Remote sensing-based spatial modelling forNatura 2000 management at large scale:
from habitat monitoring to site selection
JM Álvarez-Martí[email protected]
J.M. Álvarez-Martínez, S. García, B. Jiménez-Alfaro, A. Silió-Calzada, M. Recio, J. Barquín, B. Ondiviela, I. Pérez Silos, and J. Juanes
A need for spatial data: patterns, process, dynamics and functioning
of natural and seminatural systems (N2000)
20122010
SIOSE: Sistema de Información sobre Ocupación del Suelo de España (CNIG)
CLC (CORINE): CoORdinationof INformation of the Environment (EEA)
Land use-land cover typologies
Vectorial format: user-defined
Not homogeneous land cover patches: mapping habitats?
Limited or no temporal resolution: monitoring?
Traditional inventory: visual interpretation and digitalization
LAND COVER MAPPING
§ High cost (time, people)
§ 5 to 7 years (pilot area)
§ Difficult to update (no temporal resolution)
Traditional inventory
§ Lack of spatial data
§ Need for trainedsurveiers
§ Mixed patches
LAND COVER MAPPING
A cost-effective solution for
large scale mapping based
on optimal field surveys
(adaptive sampling), remote
sensing and habitat and
species modelling
LAND COVER MAPPING
Remote sensing-based spatial modelling
CASE STUDY: MONITORING
Management plan
Mapping broad-scale vegetation patterns in complex mountainous territoriesHabitat maps using modelling techniques in SCIàSACof Natura 2000 Network in Cantabria (NW Spain)26% of Cantabria. 25 hábitats…
3. Local actions,
cost, monitoring
1. Spatial
distribution
2. Conservation
Status
Annex I
Atlantic biogeographical
region (NW Spain)
Vegetation DB:
National maps
Regional programs
Fieldwork
Predictor layers
Limiting factors
Remote sensing
Monitoring and reporting
Conservation status
CASE STUDY: SITE SELECTION
Álvarez-Martínez et al, 2017
3. SAC definition
1. Spatial
distribution
2. Conservation
Status
Annex I, II
SWOT ANALYSIS
Cost-effective solution for large scale mapping
We have to get information in a quick, effective, homogeneous and dynamic manner
Álvarez-Martínez et al, 2017
We do not end up with available tools (year 2017 and so on) and outputs…
1] CLASSIFICATION TYPOLOGYLand use-land cover (LULC)
Vegetation types
2] OCCURRENCE DATATraining
Validation
3] PREDICTOR LAYERSEnvironmental limiting factorsRemote sensing: resolution
4] MODELLING PROCEDUREEnsemble, sensitivity analyses
Data mining tools…
HABITAT MAPPING
New modelling tools and remote sensing for vegetation mapping:
what actually matters?
1] CLASSIFICATION TYPOLOGYLand use-land cover (LULC)
Vegetation types
2] OCCURRENCE DATATraining
Validation
3] PREDICTOR LAYERSEnvironmental limiting factorsRemote sensing: resolution
4] MODELLING PROCEDUREEnsemble, sensitivity analyses
Data mining tools…
HABITAT MAPPING
Expert-based methods: a review of a estandard database
EUNIS (level 1): Heathland, scrub and tundraEUNIS (level 2): Temperate shrub heathlandEUNIS (level 3): Dry heathsEUNIS (level 4): Sub-Atlantic [Calluna] - [Genista] heaths F4.22
HABITATS DIRECTIVE ANNEX I: 4030-EU dry heathlands
EUNISF4.22
1] CLASSIFICATION SYSTEM
Patron List of Spanish Habitat types Download from MITECOResolución de 17 de febrero de 2017, de la Secretaría de Estado de Medio Ambiente, por la que se establecen tres listas patrón: la de las especies terrestres, la de las especies marinas y la de los hábitats terrestres, presentes en España - Texto de la Resolución
Borja Jiménez-Alfaro (U. de Oviedo)
EUNIS 3-5 level habitat types
Borja Jiménez-Alfaro (U. de Oviedo)
1] CLASSIFICATION SYSTEM
EUNIS typologies in Cantabria
EUNIS 2-6 level habitat types
EUNIS 2-6 level habitat types
Borja Jiménez-Alfaro (U. de Oviedo)
1] CLASSIFICATION SYSTEM
EUNIS typologies
in Central Anatolia
EUNIS 4 (6)
Patron List of Spanish Habitat typesDownload from MITECO
1] CLASSIFICATION TYPOLOGYLand use-land cover (LULC)
Vegetation types
2] OCCURRENCE DATATraining
Validation
3] PREDICTOR LAYERSEnvironmental limiting factors
Remote sensing: satellite and LiDAR
4] MODELLING PROCEDUREEnsemble, sensitivity analyses
Data mining tools…
HABITAT MAPPING
Vectorial LULC databases simplify complex landscapes mosaics by creating “homogeneous”
Landscapes patches inclufding different communities and environmental gradients…
There is a need for defining detailed (EUNIS, MAES…) typologies for large territories in order to
accomplish landscape complexity, temporal variability and locally-tailored management practices
2] OCURRENCE DATA
Occurrence data obtained from field surveys with (almost) no uncertainty:
Field campaigns (botanists)
Point-based sampling surveys
“Tablet System” H2020 Working Group of Turkey
Remotely connected to a central DB that maximizes data security and accuracy
Species caract., indicators of structure and function and dynamic update
2] OCURRENCE DATA
2] OCURRENCE DATA
İç Anadolu Pilot Bölgesi
117 449 pointsHabitats Directive
(Annex I)
224 770 points
2] OCURRENCE DATA
1] CLASSIFICATION TYPOLOGYLand use-land cover (LULC)
Vegetation types
2] OCCURRENCE DATATraining
Validation
3] PREDICTOR LAYERSEnvironmental limiting factors
Remote sensing: satellite and LiDAR
4] MODELLING PROCEDUREEnsemble, sensitivity analyses
Data mining tools…
What actually matters?
Remote Sensing (RS)
Satellite imagery:
Landsat 5TM and 8OLI 30m
Sentinel 2 A and B, 10-20m
DEIMOS-2, 4m
LiDAR derived data, 5-30m
Env. Limiting factors
topography, climate, soil
(digital soil mapping *)
3] PREDICTOR LAYERS
REGIONAL CASE STUDIES
2015-2018, MVC
LARGE-SCALE PROJECTS
35 Sentinel2A sub-scenes (20m). Sensitivity analyses with
Landsat data (30m). Final modelling: combined products (120m)
1] CLASSIFICATION TYPOLOGYLand use-land cover (LULC)
Vegetation types
2] OCCURRENCE DATATraining
Validation
3] PREDICTOR LAYERSEnvironmental limiting factors
Remote sensing: satellite and LiDAR
4] MODELLING PROCEDUREEnsemble, sensitivity analyses
Data mining tools…
What actually matters?
A DATA MINING method or modelling algorithm for habitat mapping relates
occurrence data and the process-based environmental and RS predictors
MaxEnt: SWD format, Tunning parameters, Phillips et al (2006)SDM: Multiple algorithms, Bootstraping, Naimi and Araújo (2016)
1
2 SPATIAL MODELLING
OCCURRENCE DATA
PREDICTORS
SPATIAL PREDICTIONS
MAPS
3
4] MODELLING
4030 –European dry
heathlands
0 1E 1:50 000 Local AOO
4] MODELLING RESULTS
9120 – Atlantic
acidophilous beech
forests with Ilex and
sometimes also Taxus
in the shrublayer
0 1E 1:25 000 Local AOO
4] MODELLING RESULTS
E 1:50 000
Automatic and objective: depends on the models
n habitats with good quality data
Realized AOO
4] MODELLING RESULTS
Teselado de la
vegetación en unidades
fisionómicas (manchas
homogéneas mayores
de 5hectáreas)
Automatic and objective: depends on the models
E 1:25 000 UNCERTAINTYDOMINANCE +
4] MODELLING RESULTS
VALIDATION - Confusion matrices
4] MODELLING RESULTS
Spectral uncertainty and unmixing
Landsat 8 OLI (30 m)
VALIDATION - Confusion matrices
4] MODELLING RESULTS
Landsat 8 MVC Landsat8 x2Sentinel2 x2Deimos2 x2+LiDAR +MDT
High suitability
Low suitability
4] MODELLING RESULTS
Landsat 8 MVC Landsat8 x2Sentinel2 x2Deimos2 x2+LiDAR +MDT
High suitability
Low suitability
4] MODELLING RESULTS
Landsat 8 MVC Landsat8 x2Sentinel2 x2Deimos2 x2+LiDAR +MDT
High suitability
Low suitability
4] MODELLING RESULTS
SPECTRAL SIGNATURES
Hyperspectralmeausrements
SPECTRAL SIGNATURES
Hábitat 4030 (b)
Vera
noOt
oño
C
D
E
F
Hábitat 6510 (a)
B
A
Hábitat 4020Spectral library: HABITAT TYPES
Spectral library: PHENOLOGY
SPECTRAL SIGNATURES
Hábitat 9120(F. sylvatica)
Hábitat 9230(Q. pirenaica)
Verano OtoñoVerano OtoñoB C DA
Whole TURKEY(example)
Traditional mapping system
Modelling
Economic cost
Time
Number of fıeld-
workers (2 years)
Resolution
Accuracy of mapping
products
% of habitats mapped
Monitoring
capabilities
(1) Could improve with
photo-interpretation
refinement of model
outcomes
(2) This % could easily
improve with further
research and data
(1)
7.000.000 € 2.500.000 €
5 years 2 years
486 162
< 1:50.000 < 1:50.000
80-90% 70-80%
70% 70%
Low, sampling Real-time
Good
Medium
Bad
(2)
HABITAT MAPPING: SWOT
APPLICATIONS
Monitoring the conservation status
Sites selection at the national level
Area of Occupancy (AOO)Estructural and functionalindicators
Early warning system: identification of drivers and pressures
Common cost-effectiveindicators of ConservationStatus through remote sensing
Non dependent of MemberState data (validation!!!)
PASTURES
CONSERVATION STATUS
4] MODELLING RESULTS
FORESTS
0 50 meters high Vegetation structure (LiDAR derived data)
LiDAR PNOA: 0.5 p/m2, <0.5m=NoData
CONSERVATION STATUS
1985 1990 1995 2000 2005 2010 2015 2020 … 2030 2040 2050
FOREST FIRES
VEG. RECOVERY
CONSERVATION STATUS
2030
2040
2050
Future scenarios
TIME
NATURA 2000 SELECTION
Systematic Conservation Planning principles+
Requirements of the Habitats Directive
METHODOLOGY FOR THE SELECTION OF NATURA 2000 SITES
Europe Aid project – Turkey
What – how much - where
44 habitats (EUNIS) à concurrence map (dominance+uncertainty)
WHAT TO CONSERVE?
High resolution maps: 22 Habitats (Annex I) in Central Anatolia
HOW MUCH TO CONSERVE?
Target definition by following objective criteria
WHERE TO CONSERVE?
Social and environmental costs
OPPORTUNITIES
Connectivity / Vegetation Sensitivity Index / Net primary productivity
Pollination / Evapotranspiration / Soil erosion protection
Integration of ecological processes
CONSTRAINTS
Integration of climate change
Area Lost àCentral Anatolian
gypsum steppes (E12B)
Future RCP 6.0 2060-2080
Actual
1
Actual distribution
Future (Potential)
0,5
DECISON SUPPORT SOFTWARE
Allows managing lots of spatial data in a same process to:
§ Identify the smaller number of planning units (that
means the optimal solution),
§ That are necessary to meet at the same time targets
for all the biodiversity,
§ Maximazing connectivity and coherence of the system
§ With the minimal cost
MARXAN
MARXAN PROPOSAL
Data-driven proposal of N2000 sites
Expert judgement usign a
Delphy methodology, asisted
by GIS information
Group AHABITAT THREATS
EXPERT KNOWLEDGE
Expert-driven refinement of N2000 sites
Connectivity at a regional
and interregional scale
Replication and
geographic
distribution
Closer to other
protection figuresSize (better bigger
than smaller)
Socioeconomic features
without spatial data for all
CAPA (present and future)
Compactness
(better round than
irregular-long areas)
EXPERT KNOWLEDGE
Natura 2000 potential sites in Central Anatolia
MANAGEMENT AND REPORTING
We need to get an approach that allows:
§ Information flow for updated data
§ Query system for adaptative management
A management system for N2000:
§ Automatic reports: sexenial reports (art 17)
§ Prioritised Action Frameworks (PAF)
MANAGEMENT AND REPORTING
Standard Data Form
ConservationStatus
Managementplans
Reportsproduction
Assesmentof effects
• Art. 17• PAF• Non Effects
DRIADE
Close
ZEC 00003
Fecha de actualización
Conservation status
Documento Legislativo de designación
HabitatCode
Documento Legislativo de designación
Fecha de actualización
Distribution area Structure and Function Global EvaluationFuture perspectives
Area: Ha
Mapa: Place
Region
Turkey
UE
Word
Quality of the data
Last update:
Tendency:
Observaciones:
Save
Impacto negativoR Código Presión/Amenaza OA A04.03 Abandono de los sistemas de pastoreoA A08 Uso de fertilizantes M D01.01 Sendas, pistas, carriles para bicicletas M
B02.04Eliminación de árboles muertos o
deterioradosB G091.02.03 Conducción motorizada todoterreno RB G05.01 Pisoteo, uso excesivoA
J03.01
Disminución o pérdida de las
características específicas de un
hábitat B K01.01 Erosión M
K02.01Cambios en la composición de especies
(sucesión)
R
MI01
Especies invasoras y especies
alóctonasM M02.03 Declive o extinción de especies
Impacto positivo R Código Actividad/Gestión OA A04.02.02 Pastoreo de ovejas no intensivo RM A04.02.03 Pastoreo de caballos no intensivo RA
A06.01.02Cultivos anuales no intensivos
para
producción de alimento
R
MB02.05
Sacas no intensivas (dejando
madera muerta/ árboles viejos
intactos )
R
F I M D
Área de distribución
Estructura y
funciónPerspectivas
futuras
Evaluación Global
F: FavorableI: InadecuadaM: MalaD: Desconocida
I: Intensidad; O: Observaciones
Obs:
Dry heathlands4210
Bibliography:- Sanz-Azkue I. & I. Olariaga. Uribe-Echebarría P. M., I. Zorrakin, J. A. Campos & A. Domínguez.
pq
pq
Code TR610002
Driade: N2K management system
Management plan