They give you two points on the datasheet: 580W with case temp of 25C, and 400W with a case temp of 80C. You can assume an approximate third point of 0W with a case temp of 150C, which is the maximum junction temperature. Throwing those points into Excel and getting a scatterplot shows those three points are colinear to R^2 of .98. So as a first-order approximation, you could say your maximum acceptable pulse watt-loss is 725C, minus 4.7 times the temperature above zero in Celsius.
This function can not be modified! It's a function of the internal structure of the device. Temperature reduction is the only way that value is going to improve, short of picking a different IGBT. If you have any sort of budget, or plan to buy new parts in the future, I might recommend changing to a different IGBT. That device isn't even listed on Infineon's site any more, and word here is that the entire BSM series is dead; we're having to update our UL listings that include that series. Sorry I don't have a source for that.
Now, you can mitigate the problem by reducing the actual watt-loss caused by the pulse. You can calculate your watt-loss if you know your current by looking at the output characteristics plot in the datasheet. Voltage times current is power. You'll note that Vce goes down for a given current if Vge increases. So by hitting the gate as hard as you can without exceeding its spec (20V), you may buy yourself a few amps. But not much.
As a general rule, if you're really operating on the raggedy edge like that, you need to reconsider what you're doing.