1. High-accuracy Terrain Modelling for Soil
Mapping using ALOS-PRISM Imagery
S.J McNeill S.E Belliss D Pairman
Landcare Research New Zealand
2. Motivation
S-map is a national soil database, map and information
inference system for New Zealand providing:
A complete national digital soil map
Accessible data and inferred key information
Provides the best legacy data as well as new data
S-map methodology:
Gather legacy knowledge in modern data framework
Gather new data for areas poorly mapped at present
Use data mining methods to predict high country soil
properties
Present data in an accessible form for end-users
A good DEM is required for the high country
Low cost per unit area
Better accuracy than existing DEMs generated from
20-metre contours
Satisfactory for estimation of low-order derivatives
IGARSS-2011, 25-29 July 2011, Vancouver, Canada
3. Objectives
Investigate how better DEMs can be produced for
high-country soil mapping
Emphasis is on use of existing ERDAS product suite, where
possible
Consider DEM quality for generation of complex terrain
attributes
Important considerations:
ALOS-PRISM imagery favoured due to low cost per unit area
High cloud cover means some areas cannot be covered with
optical imagery
Methodology should use a variety of data sources where
advantagous
IGARSS-2011, 25-29 July 2011, Vancouver, Canada
4. ALOS-PRISM sensor
Forward, nadir, aft telescopes with constant 2.5 m resolution
Each telescope has 4×CCD line sensors camera, each with a
separate focal centre
Raw data rate from PRISM sensor subsystem 960 Mbit/s
Far exceeds available downlink data rate of 120 or 240Mbit/s
Lossy JPEG compression implemented on-board, with
constant output data rate
JPEG block artifacts reduced by processing change in Oct'07
IGARSS-2011, 25-29 July 2011, Vancouver, Canada
5. Study data
ALOS-PRISM imagery
01 Jan 2008 (3 scenes), 20 Jan 2009 (3 scenes)
Excellent sun elevation (59o & 56o ), no cloud
ALOS-PALSAR imagery
Intended to ll gap between PRISM DEMs using InSAR
Dual-pol. data used as basis for vegetation height model
Field data
Dierential GPS feature position estimates (σ = 0.25m)
Other supporting data
Raster DEM at 25m postings from 20m contours spot
heights
Geodetic marks for height validation
IGARSS-2011, 25-29 July 2011, Vancouver, Canada
9. Imagine LPS DEM generation
Extensive facilities for data management and editing
Adaptive, multi-resolution, correlation stereo method
Generic line-array model for ALOS-PRISM sensor
Range of output options limited:
ASCII le of estimated DEM points
TIN, 3D shape model, or interpolated TIN
Only qualitative error measure available
(19 × 19 km sun-shaded subscene) (8 × 8 km sun-shaded subscene)
IGARSS-2011, 25-29 July 2011, Vancouver, Canada
10. Imagine LPS DEM point output
5595000
5590000
5585000
5580000
5575000
5570000
q Excellent
q Good
q Fair
5565000
5560000
1820000 1830000 1840000 1850000 1860000
IGARSS-2011, 25-29 July 2011, Vancouver, Canada
11. Imagine LPS DEM accuracy
Measured against geodetic marks (order 15)
Quadratic trend in error minimum in scene centre
Due to simple single-CCD line sensor model (?)
Can be corrected using post-tting of DEM data
IGARSS-2011, 25-29 July 2011, Vancouver, Canada
12. Accuracy of corrected DEM
Measured against independent contour-derived DEM
Estimated σ = 6.48m for TIN-interpolated DEM
95% equal-tail condence interval [4.46, 11.8]m
Measured against independent geodetic marks (order 15)
Estimated σ = 2.90m for TIN-interpolated DEM
IGARSS-2011, 25-29 July 2011, Vancouver, Canada
13. Interpolation of point output
Abandon use of TIN-interpolated DEM output
Use interpolation of point output, with requirements:
Enforce smoothness
Incorporate other data, where available
Rational Basis Function (RBF) interpolation
N
s (x ) = p (x ) + ∑ λi Φ (x − xi )
i
The RBF is a weighted sum of a radially symmetric Φ at the
centres xi and a low degree polynomial p
Finding λi given xi , s (xi ) very dicult for large N
Specialised software provides tting, ltering and surface
generation using approximation methods
Method used can t successively ner-resolution terrain
surface models
IGARSS-2011, 25-29 July 2011, Vancouver, Canada
14. Processing method
Reduce high-frequency JPEG artifacts and de-stripe
Import PRISM data for ERDAS LPS
Use GCPs to t stereo model using ERDAS LPS
Generate point output using ERDAS LPS DEM generation
Apply cross-track model correction for DEM height
Build uncertainty model for corrected point output
Fit RBF surface from point output with specied point
uncertainty
Generate terrain surface by evaluating RBF
IGARSS-2011, 25-29 July 2011, Vancouver, Canada
17. RBF advantages disadvantages
Advantages:
Terrain surface inherently smooth
Fitted RBF can be t successively with additional terrain
information
Accuracy of RBF-interpolated DEM not degraded compared
to TIN-interpolated DEM
Disadvantages:
Fitting process needs to be managed carefully to preserve
memory
Some tuning of RBF tting parameters required
IGARSS-2011, 25-29 July 2011, Vancouver, Canada
18. Conclusions
We have developed a pragmatic rather than the technically
best solution for DEM generation
Need to provide pre- post-processing to make exiting
software work satisfactorily
Additional processing requires little extra manual eort
For high country, results provide a useful improvement over
existing DEMs
Acknowledgements
Research funded by the Ministry for Science
and Innovation (contract C09X0704).
IGARSS-2011, 25-29 July 2011, Vancouver, Canada