Hydro Availability Scenarios

Evidence-based methodology validated with 24 years of historical data

The future will not be average. Analysis of 212 countries shows that planning for hydro variability is not optional—it’s essential for energy security.

Executive Summary

The Hydro Availability Scenarios module addresses critical uncertainty in energy planning by generating 100+ plausible future scenarios based on comprehensive historical analysis of global hydroelectric systems. This methodology transforms hydro uncertainty from an unknown risk into a manageable planning parameter.

Key Findings from Global Analysis:

212 countries analyzed with historical data (2000-2023)
74 countries with detailed monthly patterns
Major droughts successfully reproduced in scenario analysis
P10/P50/P90 framework provides robust planning boundaries

The Problem We Solve

Why Historical Averages Fail

Most energy models use simple historical averages for hydro availability, assuming every future year will match the past 5-10 year average. This approach is dangerously flawed because:

Droughts cluster: Multiple dry years often occur consecutively (e.g., Brazil 2014-2015, California 2012-2016)
Climate change intensifies extremes: Both droughts and floods are becoming more severe
Timing matters: A drought during peak demand has different impacts than one in low season
Regional correlations exist: Neighboring countries often experience similar conditions

Using historical averages leads to:

Underestimating backup needs by 10-20%
Inadequate reserve margins for drought years
Overreliance on hydro during critical periods
Systemic risk from correlated regional droughts

Our Solution: Stochastic Scenarios

Instead of one average future, we generate 100+ possible futures that capture:

Natural variability: Wet and dry year cycles
Persistence: Multi-year droughts and wet periods
Climate trends: Declining availability and increasing volatility
Extreme events: Low-probability, high-impact droughts

Understanding Scenario Outputs

The Three Key Scenarios

From 100 generated scenarios, three are particularly useful for planning:

P10 - Security Planning Scenario

What it represents: A dry future that occurs 10% of the time
Characteristics: More frequent and severe droughts
Use for: Sizing reserve margins, stress testing, ensuring system security
Key insight: If your system survives P10, it’s robust

P50 - Base Planning Scenario

What it represents: The median expected future
Characteristics: Typical mix of wet, normal, and dry years
Use for: Expected costs, base case capacity planning
Key insight: Most likely outcome for financial planning

P90 - Opportunity Scenario

What it represents: A favorable future with good hydro conditions
Characteristics: Fewer, milder droughts
Use for: Identifying export opportunities, minimum thermal needs
Key insight: Best case for hydro generation

Real-World Validation: Top Hydro Countries

Analysis of scenario outputs reveals critical planning insights validated against historical data:

**Hydro Risk Analysis: Historical vs Future Scenarios**
Country	Hydro Share	P10 Avg	P50 Avg	P90 Avg	Risk Level
Norway	96%	43.8%	46.6%	49.0%	Moderate
Brazil	65%	40.3%	43.4%	46.4%	High
Venezuela	68%	36.6%	38.7%	41.0%	Moderate
Colombia	70%	48.5%	51.0%	54.0%	Moderate
Canada	60%	50.4%	54.0%	57.2%	High
Switzerland	60%	27.6%	29.0%	30.9%	Low

Source: Analysis of 100 scenarios per country (2025-2050) validated against historical data (2000-2023)

Historical Foundation

Methodology Validation with Real Data

Our scenario generation methodology has been validated against 24 years of historical data from EMBER Climate database, covering 212 countries with annual records and 74 countries with detailed monthly patterns.

Historical Drought Documentation:

The analysis uses country-specific drought thresholds based on each nation’s historical patterns (P20 percentile) rather than arbitrary fixed values. This ensures drought identification reflects actual operational experience:

Brazil 2001: Energy crisis with 20% rationing (CF dropped to ~30%, below Brazil’s P20 threshold of 38%)
Brazil 2014-2015: Consecutive drought years, reservoir levels at 40%
Brazil 2021: Severe drought, 91-year low in some regions
European 2003: Heat wave with significant hydro impacts

Methodology Note: Drought thresholds are derived from the bottom 20% of each country’s historical capacity factors (2000-2023), ensuring definitions reflect actual operational stress rather than arbitrary percentages.

Data Foundation

Primary Data Sources:

EMBER Climate: Monthly generation database (2000-2023) covering 212 countries
Capacity Data: Installed hydro capacity by country and year
Quality Control: Systematic validation and gap-filling procedures

Coverage Statistics:

High-Quality Analysis: 74 countries with complete monthly data
Moderate Coverage: ~90 countries with regional pattern-based scenarios
Basic Coverage: ~48 countries with climate zone patterns

Derived Metrics:

For each country and month, we calculate:

Availability Factor = Monthly Generation (TWh) / (Installed Capacity (GW) × Hours in Month × 0.001)

Scenario Generation Framework

Step 1: Historical Analysis We analyze each country’s historical hydro patterns (2000-2023) to extract:

Seasonal patterns (wet/dry seasons)
Year-to-year variability
Extreme event frequency
Country-specific drought thresholds (P20 percentile)

Step 2: Regime Classification and Transition Modeling Years are classified into three regimes with realistic persistence and transition probabilities:

Dry regimes (25% probability): 15% reduction in availability, tend to persist for 2-3 years
Normal regimes (50% probability): Baseline conditions with ±10% variation
Wet regimes (25% probability): 10% increase in availability, favorable conditions

Regime Transition Matrix: Each year has a 30% probability of regime change, with equal probability of transitioning to any of the three regimes. This creates realistic multi-year drought and wet periods while preventing unrealistic regime oscillations.

Step 3: Climate Change Integration Future scenarios incorporate climate change through four distinct pathways with quantified parameters:

Climate Scenario Parameters:

Scenario	Trend (%/decade)	Variability Increase (%/decade)	Description	Use Case
Optimistic	-0.3%	+5%	Mild climate impacts	Best case planning
Moderate	-0.5%	+10%	IPCC median projections	Base case planning
Pessimistic	-1.0%	+15%	Severe climate impacts	Risk assessment
Severe	-1.5%	+20%	Extreme climate change	Stress testing

Mathematical Implementation: - Trend Factor: trend_factor = 1 + (climate_trend × decade_offset) - Variability Factor: variability_factor = 1 + (variability_increase × decade_offset) - Decade Offset: (year - 2025) / 10 for linear progression from 2025 baseline

Step 4: Stochastic Scenario Generation Algorithm

For each of 100 scenarios, we implement a Monte Carlo simulation with regime-switching dynamics to generate 25+ years of monthly availability factors:

Algorithm Overview:

# For each scenario (1 to 100):
for scenario in range(1, 101):
    # 1. Initialize regime (25% dry, 50% normal, 25% wet)
    regime = random_regime_selection()

    # 2. Generate each year
    for year in range(start_year, end_year + 1):
        # Climate trend adjustment
        decade_offset = (year - 2025) / 10
        trend_factor = 1 + (climate_trend * decade_offset)
        variability_factor = 1 + (variability_increase * decade_offset)

        # 3. Annual scaling based on regime
        if regime == 'dry':
            annual_scale = 0.85 + random(0, 0.1)  # 15% reduction ± 10%
        elif regime == 'wet':
            annual_scale = 1.1 + random(0, 0.1)   # 10% increase ± 10%
        else:  # normal
            annual_scale = 0.95 + random(0, 0.1)  # Slight variation ± 10%

        annual_scale *= trend_factor

        # 4. Generate monthly values
        for month in range(1, 13):
            base_value = mean_annual * seasonal_pattern[month]
            random_variation = gaussian_random() * std_annual * variability_factor
            availability_factor = (base_value * annual_scale) + random_variation

            # Bound between 5% and 95%
            availability_factor = max(0.05, min(0.95, availability_factor))

        # 5. Regime transition (30% probability each year)
        if random() < 0.3:
            regime = new_random_regime()

Key Technical Details:

Regime Persistence: Regimes tend to persist for 2-3 years (70% probability of staying in same regime)
Climate Adjustments: Four scenarios (optimistic: -0.3%/decade, moderate: -0.5%/decade, pessimistic: -1.0%/decade, severe: -1.5%/decade)
Variability Increase: 5-20% per decade depending on climate scenario
Seasonal Preservation: Monthly patterns scaled from historical seasonal coefficients
Stochastic Component: Gaussian random variations with country-specific standard deviations
Physical Bounds: All values constrained between 5% and 95% availability factors

Statistical Properties of Generated Scenarios

Scenario Distribution and Validation:

The 100 generated scenarios create a statistically robust distribution that enables percentile-based analysis:

P10 Scenario: 10th percentile (dry conditions, occurs ~10% of time)
P50 Scenario: 50th percentile (median conditions, most likely outcome)
P90 Scenario: 90th percentile (wet conditions, favorable hydro)

Quality Assurance Metrics:

# Statistical validation for each country:
- Mean preservation: Generated mean ≈ Historical mean × climate_trend
- Seasonal pattern correlation: R² > 0.85 with historical monthly patterns
- Variability bounds: 5th-95th percentile range matches historical ± climate adjustment
- Regime persistence: Average regime duration 2.3-2.7 years (realistic)
- Extreme event frequency: P5/P95 scenarios occur with appropriate rarity

Monte Carlo Convergence: 100 scenarios provide sufficient statistical power for: - Stable percentile estimates (±2% confidence intervals) - Robust extreme event representation - Reliable risk assessment across planning horizons

Seasonal Pattern Analysis

The methodology successfully preserves historical seasonal patterns while projecting future variability:

Seasonal hydro availability profiles — **Seasonal Hydro Availability: Future Scenarios vs Historical Data**

P10/P50/P90 future scenarios (solid lines) compared with historical monthly averages and variability ranges (orange). The analysis shows scenarios successfully preserve seasonal patterns while capturing climate-driven changes.

Annual Trajectory Validation

Long-term scenario trajectories demonstrate realistic evolution from historical baselines:

Annual hydro availability trajectories — **Annual Hydro Availability: Future Scenarios vs Historical Data (2000-2050)**

Historical time series (2000-2023) seamlessly connect to future P10/P50/P90 scenarios (2025-2050). Country-specific drought thresholds (red dashed lines) reflect actual operational experience rather than arbitrary values.

Regional Patterns

Tropical Systems (Brazil, Colombia, Venezuela): - High seasonal variability with distinct wet/dry seasons - Strong El Niño/La Niña impacts on annual patterns - Multi-year drought sequences require extended backup planning

Snow-Dominated (Norway, Canada, Switzerland): - Spring snowmelt timing critical for annual generation - Climate change shifting peak timing earlier - Storage capacity crucial for seasonal balancing

Monsoon-Dependent (Vietnam, parts of Asia): - Extreme seasonal concentration of generation - Monsoon timing uncertainty creates planning challenges - Flash drought risk during failed monsoon years

Historical Validation

Our scenarios successfully reproduce the statistics of major historical droughts:

Historical drought validation analysis — **Historical Drought Analysis: Validation of Scenario Methodology**

Left: Drought frequency vs hydro dependency using country-specific P20 thresholds. Right: Worst-case historical availability factors showing actual operational stress levels experienced by each country.

Practical Application Guide

The Three-Scenario Planning Framework

Integrated Planning Approach: Use all three key scenarios together for robust decision-making rather than selecting just one:

1. P50 (Base Case Planning) - Capacity Planning: Size base generation for median conditions - Financial Analysis: Expected returns and base case NPV - Operational Planning: Normal maintenance scheduling and operations - Market Analysis: Typical export/import patterns

2. P10 (Security & Risk Planning) - Reserve Sizing: Additional firm capacity needed for drought resilience - Financial Stress Testing: Downside risk and debt service coverage - Emergency Preparedness: Backup resource requirements - System Adequacy: Minimum reliability standards

3. P90 (Opportunity Planning) - Export Potential: Maximum surplus generation opportunities - Investment Optimization: Upside potential and market opportunities - Minimum Thermal: Reduced backup needs in favorable conditions

Practical Example: Brazil Capacity Planning

Traditional Approach (Historical Average):
Required Capacity = Peak Demand / 0.41 (average availability) = 100 GW

Scenario-Based Approach:
P50 Base Planning = Peak Demand / 0.41 = 100 GW base capacity
P10 Security Check = Peak Demand / 0.35 = 114 GW total needed
Required Reserve = 14 GW additional firm capacity

Decision: Build 100 GW base + 14 GW reserve = 114 GW total
Result: 14% higher investment but drought-resilient system

Climate Change Implications

Our analysis reveals concerning trends for 2025-2050:

Global Average Changes:

Mean availability: -5% by 2050 (moderate climate scenario)
Variability: +20% increase in standard deviation
Extreme events: 2x frequency of severe droughts

Regional Hotspots:

Mediterranean: Summer availability could drop 30%
Tropical: Wet/dry season intensification
Snow-based: 1-2 month earlier peak, reduced total

Planning Implications:

Increased backup needs: 10-20% more firm capacity required
Storage value: 2-3x more valuable for drought bridging
Regional transmission: Critical for resource sharing

Energy Model Integration Methods

Choose the integration method that matches your model complexity and computational resources:

Method 1: Full Stochastic Optimization (Recommended) - Input: All 100 scenarios with equal probability weights - Advantage: Captures full uncertainty and year-to-year dynamics - Use for: Capacity expansion models, storage valuation, investment optimization - Implementation: scenarios_1_to_100.csv as stochastic input parameters

Method 2: Three-Scenario Sensitivity Analysis - Input: P10/P50/P90 monthly values as separate model runs - Advantage: Simpler implementation, clear risk boundaries - Use for: Policy analysis, quick assessments, deterministic models - Implementation: Three separate model runs with different hydro availability assumptions

Method 3: Direct Scenario Selection - Input: Specific scenario number (e.g., Scenario #23 for P10 conditions) - Advantage: Preserves realistic year-to-year transitions and storage dynamics - Use for: Detailed operational studies, multi-year storage analysis - Implementation: Single scenario time series maintaining temporal correlations

Model Integration Best Practices: - Never use historical averages alone - Minimum use P50 with P10 verification - Size reserves for P10 scenarios - Better over-prepared than experiencing rationing - Value flexibility highly - Storage and demand response worth premium in variable futures - Consider regional correlations - Neighboring countries often experience similar drought conditions

Data Quality and Coverage

Quality Transparency:

Quality Level	Basis	Countries	Example
High	Direct monthly data	74	Brazil, Norway, Canada
Moderate	Regional patterns	~90	Bolivia (from Brazil)
Basic	Generic patterns	~48	Small island states

Limitations and Uncertainties

What We Capture

Natural climate variability and persistence
Climate change trends and intensification
Realistic drought sequences and recovery patterns
Seasonal pattern preservation

What We Don’t Model

Reservoir operations and storage constraints
Upstream/downstream water dependencies
Political water allocation changes
Catastrophic infrastructure failures

Uncertainty Bounds

P10-P90 range captures 80% confidence interval
Extreme scenarios (P5, P95) available for stress testing
Climate scenarios span optimistic to severe projections

Summary and Key Takeaways

The Core Problem Solved: Hydro availability uncertainty represents a critical risk for energy systems worldwide. Traditional planning using historical averages systematically underestimates backup needs by 10-20%, leading to inadequate reserve margins and potential energy security risks.

Our Solution: A data-driven stochastic scenario approach, validated against 24 years of historical data from 212 countries, that transforms hydro uncertainty from an unknown risk into manageable planning parameters through 100 Monte Carlo scenarios per country.

Essential Planning Principles:

The future will not be average - Historical averages mask critical variability
Use the three-scenario framework - P10 for security, P50 for base planning, P90 for opportunities
Size reserves for drought resilience - P10 scenarios reveal true backup capacity needs
Value flexibility and storage highly - 2-3x more valuable in variable hydro futures
Plan for climate change impacts - Declining availability and increasing variability by 2050

Bottom Line: Planning for hydro variability using stochastic scenarios is not optional—it’s essential for energy security and cost-effective system design in an uncertain climate future.

—

Methodology validated against historical data from EMBER Climate (2000-2023) Scenario generation covers 212 countries with 100 futures per country (2025-2050)