Hi all,
I'm a London based data scientist working on tracking cement plant activity globally as part of an international collaboration to track emissions from various point sources. https://time.com/5866881/al-gore-climate-trace-emissions/
At the moment I'm combining production indices, capacity data and satellite imagery to train my models. Given my limited understanding of the domain I could do with some help on technical questions.
Thus far I've labelled kilns (in cases where they aren't concealed by roofs) and tracked their activity over time using short wave infrared images from sentinel-2. Roughly speaking the idea is that more heat=stronger signature=greater output. Is this a naive assumption? The reason I ask is that my results so far don't match with production indices. In the US, for example, I get an uptick in activity in December despite production ostensibly ramping down during this period. This could be due to a processing error on my part but I'm wondering if there's a theoretical explanation?
My laymen's understanding is that kilns aren't switched off if there's sufficient demand. But statistics from cemnet imply that countries differ wildly in their capacity utilisation, with Spanish plants operating at around 35% capacity and US plants at around 80% for example. Is it naive to assume that this would imply Spanish plants will turn their kilns off much more frequently or is it too expensive to do this? If kilns aren't switched off, how are their temperatures affected by operating capacity. Is the temperature lower? I've noticed huge variance in the strength of infrared signatures from kilns at different plants. Even in cases where two plants both very obviously have kilns that are switched on, one might give off a much stronger infrared signature. This surprises me because I thought that kilns are relatively constant in temperature. Perhaps these observations are an artifact of the low resolution of the satellite imagery I'm using but again theoretical insight would be really useful.
A final question I have relates to the use of roofs to cover kilns. I've noticed that around 60% of kilns don't have a roof in that data I've looked at so far. It is obviously easy to see infrared signatures in these cases. What motivates the decision to use/not use a roof and in cases where I can't see kiln operation due to a roof, is there anything else I should look for?
Help really appreciated as I'm at a bit of an impasse!