Summary of relevant works Fabin Santos 05.06.2014 Deforestation on - - PowerPoint PPT Presentation

Summary of relevant works Fabián Santos 05.06.2014

Deforestation on tropical forests from image processing of Landsat 7 with scan-off and clouds cover in the sector Auca Sur, Yasuni National Park – Ecuador (master degree dissertation)

Background:

• Ecuador had for some

years, the highest deforestation rate in Southamerica (FAO, 2010)

• Lack of deforestation

methodologies which treat bad quality Landsat images Objective:

• Predict deforestation in tropical

forests with high cloud cover and scan-off Landsat images Study area: Pre-processing chain: Processing chain:

Soil – GV – NPV R – G – B

Scan-off & Data cloud gap fill with ERDAS Modeler (Leica

Post-processing:

Visual inspection and edition of Spectral mixture decomposition with ImageTools (IMAZON, 2013): *Green vegetation *Non-photosynthetic vegetation *Soils *Shades Radiance Conversion with ImageTools (IMAZON, 2013) Atmospheric Correction with FLAASH (ITT, 2013) Normalized Difference Fraction Index or NDFI (Souza, 2006) Modeler (Leica Geosystems, 2007) Haze filter Enhancement with ImageTools (IMAZON, 2013) NDFI classification with ImageTools (IMAZON, 2013) and edition of errors High resolution images & field data

• n a stratified

sampling for a confusion matrix

Modelling deforestation scenarios: 1 2 34 5 6 7 8 9 10 11 12 13 14

Finding out the drivers of deforestation & validation

• f models (Dinamica Ego,

2009) Predictors + deforestation control maps (Dinamica Ego, 2009)

Results:

• Three model scripts for recover scan-off and

cloud data gaps

• Kappa values for the deforestation maps (2000 –

2008) were over 0.76 and the calibration of the deforestation scenario achieve 0.67

• Square matrix style calculations, helps to reveals

hotspots of deforestation Other relevant issues:

• Imgtools (IMAZON, 2013) methodology was

applied for the whole Amazon Region of Ecuador and publicy on the Atlas of “Amazonia bajo Presion” (RAISG, 2012) Calculation of deforestation rates:

Appliance of “Simulate Deforestation with Patch formation and Expansion“ model (Dinamica Ego, 2009) Square matrix (25 km2) for obtain deforestation taxes (FAO & Puyravaud formulas)

More information at: http://raisg.socioambiental.org/ bajo Presion” (RAISG, 2012)

Time period Tax calculation Site identification

A framework for Mapping Potential Strata Forests on Ecuador (technical report)

Background:

• For REDD+ purposes, a

Potential Strata Forests Map was request for quantify greenhouse gases emissions

• Previous version of a

potential map (MAE, 2013) had a not clear and replicable methodology Objective:

• Update the Potential Strata

Forest Map with a replicable methodology Study area: Flowchart: Decision tree formulas:

Predictor variables collection Hexagon database system for data collection (25 ha each analysis unit) Possible unique variable combinations and regresion formulas

Data extraction:

Ecuador regions Remaining strata forests collection (total 11) Extraction of data Outliers detection and elimination Samples collection Samples Output 1 + Output 2 + 1. Data extraction 2. Decision tree creation 3. Prediction & validation 4. Map plotting 5. Discussion & edition Output N

Prediction & Validation:

Results: Map plotting:

Final result after editing the errors identified Workshop for identify the quality of the results

Other relevant issues:

• The methodology could be applied in other

distribution exercises, as for example species or crops potential distributions, also climate change scenarios and its implications over biodiversity, food security, etc.

• The output map was used on the REDD+

reference scenario of greenhouse emission calculation, achieving the transparence of the methodology applied and its replication as better data is available

• A methodology based on a set of 7 scripts

programmed on R (Ihaka, R. and R. Gentleman ,2013) were developed

• Decision tree algorithm (Therneau, T., B.

Atkinson, et al. ,2013) achieved regression

• utputs of 0.86, 0.79, and 0.90 kappa values

for the best models

Acknowledgments References

• Dinamica Ego (2009). Dinamica EGO. 1.6 ed. Minas Gerais - Brazil, Centro de Sensoriamento Remoto, Universidade Federal

de Minas Gerais.

• FAO (2010)."Global Forest Resources Assessment 2010.Progress towards sustainable forest management. Global

tables.".Retrieved 10/07/2013, 2013, from http://www.fao.org/forestry/fra/fra2010/en/.

• Ihaka, R. and R. Gentleman (2013). "The R Project for Statistical Computing." 3.0.1. from http://www.r-project.org/.
• IMAZON (2010). Imgtools - Monitoramento da Amazônia
• ITT Visual Information Solutions (2009). ENVI 4.6 ed.
• Leica Goesystems (2007). ERDAS IMAGINE 9.2 ed.
• MAE (2013). Representación Cartográfica de los Estratos de Bosque del Ecuador Continental. Subsecretaría de Patrimonio
• Natural. Quito - Ecuador, Ministerio del Ambiente del Ecuador (MAE).
• RAISG (2012). Amazonía Bajo Presión. A. Rolla, B. Ricardo, D. Larrea, J. Ulloa and N. Hernández. Sao Paulo - Brasil.
• Souza, C., D. Roberts, et al. (2005). Combining spectral and spatial information to map canopy damage from selective

logging and forest fires. Remote Sensing of Enviroment 98 (2005) - ELSEVIER, 15.

• Therneau, T., B. Atkinson, et al. (2013). "rpart: Recursive Partitioning." from http://cran.r-

project.org/web/packages/rpart/index.html.