2. SLOPE WP 2 – Task 2.1
Kick-off Meeting
8-9/jan/2014
Andrea Masini, PhD
Remote sensing and multispectral analysis
Remote Sensing Department
Flyby S.r.l.
3. Task 2.1: general description
Kick-off Meeting
8-9/jan/2014
• design of an automatic chain that provides a first level forest
inventory exploiting satellite imagery
• calculation of NDVI (Normalised Difference Vegetation
Indices) to monitor tree growth and biomass production also
in mountainous environment
• first level forest inventory used also to drive more accurate
UAV/in-situ measurements
• satellite-based data fusion with other data to achieve more
accurate results
5. Task 2.1: expected output
Kick-off Meeting
8-9/jan/2014
• Deliverable D2.01 (month 8 – August 2014) :
Report on remote sensing data collected, on the
methodologies and the algorithm to extract needed
information and on the generated output
6. Use of satellite data for forestry
Kick-off Meeting
8-9/jan/2014
Satellite imagery can be extremely
useful in the forestry sector in
particular for :
• forest health near real-time
monitoring
• accurate and wide forest
inventory
9. Kick-off Meeting
8-9/jan/2014
EO data used in the past for
• Cover Change Detection
• Mapping biophysical structure
• Mapping ecosystem services (carbon, water)
• Modelling trends under change scenarios
• Generating management plans
20. Other high resolution satellite data
Kick-off Meeting
8-9/jan/2014
We will investigate the following data:
21. Kick-off Meeting
8-9/jan/2014
Task 2.1 main objectives
• design of an automatic chain that provides a first level
forest inventory exploiting satellite imagery
• calculation of NDVI (Normalised Difference Vegetation
Indices) to monitor tree growth and biomass production also
in mountainous environment
• first level forest inventory used also to drive more accurate
UAV/in-situ measurements
• satellite-based data fusion with other data to achieve more
accurate results
28. WP 2.3: On Field Digital Survey Systems
1. Forest Mapper System (SatForm 3D, Remote
Sensing, Aerial LIDAR & Imagery)
2. Terrestrial Laser Scanning Forest Measurement
System (AutoStem Forest)
3. Real Time Forest Intelligence
36. Full Integration – ‘Closed Loop Control’
Multisource data
Tree modeling
Parameters relationsHarvest control
Forest pre-stratification
Initial area
Spatial generalization
Geostatistics
Area correction
Spatial analysis for
field plots locations
TLS recording
Field survey
Forest Mapping FIELD INVENTORY
Automated Processing
FINAL
STRATIFICATION
WEB SERVICES
DATA ANALYTICS
37. Some ExampleTrials: International
Validation & Facts
SkogForsk Sweden 2009/2013
Coillte Results
2008 UCC Stats Department
2010/2011 Industrial Trial Results
Scotland 2008/2013
Forest Research, Forest Enterprise Scotland
James Jones
Other Results
Greenwood Resources Oregon
SkoglandScap Norway
Forestry South Australia
US Journal of Forestry
Island Timberlands Canada
39. Manual control measurements of all logs
-Diameter and length
-Approximately one diameter per meter
-Average from two diameter
measurements per sampling point
At Remningstorp 34 trees were
measured by the operator using a caliper
Control trees
Example: Swedish Government Validation
40. Harvester production data
- Stem length and diameter measurements were used as reference
- Sample trees were harvested and harvester data collected
- Diameter measurements registered every 10 cm of stem
- Diameter from approx. 0.8 m height to last cut in tree
Strömsjöliden
Remningstorp
41. - GIS software onboard harvester for linking tree measurements from
harvester with TLS
- Manually registering made by the operator at the sample plots
Linking harvester measurements with TLS data
42. meterH
100
200
300
400
500
Height
0 1000 2000 3000
0
100
200
300
400
500
Height
0 1000 2000 3000
Control trees at Remningstorp, stand 343
Spruce 343-1-06Pine 343-3-12
Diameter
Harvester
Control
TLS
43. Site Species
Number of
trees
Mean trees
size1 (m3) Bias Std Dev RMSE
Remningstorp Pine 94 1.12 0.00 0.13 0.13
1.3% 11.6% 11.7%
Spruce 185 1.01 -0.01 0.11 0.11
-0.9% 9.2% 9.2%
Birch 16 0.44 -0.01 0.10 0.10
-8.0% 15.6% 17.5%
Strömsjöliden Pine 275 0.47 0.02 0.04 0.05
3.0% 8.8% 9.3%
Spruce 339 0.27 0.01 0.04 0.04
2.6% 9.4% 9.7%
Birch 29 0.21 0.01 0.03 0.03
2.6% 12.9% 13.2%
Volume estimates on individual trees
1. Volume: on bark, excluding top
Sweden Final Results: January 2013
50. Kick-off Meeting
8-9/jan/2014
Objectives
Task 2.4 Goal: To generate and make accessible a detailed
interactive 3D model of the forest environment.
The WP’s purpose is to develop methodologies and tools to
fully describe terrain and stand characteristics, in order to
evaluate the accessibility for and efficiency of harvesting
technologies in mountain forests.
51. Kick-off Meeting
8-9/jan/2014
Scheduling
Start Month: 7
End Month: 15
Deliverable: Harvest simulation tool based on 3D forest model
Total MM: 20
Task leader: GRAPHITECH;
Participants: CNR, KESLA, COAST, BOKU, GRE, FLY, TRE
52. Kick-off Meeting
8-9/jan/2014
Participants role
GRAPHITECH(10): Task Leader. It has in charge the development of tool for representing
the virtual 3D environment of the mountain forest as well as the of the virtual system
on mobile and machine-mounted displays. Finally it will be involved into the
developmet of the solution for interactive cableway positioning.
CNR(1): Definition of the “technology layers” (i.e. harvest parameters) and
methodologies to coordinate tree marking with the subsequent harvesting operations.
KESLA(1): Acting as final user in order to simulate the behaivor of own machine into the
virtual system
COAST(2): Provide the input model for the virtual system combining the information of
task 2.1, 2.2 and 2.3
53. Kick-off Meeting
8-9/jan/2014
Participants role
BOKU(2): it will be involved into definition of the “technology layers” (i.e. harvest
parameters) then on the developmet of the solution for interactive cableway
positioning.
GRE(1): Acting as final user in order to simulate the behaivor of own machine into the
virtual system
FLY(1): Provide the input model for the virtual system combining the information of task
2.1, 2.2 and 2.3
TRE(2): Development of the Forest WarehouseTM for mountain forestry and support
the deployment of the virtual system on the machine-mounted display
54. Kick-off Meeting
8-9/jan/2014
Platform Core
Using the remote data (Satellite, UAVs orthophotos and digital surface model) combined
with on field information (TLS), each single tree feature will be segmented including its
deducted geometric properties.
Task 2.1
Task 2.2
Task 2.3
3D forest model
Virtual 3D
environment
56. Kick-off Meeting
8-9/jan/2014
Functions
• Forestry measurements estimations;
The platform will allow the combination of accurate tree profile information
with up to date remote sensing data.
• Interactive system for cableway positioning simulation.
• Definition of the “technology layers” (i.e. harvest parameters);
Technological layers show technical limitations of machines and
equipment on different forest areas.
• Deployment of the virtual system on mobile and machine-mounted displays.
57. Kick-off Meeting
8-9/jan/2014
Functions
• Forest WarehouseTM (Treemetrics) for mountain forestry integration;
The Forest Warehouse is a web-based forest planning system that performs
bucking (log making) simulation through software developed by TRE.
58. Kick-off Meeting
8-9/jan/2014
3DVisualizationTechnologies
Interfaces Scenario
• Desktop
• Mobile
• In-vehicle embedded
Systems
Approaches
• Desktop Visualization Platform
with Mobile Porting
• Web-Client Visualization
Platform
Desktop Platform
• Open-Source Library for 3d
visualization (OpenInventor, Vtk,
Openscenegraph)
• 3d Engine ( UdK, Irrichlicht
Engine, Unity 3d)
Technologies
Web Client
• WebGL : implementation of
OpenGL ES 2.0 for web,
programmable in JavaScript
• Java Applet based on Opensource
Globe Nasa World wind
60. Kick-off Meeting
8-9/jan/2014
Thank you for your attention
DR. FEDERICO PRANDI
Federico.prandi@graphitech.it
Fondazione Graphitech
Via Alla Cascata 56C
38123 Trento (ITALY)
Phone: +39 0461.283394
Fax: +39 0461.283398
62. Index
62
1. Task objectives
2. Approaches for sites location and flow allocation decisions
3. Approaches to estimate traffic in existing roads
4. Proposed work plan
5. Contact info
63. 1.Task objectives
63
Task objectives:
Identify and analyze logistics elements within the forest and their characteristics
for site locations and flow allocation decisions
Integration of the data with the global forest model
Build and validate and Optimization model to allocate landings with the mills and
plants
Build a model to estimate traffic on individual sections for road maintenance and
construction purposes
To be developed from M8 to M13
Includes development of “D2.05 Road and logistic simulation module”
Due to Month 13.
Partners involved: all
ITENE (leader), GRAPHITECH, CNR, BOKU, FLY
64. 2. Approaches for sites location and flow
allocation decisions
64
The goal is to determine an optimal (minimum cost) forest logistic
network to respond future demands
The approach should determine:
Location of facilities (specially for new requirements)
Size an capacity of facilities (storage areas and processing sites)
Volume to harvest in every landing and stand area
Volume of timber to transport from landings to facilities (it gives a
first estimation of road traffic for road planning)
Volume of product to transport from facilities to demand sites
The model should consider inputs like location of landing áreas,
intermediate sites (storage, buffers), processing sites, demand sites,
demand volumes, routes, type of routes and distances between theses
sites.
65. 2. Approaches for sites location and flow
allocation decisions
65
Location of a single facility by center-of-
gravity method
Output: XY coordinates for the facility
Optimization based only on distances
Binary model (source-sink)
Useful for a first estimation of a facility location
to be supplied from specific lands
66. 2. Approaches for sites location and flow
allocation decisions
66
Location of selected number of facilities
by the exact center-of-gravity method
Output: XY coordinates of a selected number
of facilities
Optimization based only on distances
Binary model (source-sink)
Useful for a first estimation of 2 or more
facility locations to be supplied from specific
lands
67. 2. Approaches for sites location and flow
allocation decisions
67
P-median multiple facility location
Output: selected facilities from a list of
candidate sites receiving flows from other sites
Optimization based on transport costs and fix
costs, but lack of capacity constrains and other
inventory costs
Binary model (source-sink)
Useful for a first estimation of 2 or more facility
locations to be supplied from specific lands
68. 2. Approaches for sites location and flow
allocation decisions
68
Mixed integer linear programming
problem
Output: selected facilities and optimal flows
between nodes
Optimization based on transport costs and fix
costs, capacity constrains and inventory costs
Three stages model
More appropriate approach for a network with
more than 2 node types
lands in
forest
storage
and
facilities
(saw, mills,
biomass)
Demand
sites
69. 2. Approaches for sites location and flow
allocation decisions
69
Dynamic linear programming
Consider changing demand
Output:
Selected facilities
Size an capacity of facilities (storage and processing sites)
Volume of harvest in every landing and stand área
Volume to transport:
Timber from landings to facilities
Product from facilities to demand sites
Decision to expand production capacity in a specific
period in the planning horizon
Minimize total costs for timber supply and
transport, investment and operational costs,
product transport cost to demand sites, fixed
cost for capacity expansion
-
200
400
600
800
1.000
1.200
1 2 3 4 5 6 7
Period Demand Volume
lands in
forest
storage
and
facilities
(saw,
mills,
biomass)
Demand
sites
(normally
cities)
70. 2. Approaches for sites location and flow
allocation decisions
70
Previous Work
Facilities Location Models: An Application for the Forest Production
and Logistics
JUAN TRONCOSO T. 1, RODRIGO GARRIDO H. 2, XIMENA IBACACHE J. 3
July 2002
1 Departamento de Ciencias Forestales, Pontificia Universidad Católica de Chile, Casilla 305,
Correo 22, Santiago, Chile. E-mail: jtroncot@puc.cl
2 Departamento de Ingeniería de Transporte, Pontificia Universidad Católica de Chile.
3 Escuela de Ingeniería Forestal, Universidad Mayor.
71. 2. Approaches for sites location and flow
allocation decisions
71
Stand
Cable
ways
forest
lanes
72. 2. Approaches for sites location and flow
allocation decisions
72
minor
road
main
road
land
land
land
stand
stand
stand
73. 2. Approaches for sites location and flow
allocation decisions
73
Solution flow
Possible flow
lands in forest storage and
facilities (saw,
mills, biomass)
Demand sites
(normally cities)
74. 2. Approaches for sites location and flow
allocation decisions
74
INPUTS
Demands of product per each period and type of quality from demand site
DATA COLLECTION FOR THE MODEL
Positions of stands, lands, storage areas, processing sites (saw, paper mills and
biomass heating and power plants), demand sites
Volume available to harvest in every stand per quality of timber and destination (saw,
mill or energy)
Position for stand respect existing roads
Slope or grade of difficulty to access
Capacity of ground to support specific machinery
Size and availability of skyline deployment sites
Capacity and location of storage areas and buffers, and processing sites
Characteristics of processing sites and conversion facilities
Distances between different nodes
75. 2. Approaches for sites location and flow
allocation decisions
75
COST FACTORS
supply and transport operational costs
final product transport cost to demand sites
fixed cost for capacity expansion during the planning horizon
investment associated to construction of a new site
OUTPUT
Selected facilities
Size an capacity of facilities (storage and processing sites)
Volume of harvest in every landing and stand área
Volume to transport
Timber from landings to facilities
Product from facilities to demand sites
Decision to expand production capacity in a specific period in the planning horizon
76. 3. Approaches to estimate traffic in existing roads
76
Once the different sites and locations have been selected, and flows between
sites have been determined for each future period,
A Logistics Resource Planning Model will be used to determine the volume to
harvest in every period in every land, processing and transport means, and a
more precise estimation of traffic in every individual sections of road in terms of
number of trip per vehicles type (size, weight) in each period
This traffic estimation will allow to define plans for road maintenance and
construction in the forest area, taking into account the capability of roads to
accept trucks and cranes of different weights and sizes
77. 3. Approaches to estimate traffic in existing roads
77
Similarities to DRP method
Land 1
SITE: Saw Plant
X
City 1
Product demandHarvest orders
Land 2 City 2
78. 3. Approaches to estimate traffic in existing roads
78
SITE: Saw Plant X
Minumum Batch (harvest) (m3/period) 500
Lead time (number of periods) 1
Safety stock (m3) 200
Period 1 2 3 4 5 6 7
Demand Volume (m3) 400 500 600 1.000 500 600 1.000
Available Stock (m3) 700 300 300 200 200 200 100 100
Harvest recepcion (m3) - 500 500 1.000 500 500 1.000
Harvest order launch (m3) 500 500 1.000 500 500 1.000
Land 1
To harvest (m3) 500 500 1.000
Available m3 in land 1 2.000 1.500 1.000 -
Size of vehicle (m3) 10
Number of vehicle trips size 10m3 50 50 100
land 2
To harvest (m3) - - - 500 500 1.000 -
Available m3 in land 1 3.000 2.500 2.000 1.000 1.000
Size of vehicle (m3) 10
Number of vehicle trips size 10m3 50 50 100 -
79. 3. Proposed work plan
79
Understand the forestry supply chain and logistic processes. Choose a real scenario (ITENE, BOKU)
Review literature and formulate an Optimization model for logistics site location and flow allocation
decisions (BOKU)
Define a model to estimate traffic in existing roads (CNR)
Identify elements for the models:
Relevant logistics locations within the forest (GRAPHITECH, CNR, FLY, ITENE)
Gather info and contact with the different agents of the forest product processing (ITENE)
Define and analyze relevant characteristics of the logistics elements (ITENE)
Integration with the global forest model (ITENE)
Implement the Optimization model to allocate landings with the mills and plants and traffic calculation
on individual sections (BOKU)
Validation of model with a real scenario (BOKU)
Implement the model for road planning based on the amount of timber to be transported and
identification of traffic on existing forest infrastructure (CNR)