Month: June 2024

MAAP #213: Estimating Carbon in Amazon Protected Areas & Indigenous Territories

Posted on by dcadmin
Intro Image. Screenshot of OBI-WAN forest carbon reporting app.

In a recent report (MAAP #199), we presented the updated version of NASA’s GEDI data,1 which uses lasers aboard the International Space Station to provide cutting-edge estimates of aboveground carbon globally, including our focal area, the Amazon.

These lasers, however, have not yet achieved full coverage, leaving considerable gaps in the data and resulting maps.

Here, we feature two new tools that allow us to fill in these gaps and provide detailed wall-to-wall estimates of aboveground biomass for specific areas, which can then be converted to aboveground carbon estimates.

The first is the OBI-WAN forest carbon reporting app (see Intro Image), which uses statistical inference to produce mean, total, and uncertainty estimates for biomass baselines at any given scale (from local to worldwide).2

The second is a fused product from GEDI and TanDEM-X missions.3 The combination of lidar (GEDI) and radar (TanDEM-X) has started to produce unmatched maps that combine the ability of lidar to retrieve forest structure and the ability of radar to offer wall-to-wall coverage at multiple resolutions (see Figures 1-5 below for examples at 25m resolution).

Employing these two tools, we focus on estimating aboveground carbon for select examples of two critical land designations in the Amazon: protected areas and indigenous territories. Both are critical to the long-term conservation of the core Amazon (MAAP #183). We hope that providing precise carbon data will provide additional incentives for their long-term conservation.

We select 5 focal areas (3 National Parks and 2 Indigenous Territories; see list below) across the Amazon to demonstrate the power of these datasets. Together, these five areas are currently home to over 1.4 billion metric tons of aboveground carbon.

  • Protected Areas (National Parks)
    Chirbiquete National Park (Colombian Amazon)
    Manu National Park (Peruvian Amazon)
    Madidi National Park (Bolivian Amazon)
    k
  • Indigenous Territories
    Kayapó Indigenous Territory (Brazilian Amazon)
    Barranco Chico Indigenous Territory (Peruvian Amazon)

Focal Areas

As noted above, the aboveground carbon estimates below are based on the aboveground biomass estimates from the OBI-WAN forest carbon reporting app and GEDI-TanDEM-X data. Figures 1 – 5 are based on GEDI-TanDEM-X, at 25 meter resolution.

National Parks

Chirbiquete National Park (Colombian Amazon)

Chirbiquete National Park covers over 4.2 million hectares in the heart of the Colombian Amazon (Guaviare and Caqueta departments). Both datasets converge in the estimate of around 600 metric tons of aboveground biomass, equating to over 300 million metric tons of aboveground carbon across the park (80.5 tons of carbon per hectare). Figure 1 shows the detailed spatial distribution of this biomass across Chirbiquete National Park. Note that the GEDI-TanDEM-X data misses the western tip of the park.

Figure 1. Aboveground biomass across Chiribiquete National Park (Colombian Amazon). Data: GEDI-TanDEM-X

 

Manu National Park (Peruvian Amazon)

Figure 2. Aboveground biomass across Manu National Park (Peruvian Amazon). Data: GEDI-TanDEM-X

Manu National Park covers over 1.7 million hectares in the southern Peruvian Amazon (Madre de Dios and Cusco regions).

Both datasets converge in the estimate of over 450 metric tons of aboveground biomass, equating to over 215 million metric tons of aboveground carbon across the territory (126.8 tons of carbon per hectare).

Figure 2 shows the detailed spatial distribution of this biomass across Manu National Park.

 

 

 

 

 

 

 

 

 

 

 

Madidi National Park (Bolivian Amazon)

Figure 3. Aboveground biomass across Madidi National Park (Bolivian Amazon). Data: GEDI-TanDEM-X

Madidi National Park and Integrated Management Area covers over 1.8 million hectares in the western Bolivian Amazon (La Paz department).

Both datasets converge in the estimate of over 350 metric tons of aboveground biomass, equating to over 160 million metric tons of aboveground carbon across the park (85.3 tons of carbon per hectare).

Figure 3 shows the detailed spatial distribution of this biomass across Madidi National Park. Note that the GEDI-TanDEM-X data misses the southern tip of the park.

 

 

 

 

 

 

 

 

 

 

Indigenous Territories

Kayapó Indigenous Territory (Brazilian Amazon)

Kayapó Indigenous Territory covers over 3.2 million hectares in the eastern Brazilian Amazon (Pará state). Both datasets converge in the estimate of over 413,000 metric tons of aboveground biomass, equating to over 198 million metric tons of aboveground carbon across the territory. Figure 4 shows the detailed spatial distribution of this biomass across Kayapó and four neighboring Indigenous Territories. Totaling across these five territories (10.4 million hectares), the data sets converge on over 1.5 billion metric tons of aboveground biomass, and 730 million metric tons of aboveground carbon (70 tons per hectare).

Figure 4. Aboveground biomass across Kayapó and neighboring Indigenous Territories (Brazilian Amazon). Data: GEDI-TanDEM-X

Barranco Chico Indigenous Territory (Peruvian Amazon)

Barranco Chico Indigenous Territory covers over 12,600 hectares in the southern Peruvian Amazon (Madre de Dios region). Both datasets converge in the estimate of over 2 million metric tons of aboveground biomass, equating to over 1 million metric tons of aboveground carbon. Figure 5 shows the detailed spatial distribution of this biomass across Barranco Chico and two neighboring Indigenous Territories (Puerto Luz and San Jose de Karene). Totaling across these three territories (nearly 90,000 hectares), the data sets converge on over 19 million metric tons of aboveground biomass, and over 9 million metric tons of aboveground carbon (102 tons per hectare).

Figure 5. Aboveground biomass across Barranco Chico and neighboring Indigenous Territories (Peruvian Amazon). Data: GEDI-TanDEM-X

Notes

1 GEDI L4B Gridded Aboveground Biomass Density, Version 2.1. This data is measured in megagrams of aboveground biomass per hectare (Mg/ha) at a 1-kilometer resolution, with the period of April 2019 – March 2023. This serves as our estimate for aboveground carbon reserves, with the science-based assumption that 48% of recorded biomass is carbon.

2 The approach relies on the foundational paper from Patterson et al., (2019) and it is used by the GEDI mission to estimate mean and total biomass worldwide (Dubayh et al., 2022, Armston et al., 2023). The method considers the spatial distribution of GEDI tracks within a given user-specify boundary to infer the sampling error component of the total uncertainty that also includes the error from the GEDI L4A models used to predict biomass from canopy height estimates (Keller et al., 2022). For more information on the OBI-WAN app, see Healey and Yang 2022.

3 GEDI-TanDEM-X (GTDX) is a fusion of GEDI Version 2 and TanDEM-X (TDX) Interferometric Synthetic Aperture Radar (InSAR) images (from Jan 2011 to December 2020). It also incorporates annual forest loss data to account for deforestation during this time. The GTDX aboveground biomass maps were produced based on a generalized hierarchical model-based (GHMB) framework that utilizes GEDI biomass as training data to establish models for estimating biomass based on the GTDX canopy height. The combination of lidar (GEDI) and radar (TanDEM-X) has started to produce unmatched maps that combine the ability of lidar to retrieve forest structure and the ability of radar to offer wall-to-wall coverage (Qi et al.,2023, Dubayah et a;., 2023). This fused product is a wall-to-wall gap-free map that was produced at multiple resolutions: 25m, 100m and 1ha. Ongoing processing over the Pantropic region will be made available over the next months but some geographies have been already mapped such as most of the Amazon Basin (Dubayah et al., 2023). The data we used is publicly available.

References

Armston, J., Dubayah, R. O., Healey, S. P., Yang, Z., Patterson, P. L., Saarela, S., Stahl, G., Duncanson, L., Kellner, J. R., Pascual, A., & Bruening, J. (2023). Global Ecosystem Dynamics Investigation (GEDI)GEDI L4B Country-level Summaries of Aboveground Biomass [CSV]. 0 MB. https://doi.org/10.3334/ORNLDAAC/2321

Dubayah, R. O., Armston, J., Healey, S. P., Yang, Z., Patterson, P. L., Saarela, S., Stahl, G., Duncanson, L., Kellner, J. R., Bruening, J., & Pascual, A. (2023). Global Ecosystem Dynamics Investigation (GEDI)GEDI L4B Gridded Aboveground Biomass Density, Version 2.1 [COG]. 0 MB. https://doi.org/10.3334/ORNLDAAC/2299

Dubayah, R., Armston, J., Healey, S. P., Bruening, J. M., Patterson, P. L., Kellner, J. R., Duncanson, L., Saarela, S., Ståhl, G., Yang, Z., Tang, H., Blair, J. B., Fatoyinbo, L., Goetz, S., Hancock, S., Hansen, M., Hofton, M., Hurtt, G., & Luthcke, S. (2022). GEDI launches a new era of biomass inference from space. Environmental Research Letters, 17(9), 095001. https://doi.org/10.1088/1748-9326/ac8694

Dubayah, R., Blair, J. B., Goetz, S., Fatoyinbo, L., Hansen, M., Healey, S., Hofton, M., Hurtt, G., Kellner, J., Luthcke, S., Armston, J., Tang, H., Duncanson, L., Hancock, S., Jantz, P., Marselis, S., Patterson, P. L., Qi, W., & Silva, C. (2020). The Global Ecosystem Dynamics Investigation: High-resolution laser ranging of the Earth’s forests and topography. Science of Remote Sensing, 1, 100002. https://doi.org/10.1016/j.srs.2020.100002

Healey S, Yang Z (2022) The OBIWAN App: Estimating Property-Level Carbon Storage Using NASA’s GEDI Lidar. https://www.fs.usda.gov/research/rmrs/understory/obiwan-app-estimating-property-level-carbon-storage-using-nasas-gedi-lidar

Kellner, J. R., Armston, J., & Duncanson, L. (2022). Algorithm Theoretical Basis Document for GEDI Footprint Aboveground Biomass Density. Earth and Space Science, 10(4), e2022EA002516. https://doi.org/10.1029/2022EA002516

Dubayah, R.O., W. Qi, J. Armston, T. Fatoyinbo, K. Papathanassiou, M. Pardini, A. Stovall, C. Choi, and V. Cazcarra-Bes. 2023. Pantropical Forest Height and Biomass from GEDI and TanDEM-X Data Fusion. ORNL DAAC, Oak Ridge, Tennessee, USA. https://doi.org/10.3334/ORNLDAAC/2298

Qi, W., J. Armston, C. Choi, A. Stovall, S. Saarela, M. Pardini, L. Fatoyinbo, K. Papathanasiou, and R. Dubayah. 2023. Mapping large-scale pantropical forest canopy height by integrating GEDI lidar and TanDEM-X InSAR data. Research Square. https://doi.org/10.21203/rs.3.rs-3306982/v1

Krieger, G., M. Zink, M. Bachmann, B. Bräutigam, D. Schulze, M. Martone, P. Rizzoli, U. Steinbrecher, J. Walter Antony, F. De Zan, I. Hajnsek, K. Papathanassiou, F. Kugler, M. Rodriguez Cassola, M. Younis, S. Baumgartner, P. López-Dekker, P. Prats, and A. Moreira. 2013. TanDEM-X: A radar interferometer with two formation-flying satellites. Acta Astronautica 89:83–98. https://doi.org/10.1016/j.actaastro.2013.03.008

Acknowledgments

We greatly thank the University of Maryland’s GEDI team for data access and reviewing this report. In particular, we thank Ralph Dubayah, Matheus Nunes, and Sean Healey.

Citation

Mamani N, Pascual A, Finer M (2024) Estimating Carbon in Amazon Protected Areas & Indigenous Territories. MAAP: 213

MAAP #212: Machine learning to detect mining deforestation across the Amazon

Posted on by dcadmin
Amazon Mining Watch. Screen shot of the interactive mining deforestation map, displaying data for 2023.

Gold Mining is one of the major deforestation drivers across the Amazon.*

It often targets remote areas, thus impacting carbon-rich primary forests. Moreover, in most cases, this mining is illegal, given that it is occurring in protected areas and indigenous territories.

Given the vastness of the Amazon, however, it has been a challenge to accurately monitor mining deforestation across the entire biome in a timely manner.

Here we present, for the first time, the results of a new machine learning based tool (known as Amazon Mining Watch)  that analyzes satellite imagery archives to detect mining deforestation across the entire Amazon.

Specifically, the tool produces 10-meter resolution mining deforestation alerts based on the European Space Agency’s Sentinel-2 satellite imagery. The alerts currently cover each year annually from 2018 to 2023.

This data reveals that gold mining is actively causing deforestation in all nine countries of the Amazon Biome (see Base Map below). The countries with the most overall mining deforestation are 1) Brazil, 2) Guyana, 3) Suriname, 4) Venezuela, and 5) Peru.

*Note that in this report we focus on mining activity that is causing deforestation. Additional critical gold mining areas in rivers (such as in northern Peru, southeast Colombia, and northwest Brazil; see MAAP #197), are not included in this report or detected/displayed in Amazon Mining Watch.

Major Findings

The Base Map below presents the mining deforestation data across the entire Amazon. Note that yellow indicates the historical mining footprint as of 2018, while red indicates the more recent mining deforestation between 2019 and 2023.

Although the alerts are pixels and not designed for precise area measurements, they can be used to give general estimates. For example, we estimate that as of 2018, there was a historical mining deforestation footprint of over 963,000 hectares across the entire Amazon. Between 2019 and 2023, we estimate that the mining deforestation footprint grew by over 944,000 hectares (2.3 million acres).

Thus, of the total accumulated mining deforestation footprint of over 1.9 million hectares (4.7 million acres), about half has occurred in just the past five years (see Annex).

In addition, we estimate that 38% (725,498 hectares) of the total mining deforestation occurred within protected areas and Indigenous territories.

Graph 1 shows, of the total accumulated mining, over half has occurred in Brazil (55%, covering over 1 million hectares), followed by Guyana (15%), Suriname (12%), Venezuela (7%), and Peru (7%, covering 135,625 hectares).

Base Map. Mining deforestation across the Amazon, based on data from Amazon Mining Watch, for the years 2018-2023. Data: AMW, ACA/MAAP.
Graph 1. Mining deforestation across the Amazon, by country. Data: AMW, ACA/MAAP.

Case Studies

In this section, we show a number of case studies highlighting the power of this data to see the evolution of mining deforestation in the following critical areas (see Insets A-E on Base Map). In these examples, note that yellow indicates the historical mining footprint as of 2018, purple indicates the expansion from 2019-2021, and red indicates the more recent mining deforestation between 2022 and 2023.

A. Southern Peruvian Amazon
B. Brazilian Amazon – Yanomami Indigenous Territory
C. Brazilian Amazon – Kayapó Indigenous Territory
D. Venezuelan Amazon – Yapacana National Park
E. Ecuadorian Amazon – Punino zone

A. Southern Peruvian Amazon

In southern Peru is one of the largest, and likely most emblematic, mining sites in the Amazon (see Inset A in Base Map). Figure 1 shows the dynamic evolution in this area, from several large core mining zones as of 2018, with more recent concentration in the designated Mining Corridor (large area where small-scale mining is permitted by the government as part of a formalization process).

Overall, we recorded over 135,000 hectares (333,590 acres) of mining deforestation in this area. Of this total, 62% (84,000 ha) was present as of 2018, while 38% (51,000 ha) has occurred in just the past five years (2019-2023).

We also highlight that of the total mining deforestation (135,000 ha), 59% has occurred within the Mining Corridor, while 41% (55,000 hectares) is outside the corridor and likely illegal. Note how mining deforestation threatens several protected areas, especially Tambopata National Reserve and Amarakaeri Communal Reserve.

See MAAP #208 for more information about mining deforestation at this site, and how illegal mining also threatens Native Communities.

Figure 1. Evolution of mining deforestation in the southern Peruvian Amazon. Data: AMW, ACA/MAAP.

B. Brazilian Amazon – Yanomami Indigenous Territory

In the northern Brazilian Amazon, the national government recently launched a series of raids against illegal gold mining in Yanomami Indigenous Territory (see Inset B in Base Map). Figure 2 shows a major escalation and expansion of gold mining deforestation since 2018, especially along the Uraricoera and Mucajai Rivers.

Specifically, we documented the total mining deforestation of over 19,000 hectares (47,000 acres) in Yanomami Indigenous Territory. It is critical to emphasize that the vast majority (93%) has occurred in just the past five years (2019-2023).

See MAAP #181 for more information about mining deforestation at this site.

Figure 2. Evolution of mining deforestation in Yanomami Indigenous Territory in Brazil. Data: AMW, ACA/MAAP.

C. Brazilian Amazon – Kayapó Indigenous Territory

In the eastern Brazilian Amazon, the Kayapó Indigenous Territory is also facing ongoing illegal mining (see Inset C in Base Map). Figure 3 shows the continuing expansion of mining deforestation, mostly in the eastern section of the territory.

We documented the total mining deforestation of nearly 50,000 hectares (123,550 acres) in Kayapó Indigenous Territory. Of this total, 60% (30,000 has) has occurred in just the past five years (2019-2023).

See MAAP #116 for more information about mining deforestation at this site, along with nearby Munduruku Indigenous Territory.

Figure 3. Evolution of mining deforestation in Kayapo Indigenous Territory in Brazil. Data: AMW, ACA/MAAP.

D. Venezuelan Amazon – Yapacana National Park

In Venezuela, we see the continued expansion of mining deforestation in Yapacana National Park (see Inset D in Base Map). Indeed, Figure 4 shows the steady expansion of gold mining deforestation at several sites in the southern section of the protected area.

We documented the total mining deforestation of over 6,000 hectares (14,800 acres) in Yapacana National Park. Of this total, just over half (52%; 3,000 has) has occurred in just the past five years (2019-2023).

See MAAP #173 and MAAP #207 for more information about mining deforestation at this site.

Figure 4. Evolution of mining deforestation in Yapacana National Park in Venezuela. Data: AMW, ACA/MAAP.

E. Ecuadorian Amazon – Punino River

In a series of reports, we have been showing the rapid increase in mining deforestation in the Ecuadorian Amazon (see MAAP #182). One of the main sites is around the Punino River in northern Ecuador (see Inset E in Base Map). Figure 5 shows the sudden emergence of gold mining deforestation near the river.

We documented the total mining deforestation of over 500 hectares (1,235 acres) in the Punino River area. Of this total, 100% is new, all starting in 2023.

See MAAP #206 for more information about mining deforestation at this site.

Figure 5. Evolution of mining deforestation in along the Punino River in Ecuador. Data: AMW, ACA/MAAP.

Annex

As noted above, of the total accumulated mining deforestation footprint of over 1.9 million hectares (4.7 million acres), about half has occurred in just the past five years.

Methods

All data for this report were obtained from Amazon Mining Watch. We only utilized patches with greater than 0.6 mean score. We used the 2018 data as our baseline. For 2019, we masked the previously reported 2018 data to only highlight the new mining that year. We then repeated this process for each subsequent year. For example, the 2023 data masked the 2018-2022 data, indicating only new mining deforestation that year.

Citation

Finer M, Ariñez A (2024) Machine learning to detect mining deforestation across the Amazon. MAAP: #212.