Forty years of free, orbital, multi-spectral imagery already exists for your lake. Used correctly, it answers questions that a single year of in-situ sampling cannot: when did the chronic bloom regime start, where does it begin each summer, and how does it move. Used incorrectly — or sold as a substitute for sampling — it produces confident wrong answers.
This Insight is the framework we use to extract a defensible decadal record from Landsat and Sentinel, where the data are trustworthy, where they are not, and how to pair them with in-situ work so the satellite work earns its keep.
The forty-year archive nobody is using
Landsat has been imaging your lake every 16 days since 1984. Sentinel-2 every 5 days since 2017. Both are public-domain, radiometrically calibrated, and atmospherically correctable to surface reflectance. For most lakes larger than about 4 hectares, this archive contains the only decade-scale, spatially explicit record of water quality that exists — and 95% of lake-management programs have never opened it.
What can you actually pull out of it? Three things, reliably:
- Bloom phenology. When does the bloom typically start, where in the lake does it begin, and how has the timing shifted over the past two decades. Chlorophyll-a indices (NDCI, Maximum Chlorophyll Index, MCI on Sentinel-2 and OLCI on Sentinel-3) recover annual phenology with useful precision even in moderately turbid water.
- Spatial heterogeneity. Where the chlorophyll anomaly is consistently highest across years. This is almost always your internal-loading hotspot, and it almost always sits over a specific sediment regime — typically the deepest anoxic basin or a depositional zone downwind of the prevailing summer wind.
- Regime shifts. The year the lake crossed from oligotrophic to mesotrophic, or from periodic to chronic cyanobacterial dominance. Regime shifts have signatures that sequential annual imagery captures and single-station time-series sampling often misses.
What you cannot ask the satellite
The satellite cannot tell you which species is blooming, what the toxin profile is, whether the bloom is buoyant or mixed, or anything about the hypolimnion. It cannot resolve sub-pixel structure: in Sentinel-2's 10–20 m bands, a 30-meter shoreline cyanobacterial scum is one or two pixels — useful for detection, useless for quantification. And outside the visible algae signal, the chlorophyll-a retrievals saturate, fail in optically complex waters with high CDOM, and depend on calibration assumptions that break down for the lakes that need this work most.
Vendors selling "satellite-based lake monitoring" as a substitute for in-situ work either know this and are betting you don't, or don't know it themselves. Neither is the partner you want. Used as a screening and prioritization layer — not a measurement layer — the imagery is genuinely transformative.
How we build the decadal record
For a typical project, we pull the full Landsat 5/7/8/9 and Sentinel-2 archives for the lake's footprint, atmospherically correct each scene (Sen2Cor, LaSRC, or ACOLITE depending on sensor and conditions), mask clouds and cloud shadows, and compute three index families per scene: chlorophyll-a (band-ratio or two-band NIR-red indices appropriate to the sensor), turbidity or non-algal particle load, and surface temperature where Landsat's thermal bands resolve it.
The output is a per-pixel monthly time series from 1984 onward. From that we extract: annual bloom-onset Julian day, peak-bloom Julian day, growing-season chlorophyll integral, year-on-year spatial centroid of the chlorophyll anomaly, and an annual regime-detection statistic. The deliverable is a 4–8 page technical memorandum, the time-series data in CSV, a set of annual chlorophyll maps in GeoTIFF, and three to five diagnostic findings that frame the in-situ sampling design.
Where it changes the project economics
The satellite work typically costs less than a single sampling event. It does not replace sampling — it tells you where and when to sample. On lake programs where the budget has been spent on five years of monthly grab samples at one mid-lake station, the decadal imagery has surfaced internal-loading hotspots two kilometers from that station that the sampling program would never have found. On reservoir programs facing treatment-plant taste-and-odor episodes, the imagery has identified the upstream tributary embayment where the cyanobacterial seed population establishes three weeks before it reaches the intake.
For utility, lake-association, and regulatory clients, the deliverable doubles as a defensible record. A decadal trend documented from independent, public satellite imagery is harder to argue with than a contested in-situ dataset collected by competing parties.
What we deliver
ENV runs this analysis as a standalone screening engagement or as Phase 0 of a larger lake-management program. Either way the output frames a diagnostic question — typically about internal loading, sediment-water interface dynamics, or watershed-source attribution — that the next phase of work answers. The imagery is the cheapest, fastest way to make sure the expensive work is pointed at the right part of the lake.
The data has been collecting itself for forty years. The decision is whether to read it.