An Image Tells You What Exists. A Time Series Tells You What’s Changing

In 2019, when MapBiomas Alerta began operating, only 5% of deforested area in Brazil had any kind of associated enforcement action. By 2025, that share had reached 65%, a change that didn’t come from more inspectors in the field, but from a continuously processed image time series, cross-checked against public registries and delivered as evidence traceable by date and location. What MapBiomas demonstrated at the scale of the national territory, industrial operations need to replicate at the scale of the portfolio, because the difference between a one-off image and a time series isn’t one of degree — it’s a difference in the class of question the data can answer.

What a single image can’t answer

Orbital imagery answers state questions well: is there vegetation in this area, is there a structure in this polygon, is the water body turbid. For that kind of question, one good-resolution image is enough. The problem starts when the question shifts from state to trajectory. A tailings dam with an incipient crack in the slope looks no different from a stable dam in a single image, and it’s only by comparing it to the image from thirty days earlier that millimeter-scale deformation reveals itself as progressive displacement. Likewise, a PRAD area only shows verifiable environmental recovery when the historical series shows the vegetation index growing consistently over months, and incipient encroachment on a right-of-way only appears at its early stages when the system compares the current image against the reference state logged at the start of monitoring. In every one of these cases, a single image is an input. The time series is the product that drives the decision.

What multitemporal analysis delivers that a single image doesn’t

Multitemporal analysis compares images from distinct dates pixel by pixel, identifies variation above expected noise, and classifies the type of change, whether vegetation expansion, cover removal, built-surface alteration, or ground displacement. That opens up three capabilities a single image simply doesn’t have. The first is early detection calibrated by history: a system that knows a specific area’s seasonal pattern distinguishes normal growth from anomalous growth, so on a right-of-way in the Cerrado where vegetation grows faster after the rainy season, a system with no history would treat that as an event worth investigating, while a system with a two-year series already knows it’s expected behavior, and the alert only fires when growth breaks from the seasonal window, the historical pace, or the direction toward the conductor. The second is traceability with an origin date, since the time series determines exactly when an irregularity began and how fast it progressed, letting the company, in response to an ANM, ANEEL, or IBAMA notification, reconstruct the event’s full timeline with independent evidence at every stage, instead of just showing the asset’s current state. The third is separating events from trends, because a fire hotspot is a discrete event while vegetation progressively advancing toward a critical asset over months is a trend, and each requires a different response — immediate reaction in one case, preventive planning in the other — a distinction a system that only analyzes the current state can’t make.

What a long history adds that recent monitoring doesn’t capture

Continuous monitoring starting today delivers a growing time series, but starts from zero on the activation date, leaving out the context that predates the system for an asset that has existed for years. Access to historical archive imagery, which for some constellations spans a decade, makes it possible to reconstruct an asset’s state on any date within the covered period.

In a regulatory context, that reconstruction can show that an irregularity detected today didn’t exist on an earlier date, or that a reported compliance condition matches the state verifiable by independent imagery, which changes the company’s position in a TAC, licensing, or environmental audit.

In due diligence, the same history makes it possible to assess an area’s land use before a transaction. Encroachment removed, unauthorized expansion later regularized: all of it shows up in the historical series, regardless of whether it appears in the asset’s documents.

In risk management, a long history reveals patterns recent monitoring doesn’t capture, such as an area that gets encroached on every year during the same period. That pattern informs when to intensify surveillance and where to focus field resources.

Why a time series requires a different architecture than a single image

Accumulating images isn’t the same as operating a time series. For pixel-by-pixel comparison to be valid, every image needs to be registered with consistent geometric precision, since variation in capture angle, atmospheric correction, or radiometric calibration between dates introduces artifacts the system needs to distinguish from real ground change. Across a large portfolio, the accumulated volume stops being manageable by hand, which requires automated processing with a model calibrated by asset type and region, and alerts that come out already filtered by each polygon’s history, not as a generic detection applied to an entire scene.

That’s why the distinction between an image-access platform and a spatial-intelligence platform remains relevant when it comes to time series: access to historical data is a necessary condition, but the processing that turns that data into intelligence traceable by asset, by date, and by event type is what decides whether the time series becomes a decision or just an archive.

How NOR runs time series across an industrial portfolio

At NOR Space Intelligence, every asset in a client’s portfolio has a logged reference state, a historical series integrated since activation, and expected-variation parameters defined by asset class and regional seasonality. Every alert generated carries not just the event’s current state, but the historical context that makes it relevant: when the asset was last in compliance, how fast the change progressed, and whether the pattern matches a discrete event or a developing trend.

For operations that need to demonstrate continuous compliance to ANM, ANEEL, or IBAMA, this evidence chain with a verifiable timeline turns monitoring from an operational tool into a regulatory asset.

To understand how this time-series architecture applies to your portfolio, talk to a NOR specialist.