I think modeling is a way that can lead to estimating or predicting based on past water quality data, climactic and watershed conditions, observations, etc., and is typically what we do in research, especially for water quality. There are so many things that we may be able to estimate effects, but difficult to forecast, eg., chemical or industrial spills, land use change, wildfires, droughts, floods, etc.

Weather is something we apply substantial resources to achieve a degree of prediction. Much more weather data is typically collected and available to assess, predict and to a certain extent forecast. Weather forecasting has its limits, but has come along way with Doppler radar and other networks of data recording tools. We no longer rely so much on sayings like "red sky at morning, sailers warning, red sky at night, sailers delight". Those with internet access, cell phones and weather apps enjoy almost immediate updates, with basic forecast for weeks or longer.

But if forecasting water quality is a goal, the complexity and number of parameters monitored and network of data collection stations would need to increase. I don't suppose most would take offense if you prefer the term 'forecast', but it may not be as accurate as far as our abilities. Granted, some water quality elements are very stable, so our ability to predict and/or forecast these may be good. Some elements vary over several orders of magnitude with high coefficients of variation, much more difficult to predict, let alone forecast.

I believe 'forecasting' and 'prediction' mean the same thing. Using what we know to have a glimpse of the future to some degree of accuracy. What method we then employ to achieve this (machine learning or not) is irrelevant.

To me the term "forecasting" is more associated to an event that can happen in future and term "prediction" is more associated to a situation in present time. For example: if I use a set of easily measured variables to predict a result that is usually obtained with a more laborious analysis, it is a "prediction". However, if I create a model to estimate the future behavior of a system, it is a "forecasting". But all this considerations is only my opinion.

I think modeling is a way that can lead to estimating or predicting based on past water quality data, climactic and watershed conditions, observations, etc., and is typically what we do in research, especially for water quality. There are so many things that we may be able to estimate effects, but difficult to forecast, eg., chemical or industrial spills, land use change, wildfires, droughts, floods, etc.

Weather is something we apply substantial resources to achieve a degree of prediction. Much more weather data is typically collected and available to assess, predict and to a certain extent forecast. Weather forecasting has its limits, but has come along way with Doppler radar and other networks of data recording tools. We no longer rely so much on sayings like "red sky at morning, sailers warning, red sky at night, sailers delight". Those with internet access, cell phones and weather apps enjoy almost immediate updates, with basic forecast for weeks or longer.

But if forecasting water quality is a goal, the complexity and number of parameters monitored and network of data collection stations would need to increase. I don't suppose most would take offense if you prefer the term 'forecast', but it may not be as accurate as far as our abilities. Granted, some water quality elements are very stable, so our ability to predict and/or forecast these may be good. Some elements vary over several orders of magnitude with high coefficients of variation, much more difficult to predict, let alone forecast.

Yes. It is some sort of forecasting, though it can be difficult to 'prophecy,' in many cases.

Do not mix up forecasting with prediction - because prediction is based on employment of analytical tools, whilst forecasting at best - is based on hypersensitivity of of some people to sense that other can not.

The latter, however, may work correctly, but there is a considerable possibility that it will not work in a concrete case and if to speak of quality of water mistakes may be very dangerous.

Prediction is forecasting; so you are right.

Forecasting is more often used to describe regression model output. When I use a classifying algorithm I say I am predicting, this is how people in the business/finance field use the term. You can predict high default probability with a classification algorithm. But most forecasts are for finance/sales or other regression like output. You could use a random forest classifier and forecast abnormal sales in the next season but you might have a harder time explaining that to the board.