there is no "exact method". You pick a narrow band of frequencies you are interested (for instance near the cuttoff frequency of a filter), and pick a distributed element to simulate by a distributed one.
so as an example, you have a series L, shunt C, lowpass filter with a cuttoff frequency of 2 GHz.
Lets say ONE of the lumped elements in the lowpass prototype model is 2 nH.
View attachment 192216
You would simulate the above inductor of 2 nH, and see its impedance across its terminals, maybe its magnitued of S21 from 1 to 3 GHz, concentrating the most right at 2 GHz.
Then you would replace it in your simulator with a high impedance microstrip element.
View attachment 192217
Then you try to match the above to the response of the actual lumped inductor you previously simulated.
If you find something that kind of matches, you then add on a length of LOW impedance line to simulate the follow on lumped capacitor in the filter.
View attachment 192218
then keep going until the entire lowpass filter is made up of transmission line elements.
if you want to get fancy, you can add models for microstrip line width changes, that will model the parasitic shunt capacitances and series parasitic inductances. In this specific case (lowpass filter) those parasitic capacitances actually help you make the distributed sections look more lumped in nature.
View attachment 192219
the fringing capacitance adds to the shunt capacitance of the low impedance section
THIS will get you pretty close to the actual response.
at this point you either build the actual distributed element filter and measure it to optimize, OR do a full electromagnetic simulation of the structure to give it a final tweak. You might tweak this lowpass filter to have the exact cuttoff frequency you want. Or you might not like the return loss ripple, and want to improve that.