Introduction to WG5 on CwRS imagery use and alternatives and QE5 on claimed rate inside the RP Peter Viskum Jørgensen, FERV and Birger Faurholt Pedersen, University of Aarhus, DK
CwRS Imagery use and alternatives In Denmark DJF at Århus University is responsible for the CwRS and FERV is responsible for the LPIS updating and LPQA In 2010 8 zones were selected for CwRS (4 Ikonos, 4 GeoEye) It was decided to supply some of the VHR imagery with so-called Snapfotos Furthermore all reference parcels were screened for potential errors by DJF before June 1, 2010 and the reference parcels were updated by FERV using the 2010 imagery before the CwRS began Danish LPIS build on Physical blocks: approx 300.000 RP of 9 ha in avg. National ortophoto coverage from spring/summer 2010 were also available via Quick-access, when processed at the contractor from August 2010 and will fully be available late October
Wiki Cap requirements for LPIS-QA Art. 10.2.5 in the frozen version 19/9 2010 on LPIS-QA Quickbird, GeoEye-1 and Worldview-2. These sensors are suitable by default for all conditions and are labelled as "prime LPIS QA" sensors. IKONOS and Kompsat-2. These sensors can be used under certain conditions and are labelled as "secondary LPIS QA" sensors. At least with Ikonos alternative solutions might be useful!
Snapfotos Rectified aerial photos Acquired 7 April 2010 Covering 4 whole CwRS zones + part of two other zones or approx. 3000 km2 in total 40 cm resolution Both RGB and CIR Price ready to use approx. 70.000 EUR: 23 EUR/km2 Delivered 23 April 2010 to DJF and FERV No adjustment of colors and seam lines
SNAPFOTOS missing adjustments
Snap-foto overall conformance DQ Sub elements Conformance Quality Level & Tolerance Limits OK Spatial resolution <= 1m x Radiometric resolution =>8 bits/channel x Spectral Resolution Color (natural or color infrared) x General Image Quality Lack of defects and artifacts, which could prevent the visual interpretation of the image Cloud cover <5-10% x Overall clipping <0.5% at each tail x Histogram Peak +/-15% of middle value x Color balance <2% between min and max value of triplet (x) Noise Signal to Noise Ratio > 12 for each channel x Contrast The coefficient of variation of the image DN values should be in the range of 10-20% Geometric accuracy RMSEx <= 2.5 m; RMSEy <= 2.5 m x Mosaicking DN valies variation on similar area type not to exceed 10% in average (or 4% between each of the 3 channels) Mosaicking Geometric mismatches along seam lines (d) <3 pixels x Compression Lossless (TIF, LZW-TIF) Visually lossless (JPEG2000, ECW, MrSID) at last stage (storage) x (x) (x) x
SNAPFOTO 7. April 2010
IKONOS 15. April 2010
ORTO DDO2008
Geoeye 15 APRIL 2010
SNAPPHOTO 7 APRIL 2010
ORTHO DDO 2008
Conclusion on use of Snapfotos Better than Ikonos Similar to GeoEye The missing adjustments of colors, contrast and seamlines a minor problem But archive orthophotos are of much better quality
Other findings on this topic Limited information from other MS! For setting up and updating LPIS Orthophotos are used in most MS s. For the ETS Satellite images are used. These 2 sources have a different quality and will lead to different digitized area Image quality is not always adequate to perform the ETS
ETS inspection, example from Sweden ETS inspection, example 1 Satellite image 2010 inspection is obstructed due to oblique image (off nadir) and/or large tree canopys, shadows... LPIS,Copenhagen 2010 15
ETS inspection, example from Sweden Ortophoto 2008 inspection not obstructed LPIS,Copenhagen 2010 2010-09-23 16
What to discuss and clarify on the use of CwRS imagery and alternatives In some cases the inspected area is more incorrect than the recorded area. In some cases RP boundaries is delineated better from older images (1-3 years old) Solution could be to allow alignment in such cases to be done from older images Off-nadir image causes problems. Distortions noticed on several images Solution could be: Increase quality control of CwRS imagery regarding geo-referencing and ortho-rectification Sometimes it is necessary to use orthopotos form previous years, because trees cover up the border of RP, specially in case of image, which is acquired with low angle. These orthophotos form previous years allow to digitize RP border in the places where it is not well visible in satellite imagery. In some cases the 2010 satellite imagery fit well with other orthophotos, they are not shifted or rotated. Solution use older Orthophotos, when 2010 satellite imagery fails to deliver quality?
QE 5: Claimed rate inside the RP Values and expectations: At least 95 percent of the claimed parcels shall be completely claimed taking into account the last sentence of point 3.5.3 in the discussion paper on LPIS quality inspection EU requirements and methodology Problems and comments from MS The threshold is too strict Declared area needs to be clarified Is it declared area or determined area (area determined by administrative and on-the-spot controls)? The problem is that declared area can be bigger than recorded area of RP (this is overdeclaration/overclaim). If we use declared area, we can get result that ratio of declared area to the maximum eligible area >100% Is the purpose of this quality measure to find the amount of irregular declarations made by farmers (in this case we should use declared area) or the amount on land actually used in RP (in this case we should use determined area)
Problems with Declared vs. Determined area Examples on not applied, but eligible areas within reference parcels. How do we handle these in terms of QE5? Farmers might be inaccurate in delineation of the applications. What areas should be compared. The area of the polygon of the single parcels or the area that the farmer has applied in the alpha numeric data? How do we handle exceeding areas? delimitation of the reference parcel or establishing a layer of not applied areas within the reference parcel
Example from Sweden Problems with intrepretation Reference parcel is 3,11 ha and was controlled in OTSC In 2008
Example from Sweden What should be digitised? 2,14 of 3,11 ha (cereal), which Is comparable with the southern part 1. Grass or? 2. Perm. Grass betesmark? 3. Arable land! 1 2 3