Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                
SlideShare a Scribd company logo

Global Pasture Watch: Mapping & Monitoring Global Grasslands and Livestock

•
•
World Resources Institute (WRI)
World Resources Institute (WRI)

This presentation outlines a new Land & Carbon Lab research consortium, Global Pasture Watch, which will contribute to better understanding land use conversion, food production, land productivity, and impacts for biodiversity and climate change at a global scale.

1 of 75
CREDIT: PETER IRUNGU/WRI Global Pasture Watch: Mapping & Monitoring Global Grasslands and Livestock contact: landcarbonlab@wri.org visit: landcarbonlab.org Photo by: Bobo Boom/ Flickr https://creativecommons.org/licenses/by/4.0/
Webinar registration stats Over 1,170 people registered from 111 countries!
PROTECT 30% of terrestrial areas by 2030 RESTORE 300 Mha of degraded land by 2030 PRODUCE more food, fuel, fiber on less land Together with our leading network of research partners, we build and deploy open geospatial data and monitoring solutions to help accelerate implementation and financing of NBS towards 2030 global targets
GLOBAL PASTURE WATCH A consortium of research partners developing global grassland and livestock monitoring data for impact
Agenda ● Transforming the livestock sector - Andy Jarvis (Bezos Earth Fund) ● The current state of global pasture and livestock mapping - Mario Herrero (Cornell) ● Partners and use cases - Lindsey Sloat (WRI) ● Project aspects - Leandro Parente (OpenGeoHub) ● Land conversion in the Cerrado - Laerte Ferreira (LAPIG) ● Network building - Nathália Teles (LAPIG) ● Question and answer session
CREDIT: PETER IRUNGU/WRI Transforming the livestock sector Andy Jarvis Bezos Earth Fund
CREDIT: PETER IRUNGU/WRI The current state of global pasture and livestock mapping Mario Herrero Cornell
We know a lot about grasslands, but everything is highly localised and not designed for global, regional or national assessments. ● Need better pasture / rangeland extent data ● Better spatially explicit global yields, lots of work on this at the moment ● We need yield gaps of pastures (what could happen to livestock production if we intensified grasslands?) ● Need useful classifications of grasslands – linked to management practices ● Much better data for GHG inventories ● Need linkages to cropland / forest – towards a unified system: essential for ecosystems services work
Grassland systems are carbon-neutral when also considering sparsely grazed areas Chang, Herrero et al 2021 Nat Comm.
Uncertainty estimates for key variables determining carrying capacity Fetzel et al. 2017 GBC
Large discrepancies in rangeland productivity maps Havlik et al … Conant and others LPJm G-Range We know very little about extent of intensive grasslands, we know even less about yields! Models with patchy performance
Disagreement in Africa in the cropland domain between GLC-2000 and MODIS v.5 Fritz et al ERL, 2012
Global livestock production systems Robinson et al 2011
Grazed biomass from livestock (Herrero et al PNAS 2013 ) Updates to 2005 and 2010 Grass represents 48% of the biomass consumed by livestock: 2.3 billion tonnes
Boone et al. 2017 G-Range: Climate change impacts on grassland functional groups
Global Pasture Watch: Mapping & Monitoring Global Grasslands and Livestock
Some rangelands and grasslands may have high carbon opportunity costs Hayek et al. 2020
Presentation title | Presenter name 18 | Piipponen et al. 2022 GCB Workflow for carrying capacity studies Laborious! Slow!
mario.herrero@cornell.edu Thank you. Warren Hall 250C Follow our team’s research on twitter @GlobalFoodTeam
CREDIT: PETER IRUNGU/WRI Partners and Use Cases Lindsey Sloat Research Associate II, Land & Carbon Lab
Global Pasture Watch: Mapping & Monitoring Global Grasslands and Livestock
Time series 2000 – 2022+ Grassland extent & management Pasture condition & productivity Livestock density & methane emissions
Many possible applications…
Many possible applications… ✔ Understanding the global land squeeze
Many possible applications… ✔ Understanding the global land squeeze ✔ Monitoring “conversion free” compliance
Many possible applications… ✔ Understanding the global land squeeze ✔ Monitoring “conversion free” compliance ✔ Restoration planning
Many possible applications… ✔ Understanding the global land squeeze ✔ Monitoring “conversion free” compliance ✔ Restoration planning ✔ Land sector greenhouse gas accounting
Many possible applications… ✔ Understanding the global land squeeze ✔ Monitoring “conversion free” compliance ✔ Restoration planning ✔ Land sector greenhouse gas accounting ✔ Improving livestock production models
Many possible applications… ✔ Understanding the global land squeeze ✔ Monitoring “conversion free” compliance ✔ Restoration planning ✔ Land sector greenhouse gas accounting ✔ Improving livestock production models ✔ Improved grazing land management
CREDIT: PETER IRUNGU/WRI Project aspects Leandro Parente Researcher, OpenGeoHub
Use Cases Earth Observation / Gridded data Reference samples Machine learning and GPW User communities
Use Cases Earth Observation / Gridded data Reference samples Machine learning # Mapping product Temporal resolution Spatial resolution Pixel value Output layers 1 Pasture-class maps (2000–2022+) Annual 30-m Two class of grass Dominant classes, probabilities and uncertainties per class 2 Livestock density maps (2000–2022+) Annual 1-km Number of animals per hectare Mean value and uncertainties 3 Short vegetation height maps (2000–2022+) Annual 30-m Canopy height in meters Mean value and uncertainties 4 Gross primary productivity maps (2000–2022+) Bi-monthly 30-m Kg of carbon per m2 Mean value and uncertainties Deliverables and GPW User communities
Use Cases Earth Observation / Gridded data Reference samples Machine learning # Mapping product Temporal resolution Spatial resolution Pixel value Output layers 1 Pasture-class maps (2000–2022+) Annual 30-m Two class of grass Dominant classes, probabilities and uncertainties per class 2 Livestock density maps (2000–2022+) Annual 1-km Number of animals per hectare Mean value and uncertainties 3 Short vegetation height maps (2000–2022+) Annual 30-m Canopy height in meters Mean value and uncertainties 4 Gross primary productivity maps (2000–2022+) Bi-monthly 30-m Kg of carbon per m2 Mean value and uncertainties Seeded grass Natural or semi-natural grass Deliverables
Use Cases Earth Observation / Gridded data Reference samples Machine learning # Mapping product Temporal resolution Spatial resolution Pixel value Output layers 1 Pasture-class maps (2000–2022+) Annual 30-m Two class of grass Dominant classes, probabilities and uncertainties per class 2 Livestock density maps (2000–2022+) Annual 1-km Number of animals per hectare Mean value and uncertainties 3 Short vegetation height maps (2000–2022+) Annual 30-m Canopy height in meters Mean value and uncertainties 4 Gross primary productivity maps (2000–2022+) Bi-monthly 30-m Kg of carbon per m2 Mean value and uncertainties Managed (according a certain livestock animal density) Unmanaged (mostly wild animals grazing) Deliverables
Use Cases Earth Observation / Gridded data Reference samples Machine learning # Mapping product Temporal resolution Spatial resolution Pixel value Output layers 1 Pasture-class maps (2000–2022+) Annual 30-m Two class of grass Dominant classes, probabilities and uncertainties per class 2 Livestock density maps (2000–2022+) Annual 1-km Number of animals per hectare Mean value and uncertainties 3 Short vegetation height maps (2000–2022+) Annual 30-m Canopy height in meters Mean value and uncertainties 4 Gross primary productivity maps (2000–2022+) Bi-monthly 30-m Kg of carbon per m2 Mean value and uncertainties Woody vegetation fraction Deliverables
Use Cases Earth Observation / Gridded data Reference samples Machine learning # Mapping product Temporal resolution Spatial resolution Pixel value Output layers 1 Pasture-class maps (2000–2022+) Annual 30-m Two class of grass Dominant classes, probabilities and uncertainties per class 2 Livestock density maps (2000–2022+) Annual 1-km Number of animals per hectare Mean value and uncertainties 3 Short vegetation height maps (2000–2022+) Annual 30-m Canopy height in meters Mean value and uncertainties 4 Gross primary productivity maps (2000–2022+) Bi-monthly 30-m Kg of carbon per m2 Mean value and uncertainties Low productivity High productivity depends on climate & soil (potential) + management Deliverables
Use Cases Earth Observation / Gridded data Reference samples Machine learning # Mapping product Temporal resolution Spatial resolution Pixel value Output layers 1 Pasture-class maps (2000–2022+) Annual 30-m Two class of grass Dominant classes, probabilities and uncertainties per class 2 Livestock density maps (2000–2022+) Annual 1-km Number of animals per hectare Mean value and uncertainties 3 Short vegetation height maps (2000–2022+) Annual 30-m Canopy height in meters Mean value and uncertainties 4 Gross primary productivity maps (2000–2022+) Bi-monthly 30-m Kg of carbon per m2 Mean value and uncertainties + Input data ML models Source code Deliverables
Fully reproducible deliverables PENG R., 2011 Cloud-optimized and analysis-ready (ARCO), open catalog (STAC), computational notebooks, and software libraries
Use Cases Earth Observation / Gridded data Reference samples Machine learning # Mapping product Temporal resolution Spatial resolution Pixel value Output layers 1 Pasture-class maps (2000–2022+) Annual 30-m Two class of grass Dominant classes, probabilities and uncertainties per class 2 Livestock density maps (2000–2022+) Annual 1-km Number of animals per hectare Mean value and uncertainties 3 Short vegetation height maps (2000–2022+) Annual 30-m Canopy height in meters Mean value and uncertainties 4 Gross primary productivity maps (2000–2022+) Bi-monthly 30-m Kg of carbon per m2 Mean value and uncertainties Further versions Feedback loop
Area of interest Analysis Mapping products ZOOM poll questions
GPW product integration Class of grass Vegetation height Livestock density Combination + threshold values
GPW product integration Class of grass Vegetation height Livestock density Combination + threshold values 8 (preliminary) integrated pasture classes Seeded pasture Seeded pasture with shrub Managed Seeded pasture Seeded pasture with shrub Unmanaged Natural pasture Natural browse land Unmanaged Managed Semi-natural pasture Semi-natural browse land
GPW product integration Class of grass Vegetation height Livestock density Combination + threshold values 8 (preliminary) integrated pasture classes Seeded pasture Seeded pasture with shrub Managed Seeded pasture Seeded pasture with shrub Unmanaged Natural pasture Natural browse land Unmanaged Managed Semi-natural pasture Semi-natural browse land Low productivity High productivity GPP
Earth Observation / Gridded data (multi-source) Jolly, Texas, USA Global multi-source and multi-resolution data cube
20 analysts (graduate and undergraduate students) 4 supervisors ensure interpretation consistency Reference samples - Visual interpretation https://github.com/lapig-ufg/inspection-tiles
Number of animals (cattle) Gridded Livestock of the World (GLW 3 - Gilbert et al., 2018) + time Harmonized global database for livestock census data (vector format according to GADM 4.1) Reference samples - Census data
Ground and vegetation structure (ATL03 and ATL08) https://solarsystem.nasa.gov/news/534/10-things-to-know-about-icesat-2-nasas-latest-space-laser/ Reference samples ICESat-2 Laser instrument on Earth's orbits
212 sites providing measurements on flux exchange (incl. GPP) for different time periods available between 1991 and 2014 - Pastorello et al. (2020) Reference samples - FLUXNET2015
Data preparation + Data collection + Modeling Pasture / grasslands Selected areas Minimum Viable Product (MVP)
Preliminary results Pasture-class Seeded grass Natural or semi-natural grass 85 % 25 % Mato Grosso do Sul, Brazil - 2008 (-18.1242, -55.1707) Derived by Ensemble Machine Learning (Witjes et al., 2021)
Preliminary results Pasture-class Seeded grass Natural or semi-natural grass 85 % 25 % Les Gets, France - 2020 (46.1388, 6.6605) Derived by Ensemble Machine Learning (Witjes et al., 2021)
Time span 2000—2021 Witjes et al., 2022 Rabenau, Germany (50.9551, 13.6278) Preliminary results Quarterly GPP
Derived by Light use efficiency (LUE) model (MOD17A2 - Robinson et al., 2018) Modeled x In-situ (FLUXnet 2015) Preliminary results Quarterly GPP
2,139 in-situ measures Derived by Light use efficiency (LUE) model (MOD17A2 - Robinson et al., 2018) Preliminary results Quarterly GPP
Next steps for 2023 ➢ Produce the reference samples for seeded and natural / semi-natural grass ➢ Produce preliminary maps for MVP areas (U.S., Europe and Brazil) ➢ Prepare the Earth Observation / Gridded data (Landsat ARD-2) ➢ Prepare vegetation height and livestock census data ➢ Produce first version of Global Pasture Watch maps
CREDIT: PETER IRUNGU/WRI Pasture Quality and Pasture Conversion in the Brazilian Cerrado Laerte Ferreira LAPIG
A country in a fast-pace conversion mode 294.2 Mha ±1.44% are anthropized in Brazil about 35% of the national territory… ~1/3 of the anthropization in Brazil occurred in the last four decades…
90% was pasture at some point... ~160 Mha of pastures ~19% of Brazil ~40% > 30 yrs ~25% < 10 yrs ~102 Mha converted ~46% < 10 yrs
~47 Mha of pastures 2020 IBGE biome limit The threatened Cerrado biome…
~47 Mha of pastures 2020 IBGE biome limit ~45% > 30 yrs ~21% < 10 yrs ~34 Mha converted ~46% < 10 yrs The threatened Cerrado biome…
~47 Mha of pastures 2020 IBGE biome limit ~45% > 30 yrs ~21% < 10 yrs ~34 Mha converted ~46% < 10 yrs The threatened Cerrado biome… Since 2013, the Amazon pastures have surpassed the area of pastures in the Cerrado
Degraded Cerrado pastures & soy conversion potential… 2004 IBGE biome limit Soy suitability data from Agrosatélite / GBMF (2014) About 60% of the pastures in the Cerrado are at some degradation stage… & about 11 Mha of pastures, degraded at some degree, are highly suitable for soy expansion
Cerrado Pasture soil organic carbon 0,78 Gt.C 30,2 ±3,9 Mg. C.ha-1 0,33 Gt.C 29,8 ±3,8 Mg. C.ha-1 0,71 Gt.C 36,7 ±4,5 Mg. C.ha-1 Degradation class Intermediate Severe Absent
Increase in the Cerrado pasture SOC
Increase in the Cerrado pasture SOC Carbon Offset ~ 79% ~ 50%
A more efficient use of cattle pastures in the Cerrado is a key factor for Brazil to meet its GHG reduction targets, while simultaneously reducing habitat loss, addressing food security and trade goals. A slight increase in the average animal unit (AU), from ~1,12 to ~1,43, would make available approximately 11 Mha of additional land for soy expansion.
CREDIT: PETER IRUNGU/WRI Network building Nathália Teles LAPIG
Goal To establish a Research and User Network involving several sectors and organizations worldwide to talk, learn and exchange data about Grasslands.
Strategies 01 02 Outreaching to institutions, groups and people Direct contact, meetings, events, webinars and workshops Researching data on grassland Who has data on grasslands? What kind of data is out there and where?
What do we need? Local knowledge and demands For improving our products over time and understand landscape complexities Ground data for validation of our products Natural x Managed grasslands GPP; Short vegetation Feedback on product usability Product consistency and adjustment Post-Webinar Survey 01 02 03
What will we be giving back? GPW Products Capacity Building Workshops and Webinars 01 02 03
What’s next? Talk to stakeholders We will reach out to people and organizations to understand their work and needs Reference database Literature review Ground data from partners Productivity assessment Local/regional knowledge on grasslands Collaborate in real time Engage our Global Pasture Watch initiative! 01 02 03
Thank you landcarbonlab@wri.org https://linkedin.com/in/nathaliateles/ Please contact us, and take part in the survey.
CREDIT: PETER IRUNGU/WRI Questions?
Photo by: Bobo Boom/ Flickr Global Pasture Watch: Mapping & Monitoring Global Grasslands and Livestock contact: landcarbonlab@wri.org visit: landcarbonlab.org

More Related Content

Global Pasture Watch: Mapping & Monitoring Global Grasslands and Livestock

  • 1. CREDIT: PETER IRUNGU/WRI Global Pasture Watch: Mapping & Monitoring Global Grasslands and Livestock contact: landcarbonlab@wri.org visit: landcarbonlab.org Photo by: Bobo Boom/ Flickr https://creativecommons.org/licenses/by/4.0/
  • 2. Webinar registration stats Over 1,170 people registered from 111 countries!
  • 3. PROTECT 30% of terrestrial areas by 2030 RESTORE 300 Mha of degraded land by 2030 PRODUCE more food, fuel, fiber on less land Together with our leading network of research partners, we build and deploy open geospatial data and monitoring solutions to help accelerate implementation and financing of NBS towards 2030 global targets
  • 4. GLOBAL PASTURE WATCH A consortium of research partners developing global grassland and livestock monitoring data for impact
  • 5. Agenda ● Transforming the livestock sector - Andy Jarvis (Bezos Earth Fund) ● The current state of global pasture and livestock mapping - Mario Herrero (Cornell) ● Partners and use cases - Lindsey Sloat (WRI) ● Project aspects - Leandro Parente (OpenGeoHub) ● Land conversion in the Cerrado - Laerte Ferreira (LAPIG) ● Network building - Nathália Teles (LAPIG) ● Question and answer session
  • 6. CREDIT: PETER IRUNGU/WRI Transforming the livestock sector Andy Jarvis Bezos Earth Fund
  • 7. CREDIT: PETER IRUNGU/WRI The current state of global pasture and livestock mapping Mario Herrero Cornell
  • 8. We know a lot about grasslands, but everything is highly localised and not designed for global, regional or national assessments. ● Need better pasture / rangeland extent data ● Better spatially explicit global yields, lots of work on this at the moment ● We need yield gaps of pastures (what could happen to livestock production if we intensified grasslands?) ● Need useful classifications of grasslands – linked to management practices ● Much better data for GHG inventories ● Need linkages to cropland / forest – towards a unified system: essential for ecosystems services work
  • 9. Grassland systems are carbon-neutral when also considering sparsely grazed areas Chang, Herrero et al 2021 Nat Comm.
  • 10. Uncertainty estimates for key variables determining carrying capacity Fetzel et al. 2017 GBC
  • 11. Large discrepancies in rangeland productivity maps Havlik et al … Conant and others LPJm G-Range We know very little about extent of intensive grasslands, we know even less about yields! Models with patchy performance
  • 12. Disagreement in Africa in the cropland domain between GLC-2000 and MODIS v.5 Fritz et al ERL, 2012
  • 13. Global livestock production systems Robinson et al 2011
  • 14. Grazed biomass from livestock (Herrero et al PNAS 2013 ) Updates to 2005 and 2010 Grass represents 48% of the biomass consumed by livestock: 2.3 billion tonnes
  • 15. Boone et al. 2017 G-Range: Climate change impacts on grassland functional groups
  • 17. Some rangelands and grasslands may have high carbon opportunity costs Hayek et al. 2020
  • 18. Presentation title | Presenter name 18 | Piipponen et al. 2022 GCB Workflow for carrying capacity studies Laborious! Slow!
  • 19. mario.herrero@cornell.edu Thank you. Warren Hall 250C Follow our team’s research on twitter @GlobalFoodTeam
  • 20. CREDIT: PETER IRUNGU/WRI Partners and Use Cases Lindsey Sloat Research Associate II, Land & Carbon Lab
  • 22. Time series 2000 – 2022+ Grassland extent & management Pasture condition & productivity Livestock density & methane emissions
  • 24. Many possible applications… ✔ Understanding the global land squeeze
  • 25. Many possible applications… ✔ Understanding the global land squeeze ✔ Monitoring “conversion free” compliance
  • 26. Many possible applications… ✔ Understanding the global land squeeze ✔ Monitoring “conversion free” compliance ✔ Restoration planning
  • 27. Many possible applications… ✔ Understanding the global land squeeze ✔ Monitoring “conversion free” compliance ✔ Restoration planning ✔ Land sector greenhouse gas accounting
  • 28. Many possible applications… ✔ Understanding the global land squeeze ✔ Monitoring “conversion free” compliance ✔ Restoration planning ✔ Land sector greenhouse gas accounting ✔ Improving livestock production models
  • 29. Many possible applications… ✔ Understanding the global land squeeze ✔ Monitoring “conversion free” compliance ✔ Restoration planning ✔ Land sector greenhouse gas accounting ✔ Improving livestock production models ✔ Improved grazing land management
  • 30. CREDIT: PETER IRUNGU/WRI Project aspects Leandro Parente Researcher, OpenGeoHub
  • 31. Use Cases Earth Observation / Gridded data Reference samples Machine learning and GPW User communities
  • 32. Use Cases Earth Observation / Gridded data Reference samples Machine learning # Mapping product Temporal resolution Spatial resolution Pixel value Output layers 1 Pasture-class maps (2000–2022+) Annual 30-m Two class of grass Dominant classes, probabilities and uncertainties per class 2 Livestock density maps (2000–2022+) Annual 1-km Number of animals per hectare Mean value and uncertainties 3 Short vegetation height maps (2000–2022+) Annual 30-m Canopy height in meters Mean value and uncertainties 4 Gross primary productivity maps (2000–2022+) Bi-monthly 30-m Kg of carbon per m2 Mean value and uncertainties Deliverables and GPW User communities
  • 33. Use Cases Earth Observation / Gridded data Reference samples Machine learning # Mapping product Temporal resolution Spatial resolution Pixel value Output layers 1 Pasture-class maps (2000–2022+) Annual 30-m Two class of grass Dominant classes, probabilities and uncertainties per class 2 Livestock density maps (2000–2022+) Annual 1-km Number of animals per hectare Mean value and uncertainties 3 Short vegetation height maps (2000–2022+) Annual 30-m Canopy height in meters Mean value and uncertainties 4 Gross primary productivity maps (2000–2022+) Bi-monthly 30-m Kg of carbon per m2 Mean value and uncertainties Seeded grass Natural or semi-natural grass Deliverables
  • 34. Use Cases Earth Observation / Gridded data Reference samples Machine learning # Mapping product Temporal resolution Spatial resolution Pixel value Output layers 1 Pasture-class maps (2000–2022+) Annual 30-m Two class of grass Dominant classes, probabilities and uncertainties per class 2 Livestock density maps (2000–2022+) Annual 1-km Number of animals per hectare Mean value and uncertainties 3 Short vegetation height maps (2000–2022+) Annual 30-m Canopy height in meters Mean value and uncertainties 4 Gross primary productivity maps (2000–2022+) Bi-monthly 30-m Kg of carbon per m2 Mean value and uncertainties Managed (according a certain livestock animal density) Unmanaged (mostly wild animals grazing) Deliverables
  • 35. Use Cases Earth Observation / Gridded data Reference samples Machine learning # Mapping product Temporal resolution Spatial resolution Pixel value Output layers 1 Pasture-class maps (2000–2022+) Annual 30-m Two class of grass Dominant classes, probabilities and uncertainties per class 2 Livestock density maps (2000–2022+) Annual 1-km Number of animals per hectare Mean value and uncertainties 3 Short vegetation height maps (2000–2022+) Annual 30-m Canopy height in meters Mean value and uncertainties 4 Gross primary productivity maps (2000–2022+) Bi-monthly 30-m Kg of carbon per m2 Mean value and uncertainties Woody vegetation fraction Deliverables
  • 36. Use Cases Earth Observation / Gridded data Reference samples Machine learning # Mapping product Temporal resolution Spatial resolution Pixel value Output layers 1 Pasture-class maps (2000–2022+) Annual 30-m Two class of grass Dominant classes, probabilities and uncertainties per class 2 Livestock density maps (2000–2022+) Annual 1-km Number of animals per hectare Mean value and uncertainties 3 Short vegetation height maps (2000–2022+) Annual 30-m Canopy height in meters Mean value and uncertainties 4 Gross primary productivity maps (2000–2022+) Bi-monthly 30-m Kg of carbon per m2 Mean value and uncertainties Low productivity High productivity depends on climate & soil (potential) + management Deliverables
  • 37. Use Cases Earth Observation / Gridded data Reference samples Machine learning # Mapping product Temporal resolution Spatial resolution Pixel value Output layers 1 Pasture-class maps (2000–2022+) Annual 30-m Two class of grass Dominant classes, probabilities and uncertainties per class 2 Livestock density maps (2000–2022+) Annual 1-km Number of animals per hectare Mean value and uncertainties 3 Short vegetation height maps (2000–2022+) Annual 30-m Canopy height in meters Mean value and uncertainties 4 Gross primary productivity maps (2000–2022+) Bi-monthly 30-m Kg of carbon per m2 Mean value and uncertainties + Input data ML models Source code Deliverables
  • 38. Fully reproducible deliverables PENG R., 2011 Cloud-optimized and analysis-ready (ARCO), open catalog (STAC), computational notebooks, and software libraries
  • 39. Use Cases Earth Observation / Gridded data Reference samples Machine learning # Mapping product Temporal resolution Spatial resolution Pixel value Output layers 1 Pasture-class maps (2000–2022+) Annual 30-m Two class of grass Dominant classes, probabilities and uncertainties per class 2 Livestock density maps (2000–2022+) Annual 1-km Number of animals per hectare Mean value and uncertainties 3 Short vegetation height maps (2000–2022+) Annual 30-m Canopy height in meters Mean value and uncertainties 4 Gross primary productivity maps (2000–2022+) Bi-monthly 30-m Kg of carbon per m2 Mean value and uncertainties Further versions Feedback loop
  • 40. Area of interest Analysis Mapping products ZOOM poll questions
  • 41. GPW product integration Class of grass Vegetation height Livestock density Combination + threshold values
  • 42. GPW product integration Class of grass Vegetation height Livestock density Combination + threshold values 8 (preliminary) integrated pasture classes Seeded pasture Seeded pasture with shrub Managed Seeded pasture Seeded pasture with shrub Unmanaged Natural pasture Natural browse land Unmanaged Managed Semi-natural pasture Semi-natural browse land
  • 43. GPW product integration Class of grass Vegetation height Livestock density Combination + threshold values 8 (preliminary) integrated pasture classes Seeded pasture Seeded pasture with shrub Managed Seeded pasture Seeded pasture with shrub Unmanaged Natural pasture Natural browse land Unmanaged Managed Semi-natural pasture Semi-natural browse land Low productivity High productivity GPP
  • 44. Earth Observation / Gridded data (multi-source) Jolly, Texas, USA Global multi-source and multi-resolution data cube
  • 45. 20 analysts (graduate and undergraduate students) 4 supervisors ensure interpretation consistency Reference samples - Visual interpretation https://github.com/lapig-ufg/inspection-tiles
  • 46. Number of animals (cattle) Gridded Livestock of the World (GLW 3 - Gilbert et al., 2018) + time Harmonized global database for livestock census data (vector format according to GADM 4.1) Reference samples - Census data
  • 47. Ground and vegetation structure (ATL03 and ATL08) https://solarsystem.nasa.gov/news/534/10-things-to-know-about-icesat-2-nasas-latest-space-laser/ Reference samples ICESat-2 Laser instrument on Earth's orbits
  • 48. 212 sites providing measurements on flux exchange (incl. GPP) for different time periods available between 1991 and 2014 - Pastorello et al. (2020) Reference samples - FLUXNET2015
  • 49. Data preparation + Data collection + Modeling Pasture / grasslands Selected areas Minimum Viable Product (MVP)
  • 50. Preliminary results Pasture-class Seeded grass Natural or semi-natural grass 85 % 25 % Mato Grosso do Sul, Brazil - 2008 (-18.1242, -55.1707) Derived by Ensemble Machine Learning (Witjes et al., 2021)
  • 51. Preliminary results Pasture-class Seeded grass Natural or semi-natural grass 85 % 25 % Les Gets, France - 2020 (46.1388, 6.6605) Derived by Ensemble Machine Learning (Witjes et al., 2021)
  • 52. Time span 2000—2021 Witjes et al., 2022 Rabenau, Germany (50.9551, 13.6278) Preliminary results Quarterly GPP
  • 53. Derived by Light use efficiency (LUE) model (MOD17A2 - Robinson et al., 2018) Modeled x In-situ (FLUXnet 2015) Preliminary results Quarterly GPP
  • 54. 2,139 in-situ measures Derived by Light use efficiency (LUE) model (MOD17A2 - Robinson et al., 2018) Preliminary results Quarterly GPP
  • 55. Next steps for 2023 ➢ Produce the reference samples for seeded and natural / semi-natural grass ➢ Produce preliminary maps for MVP areas (U.S., Europe and Brazil) ➢ Prepare the Earth Observation / Gridded data (Landsat ARD-2) ➢ Prepare vegetation height and livestock census data ➢ Produce first version of Global Pasture Watch maps
  • 56. CREDIT: PETER IRUNGU/WRI Pasture Quality and Pasture Conversion in the Brazilian Cerrado Laerte Ferreira LAPIG
  • 57. A country in a fast-pace conversion mode 294.2 Mha ±1.44% are anthropized in Brazil about 35% of the national territory… ~1/3 of the anthropization in Brazil occurred in the last four decades…
  • 58. 90% was pasture at some point... ~160 Mha of pastures ~19% of Brazil ~40% > 30 yrs ~25% < 10 yrs ~102 Mha converted ~46% < 10 yrs
  • 59. ~47 Mha of pastures 2020 IBGE biome limit The threatened Cerrado biome…
  • 60. ~47 Mha of pastures 2020 IBGE biome limit ~45% > 30 yrs ~21% < 10 yrs ~34 Mha converted ~46% < 10 yrs The threatened Cerrado biome…
  • 61. ~47 Mha of pastures 2020 IBGE biome limit ~45% > 30 yrs ~21% < 10 yrs ~34 Mha converted ~46% < 10 yrs The threatened Cerrado biome… Since 2013, the Amazon pastures have surpassed the area of pastures in the Cerrado
  • 62. Degraded Cerrado pastures & soy conversion potential… 2004 IBGE biome limit Soy suitability data from Agrosatélite / GBMF (2014) About 60% of the pastures in the Cerrado are at some degradation stage… & about 11 Mha of pastures, degraded at some degree, are highly suitable for soy expansion
  • 63. Cerrado Pasture soil organic carbon 0,78 Gt.C 30,2 ±3,9 Mg. C.ha-1 0,33 Gt.C 29,8 ±3,8 Mg. C.ha-1 0,71 Gt.C 36,7 ±4,5 Mg. C.ha-1 Degradation class Intermediate Severe Absent
  • 64. Increase in the Cerrado pasture SOC
  • 65. Increase in the Cerrado pasture SOC Carbon Offset ~ 79% ~ 50%
  • 66. A more efficient use of cattle pastures in the Cerrado is a key factor for Brazil to meet its GHG reduction targets, while simultaneously reducing habitat loss, addressing food security and trade goals. A slight increase in the average animal unit (AU), from ~1,12 to ~1,43, would make available approximately 11 Mha of additional land for soy expansion.
  • 67. CREDIT: PETER IRUNGU/WRI Network building Nathália Teles LAPIG
  • 68. Goal To establish a Research and User Network involving several sectors and organizations worldwide to talk, learn and exchange data about Grasslands.
  • 69. Strategies 01 02 Outreaching to institutions, groups and people Direct contact, meetings, events, webinars and workshops Researching data on grassland Who has data on grasslands? What kind of data is out there and where?
  • 70. What do we need? Local knowledge and demands For improving our products over time and understand landscape complexities Ground data for validation of our products Natural x Managed grasslands GPP; Short vegetation Feedback on product usability Product consistency and adjustment Post-Webinar Survey 01 02 03
  • 71. What will we be giving back? GPW Products Capacity Building Workshops and Webinars 01 02 03
  • 72. What’s next? Talk to stakeholders We will reach out to people and organizations to understand their work and needs Reference database Literature review Ground data from partners Productivity assessment Local/regional knowledge on grasslands Collaborate in real time Engage our Global Pasture Watch initiative! 01 02 03
  • 75. Photo by: Bobo Boom/ Flickr Global Pasture Watch: Mapping & Monitoring Global Grasslands and Livestock contact: landcarbonlab@wri.org visit: landcarbonlab.org