REDD+ FOR THE GUIANA SHIELD Technical Cooperation Project REDD+ for the Guiana Shield Mathieu Rahm, ONF international Impact of Gold mining training session, 24-28 November 2014 Cayenne French Guiana
Methodology Building MONITORING THE IMPACT OF GOLD MINING ON THE FOREST COVER AND FRESHWATER AT THE GUIANA SHIELD REGIONAL SCALE
Why and how establish a clear methodology/protocol? 1. Why? The results of data processed by several operators will have to be compiled need of comparability of results To ensure comparability of results: Operators need to do the exact same steps of process A clear decision tool need to be developed for ensuring similar responses/decisions by all operators when they encounter a specific situation A detailed methodology ensure the quality of data produced 2. How? Operators follow a protocol which guides them through the process. This will ensure that nothing has been omitted By defining the exact steps that the operator will follow By providing guidelines and precise rules to the operator for all type of situation
Choices to be made - Based on: - Results of the last study - Experience and knowledge of each forest services - Methods used by the different forest services - Available data in each country (roads, settlements, deforestation maps, )?? - Need of: - Update last results - Regional methodology - Comparability of results with the last study (2000-2007) - Use of open source software - Use of SPOT images (SPOT 5 10 m resolution) and additional data
Chronology of the methodology steps 1. Image ordering and acquisition 2. Quality check of all acquired images 3. Archive of data and production of raster catalogue 4. Image pre-processing 5. Quality check of pre-processed images 6. Image processing 7. Quality check of the processing results (internal national control) 8. Delivery of results 9. Compilation, validation and accuracy assessment of results (external regional control) 10. Final delivery with assessed uncertainty
Flowchart of the steps Image acquisition and ordering NOK Quality check OK Archive Image pre-processing NOK Quality check OK Archive Image processing NOK Quality check OK Archive Results delivery
Decisions to take: 1. Image ordering 1. What level of pre-processing? 1A; 2A; 2B; 2. Maximum cloud cover allowed for selection of image? < 20%; <30%; Is the presence of Haze an issue? 3. Priority given to the most recent image for SPOT selection 4. If quality of recent images is bad, do we go back in time or do we prefer to use other satellite data like Landsat? In order terms, do we give priority to the resolution of data or to the time of acquisition? 5. If we fill gaps with Landsat image, a clear methodology need to be established. Eg: Landsat image is clipped to fill the gap of one SPOT
Decisions to take: 2. Quality check of acquired images Criteria _ images & orthorectification Coordina Is the cloud cover mentioned te in the Zero metadata correct or is it underestimated, filterin QC All bands system in Missing Cloud g Passed? making the available image Data type not Pixel useful size WGS84_U for the image purpose image cover. of the study? Histo Pixel neede Overlapp Tile_id Acq_date (5) (16 bit) (5 m) TM_33S data lines (20%) Artefacts quality noise d ing area Is the radiometric correction correct? Is the pixel resolution correct? 3338125 14/08/2011 y y y y y y y y y y y 3338124 14/08/2011 Are all spectral y ybands ypresent? y y y y y y y y 3338123 14/08/2011 y y y y y y y y y? y y 3338123 12/08/2011 Is the georeferencing y y ywithin ythe limits y of y quality y mentioned y y by the image y y 3338123 11/08/2011 y y y y y y y y y y y provider?. 3338122 12/08/2011 y y y y y y y y y n N 3338121 12/08/2011 y y y y y y y y y n N 3338120 02/05/2011 y y y y n y y y y y n y N 3338025 14/08/2011 y y y y y y y y y y y 3338024 14/08/2011 y y y y y y y y y y y 3338023 14/08/2011 y y y y y y y y y y y 3338023 11/08/2011 y y y y y y y y y n y y y 3338022 20/07/2011 y y y y y y y y y y y 3338021 12/08/2011 y y y y y y y y y n N 1. The points of control need to be established, eg:
Decisions to take: 4. Image preprocessing Based on the level of pre-processed images ordered, specify the preprocessing steps: 1. Define the projection system 2. Georeferencing needed? 3. Coregistration with last images probably needed in any case because a potential shift could lead to digitalization and surfaces errors 4. Index calculation and/or filtering for later photo-interpretation? 5. Contrast, histogram adjustment? 6. If use of other data like Landsat, other pre-processing steps needed? Eg: Layer stack
Decisions to take: 5. Quality check of pre-processed images Based on the steps identified for pre-processing, the points of control need to be established, eg:
Decisions to take: 6. Image processing (1) 1. What is the digitalization methodology to use? Automatic, semi-automatic, manual? Given the cloudy areas and the specificity of this study (update of polygons), manual classification seems the best option 2. What features do we digitalize? Mining activities impacting forest areas and freshwater Roads and settlements 3. How do we digitalize these features? What is the minimum mapping unit for each feature? Which combination of bands and/or index do we use to digitalize each feature? Infrared false color, NDVI, NDWI, SARVI, etc. How to detect each feature based on the band combinations and/or index proposed? What color, shape
Decisions to take: 6. Image processing (2) How do we digitalize it? Based on which band combination/index and what are the boundaries? How do we deal with misclassified objects of the previous study? How do we deal with clouds and cloud shadows/ no data areas? How do we report it? Based on the metadata rate of cloud or do we need to precise the location which means to localize them? How do we deal with roads and / or settlements integrated in or near mining sites? What would be the definition of human settlements? Does it integrate all infrastructures built by human? How could we integrate existing data in each country Which type of ancillary data do we use and for what purpose?
Decisions to take: 6. Quality check of the processing results (internal national control) 1. What control do we want to make? If we produce random points that need to be verified, it should: Be verified by another operator Be defined how many points we create in each class? Be define if we generate a confusion matrix? 2. Completeness control of the processing. Was all images processed? Within each image, has the complete area been analyzed? 3. What controls are commonly realized in each forest services?
Calendar of activities As planned, the methodological choices should be established by the end of this week. October 2014 November 2014 December 2014 January 2015 February 2015 March 2015 April 2015 May 2015 June 2016 July 2017 Activities Responsability W1 W2 W3 W4 W5 W6 W7 W8 W9 W10 W11 W12 W13 W14 W15 W16 W17 W18 W19 W20 W21 W22 W23 W24 W25 W26 W27 W28 W29 W30 W31 W32 W33 W34 W35 W36 W37 W38 1. Work Preparation Team constitution: prepare a job description and facilitate selection in countries ONFG Images' identification and purchase, organization of a complete file ONFG Data preprocessing Core team Preparatory work to update and detail the methodology based on existing data in the countries Literature review and exchanges with countries ONFG Methodology preparation ONFG 2. Training on data preparation and discussion on the methodology Preparation of training and training ONFG 5 days mission of the team in French Guiana Core team X Final description of the updated methodology to be submitted to partners validation ONFG Expected validation by partners Partners X 3. Data Processing by the team Preparation of the training and training ONFG 10 days mission of the team in French Guiana Core team X Processing for French Guiana territory Core team Processing for Suriname Core team Processing for Guyana Core team Processing for Amapa Core team Delivery of intermediate processing results Core team X Delivery of final processing results Core team X Support mission to countries (1 mission in Suriname/Guyana and 1 mission in Amapa) ONFG/ONFI X X 4. Data validation Identification of existing ground-truthing points ONFI Development of a validation protocole ONFI Implementation of the validation protocole and uncertainty estimates ONFI 5. Data compilation, report and analysis Compilation of data ONFG Report writing ONFG/ONFI Preparation of data files ONFG Provision of draft report ONFG/ONFI X Expected comments Partners X Provision of final report ONFG/ONFI X 6. Coordination Coordination ONFI Technical support ONFG
Photo credit: ONF Guyane Let s discuss and find the best options