If it allows for more efficient use of resources elsewhere - for example a cloud-based app that picks shorter routes for delivery vehicles - then it's not that simple. The whole system contaminates less at that point.

1. Transport planning: While this may not be save, it can mitigate impact while we find better solutions. Efficient vehicular routing logistics, like ride sharing, dial-a-ride, school bus problems.
2. Consume planning: We can also reduce how much we eat, say an app like "eat this much"stops people going overweight.
3. We can also plan how we distribute this food efficiently, think uber but for food.
4. Wageningen has shown amazing agro tech that makes them more productive. For example to save water, coordinating water sensors and weather prediction. Crop quality analysis by image processing, drones to pesticide. But that is not code alone.
5. Indirect influence, this one is the most interesting, by helping people to lift their weights and do their jobs easier, learn better, etc. They might find how to save the world. Cool examples of this are in open source. Think on the relationships of Khan academy React and live editor.

Code is like math, it's nothing but a tool and it's useless if not used properly. Code (and computing and tech in general) could (and will) certainly help to solve the problem but I don't think any solution will come from it, it will just be a way to apply this solution.

That being said, code/computing/tech can also be an issue when it comes to environmental problems. Computing requires power and you know what that means for the environment.

I think what can be done with software is limited, because it acts in the virtual domain rather than the physical one.

Generally speaking, the problem isn't a lack of software, so the solution isn't going to be more software.

Keeping that in mind, any (partial) solution from software must be either about reducing software's own negative influence or providing useful information or communication for things that do affect the physical world.

The first option mostly means writing more efficient programs. Ditch python in favor of rust. This isn't going to accomplish much.
Alternatively you could store computations for times when renewable energy sources are overproducing. You'd be storing work rather than storing energy, the later being the focus of much renewable energy research. Even if you'd get all non-critical computation in the world to 0 emissions this way, I doubt it would make a big difference.

For the second, in essence, you'd play support for people who do fix or alleviate these problems. E.g. government agencies concerned with monitoring adherence to environmental laws or a relevant department at a university or research institute, both are often looking for software engineers.