• Forestry related production in Finland, Norway and Spain
Blom Kartta Ltd. • Is one part of Blom group • About 30 employees • Offices in Helsinki and • Long tradition in forestry sector Joensuu – First infrared images 1977 • Main products: – First colour infrared images 1977 – Aerial photography – First digital orthophotos in forestry purposes 1995 – Laser scanning – First laser scanning experiments 1997 – Cartography – First digital aerial photographs 2005 – Forest inventory – First experiments in forest interpretation started in 2006 – Remote sensing based inventory system as commercial activity since 2007
Inventory process • • • • • • • •
Planning Laser scanning Aerial photography Measuring field reference data Modelling Segmentation Interpretation Create datasets to customer’s data system
By-products of inventory process • Orthorectified aerial images – If images are captured
• Updating the stand borders – We are not digitizing those borders by hands but these are updated using automatic segmentation which produce microstands – Forester can use microstands to update old stand borders
• Accurate terrain and canopy height model • It is possible to find new products and applications – Analysing soil – Use these two raster datasets to modify stand borders
Remote sensing based forest inventory + Estimation of forest characteristics is highly accurate and objective + It is cost-effective method in large areas + No sampling, it covers the whole inventory area + Processing is highly automatic + It is possible to estimate and calculate results for large forest areas very efficiently - We can not predict all variables - Fertility classes or define forest management proposals
- Nature conservation issues - Rare flora, fauna or habitats
- Expensive method in small areas
Risks which relates to this method 1. Bad weather conditions in aerial photographs and laser scanning 2. Restrictions from air traffic control 3. Short time period to capture aerial data 4. Errors in field plots locations or errors in field plot measurements 5. Quality of laser scanning data and aerial photographs 6. How the field plots represent the actual inventory area 7. To perform this inventory process by several data providers
Example data
Field plots in example inventory area
Measuring field reference data
Reference plots in inventory area
Modelling and calculation
Modelling and calculation
Different modelling methods • Single tree detection – It requires more expensive laser scanning data because of higher density, at least 3-5 pulse/m2 – This method has not tested in operative inventory process
• Regression based methods – We can not predict tree species specific results accurately – Still, it works quite well in total forest characteristics
• Non-parametric methods – We can estimate several dependent variables • Tree species and total forest characteristics
– It requires lots of field reference data • Minimum is 500 field plots
– k-NN, k-MSN
Modelling and calculation • Dependent variables of forest characteristics – – – – – –
Basal area weighted mean diameter Basal area weighted mean height Number of stems Basal area Volume Dominant height
• Also we can estimate forest characteristics in saplings – In Finland development class T2
• There is a chance to find new variables to describe forest
Calculation units • The basic calculation unit is grid-cell – Size could be eg. 16x16 meters – The general principle is that the size of grid-cell is equal to size of field reference plots – For every grid-cell we calculate all desired forest characteristics – Stand level results are generalized from the grid-cells which are inside the single stand – Grid-cells could be utilized directly in data system
Stand level generalization
Stand level results
Microstands • It is a new generation product • Microstands splits the forest inventory area as homogenous units as possible • In forested area, the size of microsegments is approximately 250 m2 – 1 hectare • Segmentation of microstands is an automatic procedure – No humans visual interpretation
• It could be restricted inside the certain area – Forest property or some other area
• It is based on laser scanning data
Microstands
Microstands
Example of utilizing microstands. Theme of total basal area.
Data transfer
Conclusions • Remote sensing based forest inventory will increase its popularity in forestry • This method is very powerful tool in right hands – It requires very accurate planning – Powerful and good implementation – Quality control for every processing stage
• The quality and accuracy of forest resources should become better • It provides new products and applications – Eg. microstands, accurate terrain and canopy height model
• Forester is still needed – It is not possible to automatize all working routines – Forest management proposals and counseling for forest owners – Checking of deciduous species in forest.
