dr. david caballero (meteogrid) “ground truth survey in spain”
TRANSCRIPT
Ground truth survey in SpainD. Caballero
ArcFuel WorkshopThessaloniki 2013
Spanish study area
• Provinces of Córdoba and Málaga, Andalucía, Spain
• Córdoba is one of the largest provinces in Spain.
• Both hold forestlands in a rough terrain, shrublands and agricultural lands
Spanish study area
Survey points
• LUCAS points (P)• Alternative points (ALT)• Complementary points (PC)• Other review activity (Fotos)
Survey points
Method
• Geo-tagged pictures• Recordings and annotations for every
photo set• GPS antenna linked (BT) to tablet with
OruxMaps and Google Earth (cached)• Taken 8 pics (N, NE, E, SE, S, SW, W, NW),
Upslope, Dwnslope, Canopy, Ground• Extra pictures showing the environment
Navigation to survey points
Geo-One GPS receiver attached to a Nikon D7000 camera
Navigation to survey points
Difficulties• Private properties! Almost no access,
everything with fences!• Topography, very steep slopes• Vegetation: dense vegetation with thorns
(Ulex sp)• Hunting activty (specially at dusk), bullfighting
bulls• Winter time, short daylight hours• Look for alternative points, jump fences and a
lot of walking -> reduced performance
Survey - Spain
• Poor accessbility -> alternative points of the same structure
• A total of 56 points, 58 surveyed, one outside study area (Seville), one missing information
Survey - Spain
• 1 EG Broad Scrub 1• 2 EG Broad Open 5• 3 EG Broad Dense 4• 7 EG Conif Scrub 1• 15 EG Mix Dense 2• 19 Shrubs 30• 20 Grasses 5• 24 No fuel 4
ArcFuel Classification
• Classification of points according ArcFuel Map: depends on precision and resolution
• Complicated in highly fragmented fuels• Suggestion: adaptative resolution
(according fuel fragmentation)• Suggestion: use of LiDAR
Classification results
• Good for urban areas, it should be tested in dense intermix areas (i.e. North of Cordoba city)
• EG Broad Scrub vs. Open: no so clear the difference sometimes. Required data on height (LiDAR)
• In general good classifying dense forests
Classification results
• Missing EG Broad Dense for EG Conif Scrub (PC03A)
• Missing EG Conif Scrub for Shrub! (P4)• Missing Shrub for EG Broad Dense (PC05A)• Missing EG Conif Dense for Agrofor (P27)• Missing EG Mix Dense for EG Broad Dense
(PC38A, PC04A)
Classification results
• Good classifying shrubs (80%), but many different formations and structures included as shrubs (see comments) Missing Shrubs for Grasses, Shrubs for Agrofor -> abandoned lands PC10A, P35ALT, PC35A, P52ALT
• Missing Grasses for Shrubs, difficult to differenciate (shrub density) P45, P47ALT
Conclusions
• 34 points well classified (60%)• 3 points very badly classified, wrong or
inconsistent data sources• 8 points badly classified, difficult to
differentate heights, densities, species• 11 with reasonable classification,
particularly abandoned agro lands, outdated data sources (19%)
Conclusions
• Need for more information on shrubs• Design and apply a robust method to
classify shrubs in the understory• Use of temporal and thematically
consistent data sources• Use of LiDAR (where available) may help in
future classifications