Surface Volume Tool - large discrepancy in results when using TIN vs Raster.

[ChristopherEarley]

I have been trying to calculate the volume of a small surface area. However, I am somewhat concerned about the differences I have encountered when using a TIN vs raster surface as the surface input. With the TIN I got a volume of 40000m cubed, with the raster (.asc) I got 50000m cubed. I know there will be differences in the result due to how each surface is constructed, but 20%? I find that a little disconcerting.

I don't know if this makes much difference but the TIN is derived from the raster in question using the raster to TIN tool. I did set the z tolerance (in the raster to TIN tool) to 0.01 so I assumed the surfaces would more or less the same.

So, does anyone know why there is such a big difference in the results? And which result do you think is more reliable?

Thanks

[DanPatterson_Retired]

As it states in the help topic for the raster to TIN tool ... Converting a raster to a TIN will not, in and of itself, produce a better surface... so have no expectations that the properties calculated from them should be in agreement.  There is preparatory work that can be done and it appears that you mention one (z-tolerance) but not the others.  In fact the z tolerance that you specified is 0.01, is that 1/10 of the z range?  (from help ...  By default, the z tolerance is 1/10 of the z range of the input raster....) nor do specify any ancillary points that you added.  I would investigate these, but don't be surprised as to the outcome...and don't assume that the raster is the correct one in any case, since I presume that it was interpolated from point data.

[ChristopherEarley]

I didn't expect they would be in complete agreement, perhaps +- 5% difference. But a 20% discrepancy makes me think me there may be a problem with my method or the tool isn't very accurate. I used a 0.01 z tolerance as I thought a 1 in 10 height difference between the raster and the output TIN wasn't all that accurate. I also increased the maximum no. of points to 3500000 to allow for greater accuracy. The raster is derived from lidar data with 25cm accuracy.

[DanPatterson_Retired]

I may be reading this wrong, but I think that specifying 0.01 might be a tad optimistic.  I am assuming meters as the X,Y and Z values which means that you want 1 cm as the tolerance.  What is the range of elevations within the dataset? 100m (therefore the default is 1/10*100 = 1 m, etc etc).  I would start conservatively, then see if you have made matters worse by being over-optimistic.