In my copy of the rules on page 11 it states that harvest failure/surplus is simply a random probability event, modified by the natural fertility of certain regions, so bumper harvests are more likely in Egypt, Poland, Prussia, Baltic States, North America (English mainland colonies), Russian Ukraine, Greece (Morea), Sicily, Hungary; grain shortages are more likely in Asia Minor, UDP, Spain, England, Tuscany, Portugal, north/central Russia, Venice, Rumelia.
It was also pointed out that some regions historically supply others so:
Egypt supplied grain to Ottoman Asia Minor/Anatolia
Poland, Prussia and the Baltic States supplied grain to UDP, Spain, England, Tuscany and Portugal.
English North America supplied England
Russian Ukraine supplied Central/Northern Russia
Greece (Morea) supplied Ottoman Europe and Venice
Sicily supplied Tuscany, Venice and Spain
So although some regions are more susceptible to harvest failure than others, they also had natural links to help them overcome its effects.
Reclaiming land "can provide rich farmland", though it doesn't say how often this can be done.
Shipping in top soil from a neighbouring region is expensive, but does "improve" the fertility of a region.
If night soil men are used in every named town then "the annual chances of a good harvest are significantly increased" However, the same does not apply to manure which is "deficient in important nutrients and less effective in this regard than fertiliser. It does however improve water retention and loosen soil."
There is also the chance of "Catastrophic Crop Failure" which is defined as a deficit of more than 100,000 tons per million population. This is more serious because it does give an increased risk of failure in the following year as there is not enough grain for farmers to plant. Another interesting note is that charitable giving of grain risks impoverishing farmers which encourages them to leave the land and increases the risk of future poor harvests!
Enclosures "increases the chance of a good harvest".
Irrigation canals give a permanent increased chance of a good harvest each year.
There are also ways to use different types of seed grain, but I haven't experimented with any of them. They seem to promise great benefits if they work, but risk a catastrophic failure if they fail. Small countries might get away with it, but for a large country like France a catastrophic failure would require 4,400,000 tons. I hate to think what the figure would be for China!
Although any bumper harvest is expressed as extra tons, I think that is for game convenience rather than as an calculation based on ordinary yield. In the rules France's population in 1700 is 22,000,000 so a normal harvest for France would be 2,200,000 tons. A good harvest is normally an extra 50,000 tons per million population so a normal good harvest for France would be 1,100,000 tons. but a bad harvest would require 100,000 tons per million population so a normal bad harvest for France would require 2,200,000 tons to be found. By using various sources for alternative food, culling wild animals, salted fish etc, this may reduce by 50%. Doing the maths it appears you are broadly neutral if you have 2 good harvests to each bad one (provided that bad one is not a catastrophic one!)
There is no doubt that famines can destroy EH very quickly and hit tax revenues. However, low EH usually leads to more recruits so there is a positive to it.
Manure does lose its potency and soil structure does deteriorate if not improved. So perhaps there is a benefit in repeating land improvement projects every 5-10 years?
I also think it is worth remembering that famines were quite usual in the 1700s and the effect is that some of your population shrinks by 5%: the nobility/church are not affected, only peasants. These peasants may well emigrate to neighbouring countries and boost their population temporarily only to move back home when food is available. So the long term effects may not be as horrific as we initially think. Perhaps famines act as a break on a very strong economy? I haven't found anything in the rules to suggest they are dependent on the weather reports.