Methodology

 MAAP has a dynamic, evolving methodology designed to continuously incorporate the latest technology. Our overall goal is to use the best available technology to detect deforestation in near real-time, determine the direct drivers of the deforestation, and understand larger-scale deforestation patterns.  The image below shows our basic methodology in graphical form. Further below, we briefly describe each step.

Current MAAP Methodology

Current MAAP Methodology

Step 1.  Analyze Deforestation Alerts

Starting in early 2016, WRI’s Global Forest Watch, in collaboration with the GLAD (Global Land Analysis & Discovery) lab at the University of Maryland, began publishing weekly GLAD deforestation alerts for Peru. While previous alert systems are based on 250-500 meter resolution MODIS imagery, GLAD alerts are based on 30 meter resolution Landsat imagery. For details on GLAD alert methodology, see: Hansen et al 2016, ERL 11:3. We analyze fresh GLAD alerts every week to locate emerging or expanding deforestation fronts in the Peruvian Amazon.

We also periodically analyze Terra-i deforestation alerts, particularly for areas outside of Peru. These alerts are based on 250 meter resolution MODIS imagery.

We also occasionally produce our own deforestation alerts, particularly for areas outside of Peru, using CLASlite software. CLASlite is an automated system that detects forest loss between any two Landsat images.

Our goal is to let these deforestation alerts guide our more targeted analysis described in the subsequent steps. We analyze these alerts in search of the most important emerging or expanding deforestation fronts in the Andean Amazon (with an initial focus on Peru). For example, we look for:

  • Deforestation alerts within protected areas or isolated indigenous communities.
  • Deforestation “hotspots” as indicated by kernel density analysis.
  • Concentration of alerts that indicate medium- or large-scale deforestation.
  • Deforestation alerts in areas known to be “hotspots”, such as the gold mining zone of Madre de Dios.
  • Linear patterns indicative of new road construction.
  • We also take requests from authorities and colleagues for targeted analysis on specific areas of interest.

Step 2. Verify Critical Deforestation Alerts

Once a potentially important or critical alert is identified, we manually verify the deforestation by directly analyzing 30 meter resolution Landsat images (note: these images are the basis of the GLAD and CLASlite alerts). We often construct a time-series of Landsat images to determine the precise time-frame of the deforestation event.  Landsat images are freely available from USGS on the GloVis web portal: http://glovis.usgs.gov/. With its recent launch, we are also increasingly using Sentinel-2 imagery to manually verify deforestation alerts. Sentinel-2 imagery is freely available from the European Space Agency.

Step 3. High Resolution Imagery to Identify Deforestation Drivers

Once we have verified an important alert, we use high-resolution imagery to identify the deforestation driver. We have three primary sources of high-resolution imagery:

  • Digital Globe via the NextView Program, courtesy of an agreement between ACA and USAID. This includes Worldview and Geoeye imagery (0.5 m resolution).
  • Planet Labs via their Ambassador Program. This includes Planet Dove imagery (3 m resolution).
  • Airbus via Apollo Mapping. This includes SPOT imagery (1.5 m resolution).

Note that this imagery is not freely available.

Analyzing this imagery, we are able to determine if the cause of deforestation is mining, agriculture, pasture, urban sprawl, or infrastructure (such as road or dam). Among agriculture, we can identify adult oil palm trees, but other crops are difficult to distinguish.

Step 4: If Persistent Clouds, Use Radar

Under Construction…