They don't. Usually you have a general topic you like to talk about, but you can't because it is all well known. So you have to give it just a little twist. T

They don't. Usually you have a general topic you like to talk about, but you can't because it is all well known. So you have to give it just a little twist. This can be a new method you try, for example a neural network that does some thing. You then explain the general topic, then give some Infos about the NN and may compare its success to an established method. This has been done in every field now, that people already get sick of it. But this generally holds, just apply an out of field method and compare. It makes for interesting content and you can regurgitate the known theory to get enough mass. If you have hard data, that is not easily available, because the process to receive it is espensive, you can publish with just some basic evaluation. Just make it a tiny bit more shiny. In your case with environmental modelling you should be in between the two. Just keep it simple and well struktured. The scientific process is the art of adding a lot of very small steps. You contribution can be tiny if you already walk a well known path. It's all luck and knowing people in the end anyway.