mapping forest degradation in madagascar · 04.02.2016 · • simple gis training (qgis,gps...
TRANSCRIPT
Mapping forest degradation in Madagascar
Igor Glushkov, Dmitry Aksenov, Olli Turunen, Titta Lassila, Angela Tarimy,
Daulphin Razafipahatelo
FANC ‐ Finnish Association for Nature Conservation
• The Finnish Association for Nature Conservation (FANC) ‐ Finland
• Transparent World ‐ Russia• National experts (Conservation international, University of Antananarivo)
• Local specialists and stakeholders (Association Mitsinjo, ValBio, Durrell, Manombo Special Reserve)
Team
Collaborators, sponsors• Ministry of Foreign Affairs of Finland• Ministry of Environment of Madagascar • Madagascar National Parks• Helsinki University• Finnair airline company• SEAS‐OI, Scanex • WCS, WWF, Duke Lemur Center, Missuri Botanical Garden
• Institute for the Conservation of Tropical Enviroment
Past and current forests mappingNational level Global level
2007Madagascar Vegetation
mapping Project,MODIS and
Landsat based
2007KEW Forest Change 1990 ‐ 2000 – 2005,
Landsat based
2008(updated 2013)Intact Forests Landscape,
Landsat based
2000‐2013UMD/GFW Global Forests Cover and Change maps,Landsat based
General goals of mapping project• Create network of ”forest mappers” from local to
international level • Create mapping and communication capacity in local level• Support Malagasy scientific terminology• Create tools for people to communicate offline/online with
thematic maps • Create clearing house mechanism for forest monitoring
(transparent, low cost, easy available)• Create series of high‐scale forests intactness / degradation
maps
Why to measure forest intactness / degradation?
• The forest biodiversity is sensitive to human influence and depends on the level of the forest transformation
• The intactness of forest is important for decision‐making on conservation priorities
• Forest transformation/degradation map is required for better understanding the transformation reasons and making decisions for more sustainable landuse from local to international level
BIG trees Neverloged
Deadwood
Signs of intactness
Natural disturbances
Slash‐n‐burn (TAVY)
Fires
Agriculture/Rice cultivation
Alien species/Plantation
Selective logging
Grazing
Mining
Old tavi
Local gold mining
• Evaluating classification criteria and adapte to local level.
• Developing of 3‐days seminars for local groups, field trips, data collecting
• Simple GIS training (QGIS,GPS ‐using)
How to measure the level of forest intactness / degradation?
from Ground
How to measure the level of forest intactness / degradation?
Classification scheme for Manombo region
7 forest classes from intact to disturbed translated to local names of forest stands types
About 20 – years cycle of forest regenerations may be reduced to 1‐2 years due to disturbance circumstances
Data to be collected in the field• GPS coordinates• Subjective proposal for the forest
class• Dominant tree species• Diameter of the biggest trees• Age and height of the upper layer
trees• Amount of dead wood (range)• Altitude (GPS)• Exposition of slope (west / east)• Young / old forest (rough
devision)• Special (tree) species• Photos
Field surveys of various forests in eastern
Madagascar by Malagasy colleagues from local
associations and by joint Malagasy‐FANC‐TW teams
From 2011‐2014 field trips,8 regions, more then 3000 GPS points, about 2000 from forest stands
How to measure the level of forest intactness / degradation?from Space
Direct measurment:Hight, Biomass.
Indirect method:Fragmentation analysis
• Global ‐ ICESat/GLAS, RadarSAT etc.Simard,Pinto et al. 2011
• Regional – airborne LiDAR dataAsner, Maskaro et al. 2011
• Global ‐ Intact Forest Landscapes maps
Potapov, Yaroshenko et al. 2008
Or combining both above with classification of space images time series
• Regional ‐ Mapping deforestation and forest degradation
Margono, Turubanova et al. 2014
• Intact / old‐growth humid forests usually have more than one layer of closed canopy (Multi‐layer – ML)
• Degraded / secondary forests humid have a single layer of closed canopy (Single‐layer – SL)
• Intact forests may have a simple canopy structure (single layer) only in specific habitats or landscape positions (SL "on top of ridges")
How to measure the level of forest intactness / degradation?
Canopy structure approach
Simple forest canopy structure classificationA. Multi‐layer (ML)
B. Single layer with big tress (SL BT)
C. Single layer (SL)
A B C
How to measure the level of forest intactness / degradation?
From ground to spaceMulti‐layer (ML)
Single layer with big tress (SLBT)
Single layer (SL)
No canopy structure, have spectral characteristics visible on images or on time series set of images
How to measure the level of forest intactness / degradation?
Canopy structure approach
• Using SPOT‐5 (also SPOT‐6/7 in the future) panchromatic channel (detailed (2.5 m ground resolution) enough to detect single big trees crowns and canopy irregularities)
• Mapping tree crown shadows marking irregularities in the canopy (a kind of “reversed” single trees crown mapping method)
• Calculating density of shadows and classifying forest stands polygons on this base
How to measure the level of forest intactness / degradation?
Calculating local minimum reflections pixels density on SPOTpanchromatic channel
SPOT 5 ‐10м
SRTM2
SPOT 5 ‐2.5м
STEP 1
Segmentation of SPOT images by spectral channels, 10m resolution (GRASS GIS module i.segment with the following parameters: min area – 10 pixels, similarity threshold = 0.03 for spectral channels).
Adding main statistical parameters from spectral image channels and DEM data to each polygon: mean, standard deviation, minimum, maximum, range to each polygon; as well as canopy complexity data.
STEP 2Creating training dataset, using high‐resolution image and field observation data as reference.
STEP 3
Creating model using the advanced decision trees algorithm ‐ R:randomForest.model of classification, data statistics for each class, estimate model quality and uncertainties.
Applying model to whole set of data and export result map (QGIS).
Employing the Random Forests algorithm for classifying forest stands by all their parameters measured by DEM and SPOT
images
Data sources
Space images Base images more then 70 scenesSPOT 5 from 2009‐2014 (SEAS‐OI)More then 900 sub‐scenes WV1,2 from 2011‐2013 (SCANEX)
For the map‐based portal: Landsat 8http://maps.transparentworld.ru/madagaskar/
The classification schemeCanopy coverage
High 4.1a. Savoka &Savoka with single
trees4.1b. Pure bamboo
thickets ‐ ??4.1c. Pure ravenala
thickets ‐ ??4.1d. Filippia thickets
‐ ??
3.1a. SL closed –mountain forests3.1b. SL closed –lowland forests3.1c. SL closed –littoral forests
3.1d. – Eucalyptus plantations3.1e. – Pine plantations
‐‐(not found yet) 1.1. ML closed
Medium
4.2. Mosaic of ramarasana, crops, savoka & woodlands
3.2a. SL sparse – high altitude forests
3.2b. SL sparse –slopes & valleys
2.1a. SL with big trees – top of ridges
forests2.1b. SL with big trees – lowland
forests2.1c. SL with big trees
– littoral forests
1.2. ML with gaps
Low
1.3. ML with large gaps
Very low 4.3a. Bare ground4.3b. Ramarasana4.3c. Roranga
‐‐ ‐‐ ‐‐
Non‐forest Single layer SL with big trees Multilayer
Canopy structure complexity
dark green – most intact ML;bright green‐mixed intact ML and SL; yellow – degraded SL
Map‐based portal
Thank you!Misaotra!Spasibo!
Igor [email protected]
Dmitry [email protected]
Angela [email protected]
Olli [email protected]
Daulphin [email protected]