Can't fit design in device

The next compatible device is the EPM570T100.

There is some room to gain some LE's. You are using 4 separate PWM's. THe compiler reduces this to 3 as the REAR and the NAV have the same dimming settings (for the moment), this explains the 'sudden' 47 difference. If you consolidate the PWMs into one tick counter and 4 (better 'n') comparison operators you can save up to (n - 1) * 16 LEs. In theory Quartus II could do that on its own ... I think if you re-write the PWM-module so that the PWM_ENA is used inside the 'if rising_edge( Clk ) - end if;' construct it will share the ticks counter. A quick try confirms this, but I didn't check the functionality.

Thanks a lot everyone! I'll try all your suggestions.

Replacing "<= "with "=" in if statements gave significant improvement.

Way off : Your intend to use clocked process but you wrote process(clk, other_signal) wich is not what you want (if we look inside the process). You may write

process(clk) -- no other signal in sensibility list
begin
   if rising_edge(clk)
      -- synchronous sequential statement here
   end if;
end process;