LCOE_2050'] - merged['LCOE_base']) / merged['LCOE_base']
merged['ΔAEP_pct']  = 100 * (merged['AEP_2050']  - merged['AEP_base'])  / merged['AEP_base']
print(merged[['ΔLCOE_pct', 'ΔAEP_pct']].describe())

For the default Jamaica example, expect ΔLCOE ≈ −5 % and ΔAEP ≈ +1.8 % under SSP2-4.5 / 2050.

Caching

  • CMIP6 monthly thetao subsets land in ~/.otex/cache/cmip6/ as Parquet files. Each (model, scenario, period, depth, bbox) request becomes one ~1-5 MB file.
  • The first call per region/scenario downloads ~10-50 MB (3 models × baseline + future window × 2 depths) and takes 5-15 minutes.
  • Subsequent calls — including the same scenario at different target years that fall within an already-fetched future window — are sub-millisecond.

Reproducibility

OTEX pins specific CMIP6 dataset versions in otex/data/climate.py::_DEFAULT_CATALOG (e.g. CanESM5 v20190429, EC-Earth3 v20200918, MPI-ESM1-2-LR v20190710). To use newer versions or substitute models, override the catalog via the OTEX_CMIP6_CATALOG environment variable.

Caveats

  • Delta-method assumes stationarity of GCM biases. This is the standard assumption in regional climate impact studies (IPCC AR6 WGII) and is generally valid for ocean temperature deltas at the regional scale used here.
  • Models cover 2015-2100. Target years before 2015 or after 2100 raise an error from the Pangeo Zarr selection.
  • The future window is 30 years centred on the target year. A request for target_year=2050 averages 2036-2065 in each model. Override via ClimateConfig.future_window_years.