I am going to make some assumptions. The fact that there are ground forces means that they are fighting other ground forces. This means that the enemy will avoid using orbital bombardment methods with extremely high yields. The higher the yield, the more of an early warning you need. This means small impactors.
I am also going to assume that these projectiles are mostly unpowered and unshielded. A clever enemy could make a very deadly orbital strike weapon, resistant to standard detection techniques and the ability to course correct. However, you have to ask yourself if it would be cost effective. The price of a single advanced warhead could pay to fling dozens if not more small "dumb" rocks to the surface.
I do think there is one method that may work, but first I want to dismiss some of the methods that could possibly work.
A series of cameras on your vehicles pointed skyward would be able to spot the heat of an incoming orbital attack hitting the atmosphere. Decent enough computer systems could calculate the impact point.
The period of time it will be visible can very quite a lot based on the speed and angle of entry. However, on Earth like planets, the atmosphere will be about 60 miles thick. At speeds of around Mach 1, it would take only about 6 minutes if directly incident (what is going to happen to its speed will depend a lot on the aerodynamics, ect, though it will likely slow down). Likely our projectiles will initially be going much faster. Geostationary orbit is close to Mach 4 for example. The Russian meteor a few years back was going close to mach 100.
So a warning time of about a minute might be possible in some situations. That doesn't really seem like enough time to me.
Outside of the Earth's atmosphere they will likely be to dim to detect with small mobile optics.
One way that we track NEO (Near Earth Objects) is Near IR. Things in space will absorb heat from the sun. They then will reradiate that heat.
You may wish to read up on WISE and NEOCAM. WISE is an orbital satellite with a 40cm diameter lens (about 1 and a third feet). However, there is no reason you couldn't have similar detection on the surface of the Earth. You will need to kick up the sensitivity a lot, but adaptive optics have gotten a lot better.
Based on its field of view, if something similar was mounted on a surface based vehicle it would be able to image the entire hemisphere it was in with about 2,000 images (If I did my math right, 47 arcminute FoV). At 11 seconds an image that is 6 hours to image a hemisphere . Way TO SLOW. You are really likely to miss a threat.
However, if a group of soldiers had 3-4 vehicles, with an array of like 3 of these on top and they were all working in unison? 30 minutes to image a hemisphere.
This detection technique can be used to track things as far out to the keiper belt (they are to cold past there). Not atmospheric interference would greatly reduce the range.
The problem becomes one of resolution. An object about 15m in diameter (the Russian impactor) would need to only be about 2km away to be seen (far too late). In practice it can only practically detect things 150m or larger in diameter.
NEOCam is a new satellite that has been designed that is only slightly larger and uses much more modern technology (though same basic idea) It has a much better resolution, designed to detect things down to 30m. Hypothetically, it could spot a 15m impactor about at about a third the distance to the moon based on its resolution. NEOCam also has a much wider field of view (almost ten times), but I believe takes pictures much slower.
We can assume the future will see continued development on this front.
There is a lot of story potential with this type of detection. They require cooling (WISE requires Significant cooling). A detachment unable to resupply would risk running out of cooling. The ability of a detachment's ability to detect would also decrease if they lost vehicles, degrading their knowledge over time. Without communication to forces in the other hemisphere, they would also be blind to half of space. An impactor could be hiding just around the corner.
During the day impactors would be more visible (they would be heated more), but there would be a blind spot where the sun is.
You can check out This link if you want to do the radar calculations yourself. However, with a power source of only 100kw as stipulated in your question, the system would be able to detect a small 15m impactor at about 50km away. You would honestly need a terrible amount of power to detect an orbital impactor at a useful distance with the type of radar that would be mobile. So radar is out.
This one is a bit more out there, if you don't want people calling you on getting any of the engineering wrong. Here you can read about using magnetic fields to detect asteroid collisions. Here you can read about using lasers to measure magnetic fields. enter link description here.
It may be theoretically possible to have some sort of LIDAR array looking for magnetic perturbations caused by these impactors. A system like this would have similar limitations as the IR solution.