Nimpact Academy - Module 4: Algae Prediction & Bloom Risk

The Algae Problem: Why It Matters

Harmful algal blooms (HABs) are one of the most serious water quality threats worldwide, causing beach closures, fish kills, and human health risks. Nimpact uses water-body-specific machine learning models trained on thousands of beaches to predict algae concentrations from satellite-derived environmental parameters.

Two Key Algae Metrics

Cyanobacteria Index (CI)

Measures concentration of blue-green algae (cyanobacteria)—the most harmful type. Some species produce toxins (microcystins, anatoxins) that can cause:

Liver damage in humans and animals
Neurological effects
Skin irritation and respiratory problems
Pet deaths from drinking contaminated water

Units: Cells/mL (typical range 0-100,000)

Chlorophyll-a Concentration

Measures all algae types (green, brown, blue-green) via their primary photosynthetic pigment. High chlorophyll indicates:

High biological productivity (eutrophication)
Nutrient enrichment (phosphorus, nitrogen)
Potential for oxygen depletion and fish kills
Reduced water clarity

Units: μg/L (typical range 1-100 μg/L)

Why Prediction Is Challenging

Unlike temperature and clarity (which satellites measure directly), algae concentrations cannot be reliably measured from space for inland waters. Here's why:

Patchy Distribution: Blooms concentrate in mats and streaks, not evenly distributed
Depth Variation: Some algae float (surface scums), others suspend at various depths
Spectral Similarity: Algae reflectance overlaps with suspended sediments and CDOM
Atmospheric Interference: Haze and aerosols corrupt the weak algae signal

    Nimpact's Solution: Rather than trying to measure algae directly, we predict concentrations using environmental drivers (temperature, clarity, nutrient proxies) that CAN be measured accurately. This "indirect measurement" approach achieves better accuracy than direct spectral analysis.

Water-Body-Specific Models

Nimpact uses three separate machine learning models because algae dynamics differ fundamentally across water body types:

# Algae Prediction Models (Random Forest) RIVER Model (n=1,247 training sites): - Cyanobacteria: R² = 0.69 (geography-based) - Chlorophyll-a: R² = 0.49 (temperature, clarity) - Rivers show elevated levels due to upstream nutrient loading LAKE Model (n=2,183 training sites): - Cyanobacteria: Not predictable (nutrient-limited) - Chlorophyll-a: R² = 0.55 (temperature, clarity) - Lakes show summer bloom patterns when warm + nutrients TIDAL Model (n=891 training sites): - Cyanobacteria: Not predictable (salinity inhibits) - Chlorophyll-a: R² = 0.61 (temperature, clarity, tidal mixing) - Coastal blooms driven by upwelling and tidal cycles

This water-type stratification is critical. A river with high cyanobacteria is normal (upstream sources), while a lake with the same concentration indicates a serious local problem.

Page 1 of 2

Summer Bloom Analysis and Risk Assessment

The Summer Bloom Window

For lakes, Nimpact performs specialized summer bloom risk analysis because most severe blooms occur during the June-September warm season. This analysis is particularly important for northern prairie and temperate lakes.

The Northern Prairie Paradox

Northern lakes like Eagle Lake (SK) experience severe annual blooms at surprisingly low temperatures:

Peak Temperature: Only 15.5°C (would be considered "cool" in southern lakes)
Result: Annual beach closures, toxic algae warnings, economic impacts

Why? Cold-adapted cyanobacteria species, high agricultural nutrient loading, shallow depths allowing sediment nutrient release, and long summer daylight hours (16+ hours) for photosynthesis.

Summer Risk Factors

Nimpact evaluates multiple risk factors to predict summer bloom probability:

Temperature Risk (Most Critical)
- Max temp ≥ 15°C: HIGH RISK for northern lakes
- Max temp ≥ 20°C: HIGH RISK for all lakes
- Max temp ≥ 25°C: EXTREME RISK—expect severe blooms if nutrients available
Thermal Amplitude (Seasonality)
- Range ≥ 8°C: Seasonal lake—predictable bloom window
- Range < 8°C: Thermally stable—blooms less predictable
Shallow Water (Nutrient Mixing)
- Secchi < 2m: Shallow enough for wind to mix nutrients from sediments
- Turbidity > 2 NTU: Potential organic matter/existing algae
Geographic Location
- Latitude 45-60°N: Northern temperate zone—prime bloom geography
- Prairie regions: Agricultural nutrient loading + shallow lakes

Risk Level Classification

# Bloom Risk Tiers
LOW:     Cool water (< 12°C peak) OR oceanic (high salinity)
MEDIUM:  Moderate temps (12-20°C) with some risk factors
HIGH:    Warm temps (> 20°C) + seasonal pattern + shallow
EXTREME: Very warm (> 25°C) + all risk factors present

Measured vs. Predicted Data

Nimpact uses both measured satellite data (when available) and machine learning predictions:

Data Source Hierarchy:

Direct Satellite Measurement: Chlorophyll-a from Sentinel-2 spectral analysis (when conditions allow)
ML Prediction: When direct measurement impossible or unreliable (most cases for inland waters)
Unavailable: Cyanobacteria for tidal/lake (requires nutrient data not available from satellites)

For rivers, cyanobacteria predictions achieve good accuracy (R² = 0.69) because they're primarily driven by watershed geography and upstream sources—factors we can model. For lakes and tidal areas, cyanobacteria blooms are sporadic and nutrient-dependent, making prediction from satellite data alone impossible.

Health Guidelines and Action Thresholds

# WHO / EPA Recreational Water Guidelines Chlorophyll-a: < 10 μg/L: Low risk, excellent conditions 10-25 μg/L: Moderate, monitor for blooms 25-75 μg/L: Elevated, consider warnings > 75 μg/L: High risk, limit recreation Cyanobacteria: < 20,000 cells/mL: Low risk (vigilance mode) 20,000-100,000: Moderate (alert level) > 100,000: High risk (closure consideration) > 1,000,000: Dangerous (immediate closure)

    Nimpact Role: Pre-screening tool to identify beaches that need water quality sampling and laboratory analysis. High predicted risk triggers professional testing, not closure—satellites screen, labs confirm.

Page 2 of 2 - Ready for Quiz

Module 4: Algae Prediction & Bloom Risk