JRC Experience with Area Frame Sampling

Abstract:

WAYNE FULLER LECTURE

The JRC started working on Area Frame Sampling around 1988 in the context of a pilot project for improving agricultural statistics with Remote Sensing.  Pilot surveys took place during some 5 years in various EU landscapes aiming at defining an appropriate methodology, evaluating the cost-benefit ratio and favoring the use of Area Frames in the national official statistics in Europe.

In 2001, the European Statistical Office launched an EU wide Land Use Land Cover Changes Area Frame Survey called LUCAS. This survey, still on-going today, was based on a grid point survey, aiming at providing crop area estimates at EU and national levels reaching relative accuracy respectively better than 1% and 5%.

The objective of the lecture will be to review the lessons learned on questions like:

 1) For segments approach

- Why did JRC choose squared segments versus the US physical limits standard?

- Does the segment size really matter?

- Was the landscape influential on the precision of crop area estimates?

- Did we find clear relation between variances and crop areas?

- Can we reach CV’s below 10% at reasonable costs?

 2) For points approach:

- What do we lose when replacing full segment survey with a subset of points?

- Optimizing the number of points per cluster.

- Comparing expected and observed variances in the LUCAS survey 2001.

- How far is systematic sampling from the efficiency offered by stratification?

- Gain of precision obtained in the Italian AGRIT project with point sampling.

 3) Merging Area Frame with Remote Sensing imagery:

- Some drawbacks of square segments.

- The importance to average the segments (full use of 0-1 range).

- Systematic decrease of all crop areas after regression: an artifact?

- Was the use of Remote Sensing cost efficient?

- Merging point frame and RS: the calibration estimator and its variance.