With the harvest split equally over n sites, you can lose a whole number of sites and still have enough food only if n times the spare fraction (1 minus the share you need) reaches that number. If you need 60% of the stock to reach the next harvest, 5 sites survive the loss of 2, while 4 sites survive only 1. Adding a site helps only when it raises the number of losable sites by one; otherwise it adds visits, tracks and chances of loss.

Multiple storage sites are separate stores of a harvest, planned so that the sites that survive any plausible loss still hold enough food to last until the next harvest.

Start From The Share You Need, Not From The Number Of Sites

The parent article gives a fixed rule: 3-5 sites, none above 30%. That counts sites without asking how much food has to survive. Start from the other end. Let N be the fraction of the stock you must keep to reach the next harvest (planting seed included). With n equal sites, each holds 1/n of the stock. The number of sites you can lose is

m = the largest whole number not greater than n x (1 − N).

Sites nNeed 50%Need 60%Need 80%
3110
4210
5221
6321

Worked case, need 60%: for 4 sites, 4 x 0.4 = 1.6, so m = 1; losing two leaves 50%, below the need. For 5 sites, 5 x 0.4 = 2.0, so m = 2; losing two leaves exactly 60%.

Chance Of Falling Short

If each site is lost independently with chance p, you fall short when more than m sites are lost. The chance is the sum of the binomial terms for k > m lost sites. For need 60% and p = 0.2 (an illustration, not an estimate):

Sites nLosable mChance you fall short
100.200
200.360
310.104
410.181
520.058
620.099

Working for n = 4: chance of losing none is 0.8⁴ = 0.410; chance of losing exactly one is 4 x 0.2 x 0.8³ = 0.410; the sum is 0.819, so the chance of losing two or more is 1 − 0.819 = 0.181. For n = 3 the chance of two or more is 3 x 0.2² x 0.8 + 0.2³ = 0.096 + 0.008 = 0.104.

The result is not what the parent article’s rule suggests. Going from 3 sites to 4 raises the chance of falling short, from 0.104 to 0.181, because m stays at 1 and there are more sites to lose. Two sites are worse than one. The count improves only where m steps up: here at 5 sites. This depends on need and on p. Pick n from your own N, not from a fixed range.

Where The Independence Assumption Fails

The table assumes sites are lost independently. Real sites fail together when:

  • they sit on one route, one watercourse or one floodplain;
  • the same person tends and visits all of them, leaving one trail;
  • one rodent species or one damp spell affects all of them;
  • one person alone knows where they are, and is lost.

Each shared cause turns several sites into one for that risk. Choose sites that differ in route, ground, type (root cellar, pit, dry loft, a neighbour’s store) and who knows them. See Cache Placement and Hidden Food Caches for hiding each one.

Records And Rotation

  • Who knows. Record each site’s location and contents in two places, and make sure at least two trusted people know each one. A site only one person knows is lost when that person is. The wider problem of knowledge that dies with its holder is in Knowledge Preservation.
  • Rotation. Eat the oldest stock first and refill; a site that is never opened is not checked for mould or rodents. See Food Storage Infrastructure.
  • Visits. Every visit leaves a track. Stagger the visits and vary the route.

Common Mistakes

  • Adding sites without recomputing m. Going from 3 to 4 sites at need 60% leaves m at 1.
  • Equal visits to all sites. A worn path points to all of them.
  • Counting a shared barn as two sites.
  • Planning only for raids. Rot, flood and fire also take stores.

Opening a sealed or underground store

Do not enter a pit, cellar or sealed vessel that has held respiring vegetables, fermenting food or grain without ventilating it first: carbon dioxide displaces oxygen without any smell, and a person can collapse in seconds. Never test the air with a flame, and never go down alone. See Underground Storage.

What’s Next