# What is principle of inverse modeling?

I read a few paper related to emission inventory talking about the inverse modeling. It was another method to estimate the emission from ground(called top-down method). The satellite sensor usually support the product with vertical column density(VCD) for certrain species.

The target here seems to be generate the relationship between emission and VCD. In my opinion, there are some uncertainty hidden here:

(1) the gridded VCD of certain species can be formed by different precursors, transported from neighborhood area.

(2) The chemistry system in the model also has some inherent bias which would miss some crucial source or sink process.

(3) When we set source inventory as input for chemistry transport model, the height information and the specific location are two crucial parameters. In the area of inverse modeling, we know nothing about the information of real sources(i.e fire spot, power plant).

How to generate a relative precise relationship between satellite data and source emission in consideration of these doubts?

• Generally speaking, if you have a model, you are given an initial condition and you find a solution in time. In inverse modeling, you are given a result, and you find the initial condition from whence it started. It is effectively modeling with a negative timestep. You are right on the first point. For the second point, if you do a top down approach, (I am uncertain of this), you aren't particularly interested in locations. That is more for bottom-up approaches. Sep 28 '16 at 19:12
• @BarocliniCplusplus that is a good start at an answer. I think the ability to use inverse modeling for location (or source) specific questions would depend on the pollutant in question.
– f.thorpe
Sep 29 '16 at 3:47
• To (3): It could be an approach to prescribe major point sources (their location, not the magnitude of emissions). To (1): We need species that are observable by satellites (or derivable from satellite data) and their reaction pathways are not too complicated (In the best case, they are also short living substances.). We should not do it for each species. NO2 is a candidate (as @jvir stated). To (2): yes, a problem. We have this problem also when we model forward in time ;-) . Dec 6 '16 at 17:37