Learn about the MapBiomas Alerta method

MapBiomas Alerta Perú is a collaborative initiative that transforms early deforestation alerts into verified, accurate, and contextualized data. Processing this set of alerts involves rigorous validation, the refinement of cartographic boundaries using high-resolution satellite imagery (PlanetScope with 3.7 m spatial resolution), the generation of detailed reports, and the publication of final results on a single open-access platform ( https://plataforma.peru.alerta.mapbiomas.org/   ).

DESCRIPTION OF THE STEPS

The MapBiomas Alerta Perú process encompasses the following steps: compilation, validation, refinement, cross-referencing with public data, auditing, and publication of alerts and deforestation reports (Figure 1).

Figure 1. Methodological process of MapBiomas Alerta Perú to compile, validate, refine, cross-reference data, audit, and publish deforestation alerts in Brazil

Step 1: Alerts produced for the Amazon biome are collected, and information from early deforestation alerts originating from the official monitoring system in Peru is organized and consolidated. This system uses Landsat satellite imagery with 30-meter resolution to detect and generate deforestation alerts.

Step 2: Validation and selection of before and after images

The validation process is carried out in two stages. The first is automatic, eliminating all deforestation alerts that are less than 1 hectare. The second stage consists of a visual inspection by trained analysts, organized into teams, supported by monthly high-resolution image mosaics from the Planet constellation (with 3.7-meter resolution) and Sentinel 2. In this phase, alerts that turn out to be false positives can also be discarded, recording the reason for rejection. Common reasons for discarding include.

• Agriculture: Alerts generated by pre-existing agricultural activity.
• Degradation: Natural losses, forestry activity (temporary roads, logging), or degradation events.
• Previously altered: Areas that had already been previously impacted (land use change in Non-Forest).
• Relief shadows: Detections caused by orographic shadows.
• Season change: Seasonal variations (drought/humidity) or changes in river water levels that generate false positives.
• Fires/burns: Forest fires that do not necessarily cause a land use change.
• Image unavailability: Absence of clear images due to cloud cover or other interference.
• Reforestation: Alerts caused by the logging of forest plantations.
• Duplicate: Cartographically and temporally overlapping alerts.

Only when visual inspection confirms the anthropic deforestation event is the alert considered pre-approved. From that moment, pairs of high-resolution satellite images (one showing native vegetation before and another after deforestation) are selected to move on to the next stage.

Step 3: Validation and refinement on high-resolution images

Once alerts have been pre-approved and high-resolution images selected, the spatial boundaries are refined. This step is crucial to more precisely delineate the area effectively deforested.

Refinement can be done in two ways:

• Automatic: Using advanced algorithms such as Random Forest, the system classifies the change area based on training samples from the high-resolution images (deforested and non-deforested zones). Then, the final results are converted into a refined polygon, which goes through a simplification process to remove excessive vertices.
• Manual: In complex cases, analysts can manually draw the polygon of the deforested area, using the "Accumulated Non-Forest" layer as an unconditional reference.

During this process, the analyst also identifies and records the causes of deforestation, such as agriculture, mining, livestock, road infrastructure, urban expansion, among others.

Figure 2. Example of Planet images before and after deforestation, along with the refined alert polygon

Step 4: Auditing
Each refined polygon goes through an auditing process performed by the technical supervisor. At this stage, the possible need to make additional adjustments before the final publication of the confirmed deforestation is evaluated.

Step 5: Cross-referencing with public territorial databases
The refined polygons are spatially overlaid with information from territorial categories, including native communities, indigenous reserves, peasant communities, natural protected areas, buffer zones, regional conservation areas, private conservation areas, ecozones, fragile ecosystems, among others. The alerts are also linked to administrative boundaries at the district, provincial, and departmental levels. These cross-references qualify the alerts and allow for the generation of technical reports based on relevant information for user institutions.

Step 6: Publication

All confirmed deforestation polygons are published on the MapBiomas Alerta Perú Platform, with weekly updates. Reports are available for each confirmed deforestation (with an area greater than 1 ha) and for each alert cross-referenced with a territorial category. The reports contain the following information:

• Deforestation alert code;
• Original source of the alert (detection system);
• Biome, department, province, and district;
• Deforestation area;
• Image and date before deforestation;
• Image and date after deforestation;
• Overlay of deforestation with: native communities, indigenous reserves, peasant communities, natural protected areas, buffer zones, regional conservation areas, private conservation areas, ecozones, fragile ecosystems, others;
•  MapBiomas collection 3 land cover and land use history;
•  Landsat image history for the evaluated area;
• Data sources used in spatial cross-referencing. 

Cancellation and Correction of Post-Publication Alerts

Under certain circumstances, alerts published on the MapBiomas Alerta Perú Platform can be corrected or even canceled. Whenever there is a formal indication or a justified request pointing out possible errors associated with the alerts, whether from environmental agencies or platform users, the technical team conducts an exhaustive analysis of said alerts.

This analysis is carried out by reviewing the Planet images and, when necessary, various complementary sources, such as images from other satellites (Sentinel, Landsat, etc.), high-resolution satellite imagery available on Google Earth, and the MapBiomas Land Cover and Land Use Cartography. In cases where it is confirmed that the published alert does not correspond to an actual deforestation or forest conversion event (regardless of timber extraction, legality, or liability), the alert is canceled. This implies that it is removed from the map and platform statistics, but remains in the database for individual consultation using its unique identifier code.

METHOD LIMITATIONS
Like any method, MapBiomas Alerta has some limitations that must be considered when applying its data:

1. Processing time: The importation of alerts (detection systems) is carried out monthly. Part of the validation and processing is done individually and visually by trained analysts, so the validation and processing time can vary depending on the biome and the time of year. This period can range from 30 to 60 days from detection by the source system until publication on the MapBiomas Alerta platform. The objective of MapBiomas Alerta Perú is to increase certainty regarding confirmed deforestation data and offer reports ready for remote surveillance. Rapid field surveillance operations, aimed at detecting flagrant deforestation, can be planned directly with existing detection systems.
2. Alert omissions: Deforestation is validated and refined based on the previous existence of an alert captured by an external detection system. Possible omissions by these systems in detecting deforestation also affect the alerts evaluated by MapBiomas Alerta Perú.
3. Underestimation of deforestation speed: When validating and refining an alert, a search is made for a pair of good quality Planet satellite images, from before and after the deforestation. The "before" image is the most recent within a period of up to 6 months prior to detection, and the "after" image is the closest to the end of the deforestation process. The presence of clouds can increase the period between both images for days, weeks, and even months. This does not alter the assertion that the deforestation event occurred in the period between the two images, but it can affect the calculation of the average speed at which the deforestation actually occurred.