Digital Soil Mapping in JRC Endre Dobos

1
Digital Soil Mapping in JRCEndre Dobos
2
The most recent activities in digital soil mapping

• European SOTER
• 1 1 M soil database
• 1 250.000 scale Georeferenced Soil Database
• Danube database development (Beata Huskova 3 pm.)
• Profile data extrapolation study (OC)
• Formation of a DSM core-group to discuss the
  needs and tools for soil data derivation and
  update

3
Formation of a DSM core-group to discuss the
needs and tools for soil data derivation and
update
Participants Daruossin, Joel Dobos, Endre
King, Dominique Mayr, Thomas
Montanarella, Luca Vrcaj, Borut
15-19 March, 2004 Ispra
4
Identified objectives

• Developing a quantitative procedure adaptable for
  different scales
• Completing the European SOTER at the scale of
  11million or 15 million
• for the 1250.000 scale soil database for Europe
• Danube database profile data inter-,
  extrapolation (appr. 1 250.000)
• Upgrade the 11 million soil map?
• Harmonization of the procedures for the three
  addressed scale.

5
Potential products to create

• A polygon database of the soil mapping units
  (objects soil body association)
• The addressed scale 1.250.000 to 1 5 million
• A soil property database
• Raster based thematic layers of basic soil
  properties
• profile-based information soil types
• horizon related information
• start and end depth, horizon designation based on
  the diagnostic horizons and properties (from
  WRB), SOM, texture ( of sand, silt, clay), pH
• 3. Accuracy assessement/Quality measures layers
  of data confidency
• (Pedotransfer functions can be used for deriving
  further information not contained in the base
  dataset)
• Metadatabase
• explanation-interpretation for the users
• procedure description used for developing the
  layers

6
European SOTER
7
Assessing land degradation processes
8
(No Transcript)
9
(No Transcript)
10
Deliverables

• Procedure and delineation of DEM on a scale of 1
  1M
• Cooking book
• European physiographic unit coverage.
• SOTER database for Europe (as a continent) at the
  scale of 1 5M
• Pilot study of South America
• Revise the methodological problems, rewrite the
  manual
• Publish the version one of the EUSOTER
• Raising the interest for the countries with no
  SOTER coverage to follow the new procedure and
  complete their part of the database.

11
Objectives

• To derive a quantitative, DEM based procedure to
  delineate SOTER physiographic units at the scale
  of 11 million, following and translating the
  original criteria defined in the SOTER Manual of
  Procedure.

12
Material

• GTOPO30 and SRTM Digital Elevation model
• Spatial resolutions
• 1 km
• 90 m

13
Pilot areas
Europe Carpathian-basin
14
Pilot areas
Carpathian-basin
15
(No Transcript)
16
SOTER

• Mapping Units are defined by physiography and
  lithology

• Physisography is charateized by four
  differentiating features
• Slope
• Relief intensity
• Hypsometry (the combination of relief intensity
  and altitude)
• Dissection

17
SRTM DEM - based Procedure
Reclassified layers (90 m)
Block majority (block size 990 m, grid
resolution 90 m)
Resampling to the block size of 990 m
Focal majority with 4 and 6 cells radius circles
Focal majority with 3 cells radius circles
Elimination of the polygons under the size
threshold using the minimum Euclidean distance
procedure and line simplification
Final terrain unit polygon system
18
Regional slope classification
Original SOTER Quantitative procedure
Flat 0-2 0-2
Gently undulating 2-5 2-5
Undulating 5-8 5-8
Rolling 8-15 8-15
Moderately steep 15-30 15-30
Steep 30-60 30-60
19
Classified slope
Resampled, classified slope
20
Relief Intensity
median difference between the highest and lowest
point within the terrain per specified distance.
Units are m/km, m/slope unit, m/2 km Changes of
the approach interpret relief intensity on an
aerial basis.
0-50 m/area of a 1 km diameter circle
50-100 m/area of a 1 km diameter circle
100-300 m/area of a 1 km diameter circle
300- m/area of a 1 km diameter circle
21
(No Transcript)
22
(No Transcript)
23
Hypsometry
Level and sloping lands with RIlt50m/slope unit Sloping lands with RIgt50 m/slope unit Steep and sloping lands with RIgt600m/2 km
lt 300 m lt 200 m 600 -1500 m
300 - 600 m 200 - 400 m 1500 - 3000 m
600 -1500 m 400 lt m 3000 - 5000 m
1500 - 3000 m
3000 lt .m
24
Hypsometry class Elevation range (meters above sea level)
1 Up to 10
2 gt 10 - 50
3 gt 50 - 100
4 gt 100 - 200
5 gt 200 - 300
6 gt 300 - 600
7 gt 600 - 1500
8 gt 1500 - 3000
9 gt 3000 - 5000
10 Above 5000
25
(No Transcript)
26
Dissection

• 0-10, 10-25, and over 25 km/km2
• Potential drainage density (PDD)
• 0-7 and 8-49
• I. Deriving surface drainage system based on
  digital elevation data.
• 1. flowdirection,
• 2. flowaccumulation,
• 3. recoding
• cells having flowaccumulation value higher than a
  certain - resolution dependent - value are
  assigned a value of 1 (stream cells) while all
  others are assigned value 0 (background).
• II. A specified sized moving window is sent
  through the image for counting the stream cells
  within a given area around each cell and the
  output value of this function is assigned to the
  corresponding cell.

27
(No Transcript)
28
The PDD class ranges
Class PDD value range
1 less dissected areas, convex surfaces 0-90
2 more dissected areas or depressions, concave surfaces Above 90
29
(No Transcript)
30
(No Transcript)
31
(No Transcript)
32
Aggregation procedure

• Selecting the polygons under the minimum size
  limit
• Minimum Euclidean distance
• Calculating the four mean terrain variables for
  each polygons
• Calculating the Euclidean distance for each
  polygon pairs
• Dissolving the bordering arc between the polygons
  having the smallest Euclidean distance

33
Generalization procedure eliminating the small
polygons
34
Line simplification procedure

• 0.2 mm separability distance between features on
  the printout.
• Displacement of the vertices with maximum 200
  and 1000 m in ground units respectively for the
  11 and 15 million scales

35
(No Transcript)
36
(No Transcript)
37
(No Transcript)
38
Filling in the database
39
The HunSOTER Database.
The raw AVHRR image.
The DAFE transformed image.
40
EUROPEAN SOIL BUREAU
EUROPEAN SOIL DATABASE 11 MILLION
DAFE TRANSFORMED AVHRR-DEM IMAGE
41
Classification result (average likelihood
probability 65.4) European Soil database
42
(No Transcript)
43
(No Transcript)
44
SOTER physiographic polygons derived from GTOPO30
45
1 1 M soil database
46
Limitations

• Success story
• MARS-CGMS
• Soil vulnerability
• Organic carbon content
• Soil erosion assessment

/5
47
SMU1232 STUs spatially not locatedSemantic
database STU026 (50) STU376 (30) STU030
(20)
Regionalization of soil data, Borut Vrcaj

Soil mapping unit 1986
Soil mapping unit 2003


Disaggregated raster STU map
48
Combined STU map of the area
Regionalization of soil data, Borut Vrcaj

STU


SMU

• The methodology was developed and tested using
  different data
• 11M Soil map of Europe (ESB/JRC)
• 125.000 map of Slovenia (CSES)
• 110.000 map of SPIN Postojna Test Area (CSES)

49
1 250.000 Scale Georeferenced Soil Database
50
Pilot areas

• Requirements
• Size ranging between a 100-150 km by 100-150 km
• Should represent a great variety of natural
  (soil, terrain and parent material from the
  manual of procedures)
• Should represent a great variability of data
  availabilty
• English
• Size km2 each
• French
• Slovenian
• Hungarian

51
Potential procedure to create the test area
databases

• Data collection and preprocessing
• Delineation of the soil-landscape units
  (following the 1250.000 georeferenced soil data
  base for Europe) based on the DEM (later to be
  refined by lithology, etc.)
• Extrapolating the profile information to create
  the raster based thematic layers
• Assigning the soil property and terrain
  information to the polygons
• Attaching a complementary, ancillary dataset to
  improve model/QA/describe other environmental
  parameters (climate, landuse)?

52
Danube database development
53
Danube database developmentthe creation of
raster based, soil property layers

• Procedure
• Stratification (based on a degraded datasets)
• geomorphology,
• Unsupervised fuzzy K-mean classification
  (Precision farming website of Australia)
• Paying attention to the landscape position of the
  profile (MacMillan)
• Derived from the DEM (option No. 2)
• Backward SOTER delineation procedure, using a
  modified SOTER procedure having the Hypsometry,
  relief intensity slope and PDD as factors.
• Lithology
• based on the parent material classes of the 1 1
  million European Soil Database
• revised soil region map of Europe, BGR (option
  No. 1)
• Further subdivision of the physiographic polygons
  could be done using expert knowledge (in case if
  there is no lithological information
• Analysis of the profile data
• Data distribution and representativity
• Modelling procedure for spatialization
• (regression, linear modelling)

BRAIN STORMING EXERECISRE
54
Soil organic matter estimation pilot study
55
(No Transcript)
56
The methodology used to derive the soil organic
matter map

• Data processing
• Calculation of the total soil organic content on
  a t/ha basis from the layer based data of SOM ,
  soil density and layer depth variables ofthe TIM
• Logarithmic and square root data transformation
  for the variables having non-normal-like
  distribution .
• Estimation of the SOM values for the unknown
  areas with regression-kriging.

57
The data used to derive the soil organic matter
map

• Remote sensing data
• Two dates (May and Sept of year 2000) of MODIS
  using the 7 reflective bands and the NDVI (bands
  1-7, resolution 500 m) and one thermal infrared
  band (band 31, original resolution 1 km, but
  resampled to 500 m)
• Digital Elevation Data
• SRTM-30, 1 km resolution, resampled to 500 m
• Derived parameters
• Elevation
• Specific catchment area
• Profile curvature, planar curvature, complex
  curvature
• Relief intensity
• Slope
• Potential drainage density (PDD)
• Flow accumulation
• Aspect
• Geographic position representing the climatic
  changes
• Easting
• Northing
• Soil Monitoring System for Hungary (TIM)
• Almost 1300 georeferenced soil profile data
• Evenly distributed all over Hungary, representing
  all potential landscape positions, landuses and
  parent materials.

58
Statistical Procedure, Regression kriging
Methods

• Linear Regression
• Square root transformed NDVI from Sept. 2000
  Sqrsndvi
• MODIS band 5, May 2000 May95hu
• Square root transformed PDD Sqrtpdd
• PDD Pddnd4hu05
• Logaritmic transformed altitude Lndem
• Relief Intensity Ridemndhu05
• MODIS NDVI, May 2000 Mayndvi
• MODIS band 3, May 2000 May93hu
• Profile convexity Prcurv100hu05
• Square root transformed MODIS band 1, May 2000
  Sqrm1
• MODIS band 4, May 2000 May94hu
• Aspect Aspecthu05
• SOM_g/m2 ( Sqrsndvi -1829.276) - (May95hu
  1.469) - (Sqrtpdd 7626.015) (Pddnd4hu05
  1240.775 ) - (Lndem 12097.897) (
  Ridemndhu05 54.247 ) ( Mayndvi 105.134
  ) - ( May93hu 33.099 ) - (Prcurv100hu05
  9179.56 ) ( Sqrm1 2556.134) - ( May94hu
  3.998 ) - (Aspecthu05 10.738 ) 87921.241
• R2 0.238, significant
• Kriging of the regression errors

59
(No Transcript)
60
Summary and perspective

• Spatial Soil Information System

61
Spatial Soil Inference System
Expert knowledge Digital Mapping tools
/5
62
PROPOSALfor new working group on DIGITAL SOIL
MAPPING
63
Ground data plotted in relation to modelled OC
values for three areas in UK
Cambridgeshire - Norfolk border, East Anglia
Monmouthshire South Wales
Otterburn Northumberland N England
lt1 1-2 2-6 6-12.5 12.5-25 25-35 gt35
lt1 1-2 2-6 6-12.5 12.5-25 25-35 gt35
lt1 1-2 2-6 6-12.5 12.5-25 25-35 gt35