Methods to define confidence intervals for kriged values : Application on Precision

In Precision Viticulture, the increasing number of information sources calls for methods to combine them and to generate information for professionals. Each data is located on its own geographical support, and geostatistics are mostly used to produce estimates on a common grid. Estimated values have to be associated with confidence intervals; using kriging variance to compute them is generally considered to be efficient. This paper discusses this approach and presents two original methods to compute confidence intervals without using kriging variance. These methods are based on iterations of the kriging step, and takes into account inaccuracy of variogram estimation or measurement variance. They have been applied to yield data. Results have been compared to those obtained with kriging variance.