The below represents prospective testing of satellite imagery, media, and ground source fusion to anticipate what is a seasonal and routine appearance of diarrheal disease in Bangladesh. This was an actual "live" operational drill we did, and the results were encouraging.
I have personally published on the concept of "Remotely Sensed Epidemic Intelligence (RSEPIS)" years ago, where the emphasis was on primary use of remote sensing products to "forecast" disease. The interesting thing was, in an academic treatment of the problem one obsesses with probabilities and proofing of models... almost to the point of operational irrelevance. In an operations environment, we had a much higher tolerance for false positives and model ambiguity. Therefore, if a phenomenon such as cyclone-triggered diarrheal outbreaks was considered "plausible" for a given area of the world, then that was all we needed to know to begin operational evaluation of an indicator life cycle that happens to include remote sensing-derived storm information.
Note also in the below example, an intuitive as well as data / statistical-driven deep understanding of disease reporting baselines are absolutely critical to realize an effective marriage of satellite imagery-derived information and disease surveillance.
Lastly, a point about "forecasting". That word has gotten me into so much trouble throughout my career due to the social expectations associated with the word. Almost implies a crystal ball-like power to predict. During my time at NASA, the scientists were obsessed with expressing probabilities. Understandable, when considering their perspective, however we as a society process meteorological information on a daily basis, with lives and billions of dollars at stake, within a sea of uncertainty. Therefore, we adopted the term "anticipatory analysis", to emphasize not so much prediction but the potential of a sequence of events leading to an unpleasant outcome regarding a particular disease.
This is but one example, where the anticipation window was 6 days. With an increasingly solid hypothesis of a cyclone forming south of Bangladesh and about to swing north to impact the country, a data review to check probabilities of 1) reports of an epidemic of diarrheal disease and 2) associated social disruption parameters occurred in the days prior to actually seeing reports. In other words, issuance of an advisory at (-6 days) with daily updates that included expected parameters to be observed in the days to come.
We participated in several other similar examples in various locations of the world.