LPISQA Legacy/ETS/Zones

From Wikicap - European Commission

How many images do we have to use?

The LPIS QA framework aimes at providing a quantitative, unbiased and precise state of the LPIS. In order to provide a cost-efficient sampling scheme for an 'educated' decision about the quality ISO 2859 standard has been used. Both elements require strict random sampling.

The default LPIS QA is based on imagery, therefore the sampling process is linked to it. In fact the samples follow the "clustered design", where only the first inspected parcel of the site is selected randomly and where all subsequent parcels are geographically related (i.e. there is a maximum distance from the first parcel). We assume that this geographical relation can be neglected as the same process and the same data sources apply to the whole LPIS territory. However, MS should test this assumption ex post. In order to minimise the potential risk where the assumption is false we recommend to use all suitable imagery available. Suitable means: recent, VHR, JRC-specifications compliant, and does not have LPIS-related risk)

Is there a difference in quality of orthoimagery used for LPIS creation and LPIS QA

The difference in quality can expressed in terms of product type (scale, resolution, viewing angle) timing, geometric, and radiometric quality.

Geometric quality: In contrast to aerial orthoimagery, JRC found that the quality of the CwRS VHR orthoimagery is very much dependent on the ortho-rectification process; i.e. the existence and the quality of ancillary data (GCPs, DEM). However, the producer does not always have direct control over these latter. In visual inspection, an increased relative geometric accuracy is required, which allows a better alignment between the vector and raster data. Consequently, more stringent requirements are proposed for the orthoimagery used in the LPIS QA. Some recommendations in that respect were already given on the Wiki article Use of Orthoimagery.

Radiometric quality: Focusing on geometric quality often too little attention is given to radiometry, which leads to poor colour balance and the loose of details in the image. Unfortunately, there are to date no clear and standardized metrics in respect to the quality check of the radiometry. To fill this void, JRC already revised its Guidelines for Best Practice and Quality Checking of Ortho Imagery and the Orthoimage technical specifications for the purpose of LPIS. Further work is being done on this topic and the results will be published as soon as they are available. For immediate guidance, we have compiled a dedicated article with thoughts and experiences on these sub-elements.

Timing: LPIS QA images are less dependent on the timing inside crop season. A longer acquisition window is possible than for CwRS. In any case, it is paramount that no deterioration of the LPIS quality assessment is allowed. The Commission services therefore accept the following options for LPIS QA, in case the CwRS imagery can be doubted:

  • Purchasing imagery only of a very high quality (uplift of CwRS zones)
  • MS can purchase better imagery if they see fit

Can we use VHR data with lower elevation angle than recommended?

The range for the elevation angle of VHR acquisition given in "Control zone and imagery selection " It is a recommendation, rather than a strict specification. On flat terrain with large, unfenced/unobstructed fields, indeed the elevation angle might cause no adverse effects and be fully suitable for LPIS QA. So, you can use such imagery if the usability of the product is justified.

NOTE: The prime objective of the initial acceptance of the orthoimage prior to the ETS (Annex II, Section 1) is to check the output of the ortho-production process, and not the quality of the raw/input image data.

Different datasets (orthoimages) used for the LPIS update and the ETS, Are the mapping results comparable?

The input for testing (ETS) should be independent from the data used for creation/updated of LPIS. They need to be appropriate in terms of positional accuracy (i.e. they need to be correctly orthorectified and need to be in the same reference system, comply with the recommendations). In case a MS doubts that the imagery obtained in CwRS fits the purpose it can acquire dedicated imagery according to its LPIS specifications for its LPIS QA zones, ensuring randomness and currency of the latter.

Oblique VHR imagery acquired over hilly-mountain areas

For the hilly regions, there is no option to exclude 'ex ante' non-nadir imagery because of the topography alone (which is indeed also affecting the RP creation). In this light JRC considers to be appropriate

  • use an objective topography parameter to decide on the minimum elevation angle for each and every zone
  • analyse the hilly and flat areas separately 'ex post' in your assessment report.

Should we alter the CwRS programming parameters to comply with the LPIS QA imagery recommendations?

In general: NO

The sample pre-selection generates a list of parcels with a population more than three times the prescribed sample size. If some of the recommendations e.g. capture angle, are not fulfilled by a particular image acquisition, you can choose "ex post" to skip the inspection of the parcels that fall in the unsatisfactory image. This is similar when the zone has no image coverage. According to the ETS rules when you decide to skip a zone you have to motivate this decision referring the failure to meet the recommendations.

NOTE 1: The historical success rate for CwRS image capture is more than 90%. Consequently ignoring some zones is not expected to jeopardize the availability of imagery for inspection

NOTE 2: These LPIS recommendations do not have an impact on controlling farmers' applications. Zones "skipped" for LPIS QA are still valid for CwRS.

NOTE 3: Contact JRC only when you fear that your current CwRS image programming will provide you less than threeim images that meet the recommendations.

From 2012 on, the CwRS ordering programme allows for a MS to identify 3 CwRS zones where an uplift to LPIS specifications is applied. This planning increases the likelihood that images within recommended specifications will be captured. Furthermore, any image in the “non-uplifted” zone has its probability of complying

What are the general rules for discarding/accepting control zones (or part of them) for LPIS QA

It is not possible to discard zones without a clearly demonstrated LPIS risk (before sample pre-selection) or failed acquisition (after sample pre-selection). The LPIS under all zones is made by one set of rules so inclusion of as many zones as possible is needed for better representativeness.

The criteria for discarding zones must objective, relevant, and systematic and it should not result in a bias or manipulation of the LPIS QA results. Excluding areas simply because they are more challenging is not valid as this would result in testing the given LPIS in optimal conditions only.

NOTE: We use the concept of “skipping” during the parcel inspection loop and “discarding” during LPISQA zone selection.

Time gap between the date for the sample and the date of the image

Until 2012, the relevant date of the sample used to be the date of pre-printed forms, which obviously does not equal with the date when the imagery used for the forms was captured. From 2013 on, it is the date at closure of the aid applications (i.e. the crosscheck).