During the last weeks, I am trying to develop an application within my current working project, which aims to simulate smoke dispersion of fire incidents.

**Software used**: ArcGIS for Desktop/Server (version 10.2).

**Description**: My inputs are netCDF files containing data with concentration of airborne species in μg/m3. Available dimensions are time and altitude. Usually the values range from 0 to 500 including decimal digits. I am trying to figure out which method is the most appropriate to visualize point data to surface of this phenomenon. For example let’s suppose we want a raster which illustrates the smoke dispersion at a specific timestamp, for a specific altitude. I have tried several interpolation methods (IDW, Spline, Natural Neighbor and Kriging). I’d rather prefer to have an output of a smooth, real interpolation rather than an “eye-pleasing result”. Although for start, both ways are acceptable for me. 

**The problem** is that, after using Kriging, some values of the resulting surface can be higher than the maximum of the input values (and lower than the minimum of the input values). Also the predicted values of intermediate points are not estimated properly. Is the Kriging interpolation the most appropriate method? If that’s the case, then should I have to parameterize semivariogram parameters such as nugget, partial sill, major range, in order to keep unchanged observed values as also unstretched predicted values after kriging interpolation? Can anyone help me define these parameters for the following sample data?
If anyone has another method on visualizing smoke dispersion I would like to know.

**Sample data**: I uploaded a working case in arcgis format (mxd + gdb + netcdf files included) in the following link: [http://www.filedropper.com/casestudyfiles][1]

**General Info & Parameters Used**:
Coordinate System: WGS’ 84
Grid 250 x 260 (columns x rows)
Grid point distances (in decimal degrees):
Upper h: 0.0453645
Lower h: 0.047912
Upper x: 0.029341
Lower x: 0.0332265
Upper y: 0.0345845
Lower y: 0.034532
where: 
x: horizontal distance
y: vertical distance
h: diagonal distance

Sample of concentration values descending: 
[2.3348, 1.3012, 1.0288, 0.6168, …, 0.0084, 0, 0, 0, 0, …, 0]
Kriging Parameters used:
Kriging Method Ordinary
Semivariogram model: Exponential
Output Cell Size: 3.58615417480469E-02
Search Radius Variable
Number of points: 8
Max Distance: 0.100

  [1]: http://www.filedropper.com/casestudyfiles