A stepwise approach to reference levels
Download as PPTX, PDF2 likes1,418 views
The document discusses a stepwise approach to assessing reference levels for deforestation and forest degradation across various countries, using regression models to predict future rates based on historical data. It highlights the significance of forest definitions and the importance of data quality in evaluating national deforestation estimates, particularly in Indonesia. The analysis also emphasizes the integration of socioeconomic factors to improve predictive models and better understand the drivers of deforestation.
A stepwise approach to reference levels
- 1. A stepwise approach to reference levels Louis Verchot, Arild Angelsen, Martin Herold, Arief Wijaya
- 2. A stepwise approach to FREL/FRLs
- 3. Criteria for comparing country circumstances and strategies
- 4. Deforestation/degradation drivers for each continent AMERICA -2% -4% AFRICA ASIA -2% -1% -7% -11% Deforestation -10% -39% -13% -41% -7% -36% -57% -37% -35% 4% 4% 8% 17% Degradation 6% 26% 7% 20% 9% 67% 70% 62% Deforestation driver Forest degradation driver THINKING beyond the canopy
- 5. RLs using regression models – Simple, easy to understand and test new variables – But, data demanding – Predicting deforestation in a period: Pt – Pt+1, based on deforestation in the previous period Pt-1 – Pt and a set of other factors (observed at time t). – Using structure (coefficients) from the estimated regression equation to predict deforestation in period Pt+1 – Pt+2, based on observed values at time t+1 2000 2004 2005 Historical deforestation 2009 2010 Estimated/Predicted deforestation Regression model Predictive model, based on structure from regression model 5
- 6. Step 1 case for 4 countries using FAO FRA data Indonesia 3,500 Forest C stock (Mt) Forest C stock (Mt) Cameroon 3,000 2,500 2,000 1,500 1,000 500 0 18,000 16,000 14,000 12,000 10,000 8,000 6,000 4,000 2,000 0 1985 1990 1995 2000 2005 2010 2015 2020 2025 1985 1990 1995 2000 2005 2010 2015 2020 2025 Year Year Forest C stock (Mt) Forest C stock (Mt) Brazil Vietnam 1,500 1,200 900 600 300 70,000 60,000 50,000 40,000 30,000 20,000 10,000 0 0 1985 80,000 1990 1995 2000 2005 Year 2010 2015 2020 2025 1985 1990 1995 2000 2005 2010 2015 2020 2025 Year
- 7. Step 2: Brazil Predict deforestation rates for legal Amazon 2005- 2009 Category Deforestation rate (2000-2004) Trend variable Deforestation dummy Forest stock Forest stock squared Log per capita GDP Agric GDP (%GDP) Population density Road denisty R2 N Regression coefficient 0.395 -0.136 -0.373 2.18 -1.8 -0.034 0.28 0.081 0.039 -0.145 -0.773 4.756 -3.826 -0.13 0.28 -0.81 0.076 0.831 3595 0.789 3595
- 8. Step 2: Vietnam Predict deforestation rates 2005- 2009 Category Deforestation rate (2000-2004) Trend variable Deforestation dummy Forest stock Forest stock squared Population density Road denisty R2 N Regression coefficient 1.464 -0.006 -0.011 0.067 -0.189 -1.177 0.004 0.003 -0.031 0.260 -0.463 1.036 -0.001 0.515 301 0.052 301
- 9. Preliminary conclusions Historical def. is key to predict future deforestation – Coefficients below one misleading simple extrapolation can be Some evidence of forest transition (FT) hypothesis – Robustness of FT depends on the measure of forest stock FT supported when forest stock is measured relative to total land area, otherwise mixed results emerge Other national circumstances have contradictory effects Contradictory relationships may be linked to data quality and interrelations of econ. & institutions differ THINKING beyond the canopy
- 10. In-depth case study: Indonesia definitions matter FAO forest definition – minimum 10% crown cover, minimum 0.5 ha and minimum height 5 m Indonesia national forest definition – vegetation cover dominated by intertwined tree crowns with canopy cover of more than 60% Indonesia – vegetation cover dominated by trees, with canopy cover between 25 and 60% is defined as bush Natural forest definition – no plantations THINKING beyond the canopy
- 11. Forest definitions affect estimates of deforestation THINKING beyond the canopy
- 12. THINKING beyond the canopy
- 13. Assessment of national REL/RL for Indonesia Cumulative Emission from LUCF 2000 -2009 (in Gg CO2e)* Source Methods 3,140,033 FRA country report (EF = 138 ton C/ha) 7,443,064 IPCC Guidelines 2006 3,468,150 Carbon Book keeping model (RS + Field) MOF (official) 1,760,000 Approach 1 + NFI (Tier 1 or 2) MOF + Saatchi (CIFOR) 1,811,396 Approach 1 + Global EF (Tier 1 or 2) FAOStat MoE - Second National Communication to UNFCCC Winrock International (Harris, 2012) * does not include peat emissions and peat fire
- 14. Comparison of national deforestation estimates
- 15. Validation of deforestation maps of Indonesia Source: Wijaya, et.al, (In prep)
- 16. Validation of deforestation maps 1000 Annual Deforestation (x 1000 ha) 900 800 700 600 500 400 300 200 100 0 Indonesia MOFOR Indonesia Hansen Indonesia JRC Indonesia Mean
- 17. Previous deforestation rates are good predictors of future rates Using deforestation rates in 2003 to 2006 to predict deforestation in 2006 to 2009 National Bali Java Kalimantan Maluku & Papua Log his def. 0.942 0.781 1.270 1.059 1.187 0.563 1.032 R2 0.574 0.517 0.187 0.869 0.848 0.589 0.524 372 32 114 43 25 47 111 Num. of obs Sulawesi Sumatera
- 18. Including socioeconomic factors improves the regressions National 0.289 10.121 Bali 0.507 -2.019 Java 0.532 27.345 Kalimantan 0.277 23.192 Maluku & Papua 0.662 1.166 -8.829 2.342 -43.279 -19.797 6.328 -8.653 -19.510 1.432 0.456 -0.255 -0.038 0.381 -1.136 1.688 0.033 0.015 -0.027 0.032 0.002 0.004 0.069 Log Pop. den. -0.357 0.291 0.145 0.089 -0.738 -0.404 -0.853 Road density -2.816 -4.355 0.000 0.494 5.134 6.912 1.089 R-square Num. of obs 0.777 371 0.665 32 0.549 114 0.980 43 0.965 25 0.707 47 0.858 110 Log his def. Forest stock Forest stock sq Log District GDP per capita Agric. GDP Sulawesi 0.299 14.658 Sumatera 0.116 18.523
- 20. Observations so far… Forest definition matters Selection of minimum mapping unit is important to determine the smallest units of deforested areas Different satellite image classification methods may result in different estimate There are several useful approaches to integrating drivers of deforestation and forest degradation into assessments of RELs
- 21. Thank you