A categorical, improper probability method for combining NDVI and LiDAR elevation information for potential cotton precision agricultural applications

An algorithm is presented to fuse the Normalized Difference Vegetation Index (NDVI) with Light Detection and Ranging (LiDAR) elevation data to produce a map potentially useful for site-specific management practices in cotton. A bi-variate Gaussian probability density distribution is modified to predict an improper probability distribution that also incorporates categorical variables associated with quadrant direction from the population means for the NDVI and elevation data layers. Water availability, influenced by slope and relative changes in elevation (as captured by the elevation data layer), affects crop phenology (as captured by the NDVI data layer). Thus, this fusion procedure results in a map potentially describing the joint effects of NDVI and elevation on cotton growth in a spatial and temporal way.

► We build a map for statistical analyses and site-specific decisions for cotton.
► We use both the NDVI and LiDAR elevation to obtain the geographical information.
► We combined data with incompatible units of measurement.
► A categorical, improper probability algorithm is described as a solution.