Dear Rajali,

in case there is no historical data, you need additional assumptions, I'm afraid. Sometimes, you may find analogies or there are mechanism, that lead you to meaningful assumptions. As heavy tailed distributions are particularly common in the modeling of critical events, the assumption of a normal distribution may turn out to be quite conservative... (see e.g. DOI: 10.1007/978-90-481-9171-0)

Best,

Thomas

I think it is important to recognise that the distribution accuracy is not the main goal. However, the argument why normal distribution might be a good idea is not very difficult to understand.

mathematically speaking:

There are various disasters D1,...,Dn given, each with various impacts I(D1),..., I(Dn). Let assume the function I(.) has values in real numbers and therefore can be ordered (e.g. a single burned house has far less impact than a tornado).

Assumption 1: I(.) exists and has values in real numbers

Let assume that given I(D), the demand distribution is certain (but not necessarily observable ex ante).

Assumption 2: Demand density only depends on the impact-value of a disaster.

Logically between two similar disasters Di and Dj (mathematically inside an \epsilon-neighbourhood), the impacts are expected to be similar as well I(Di) \approx I(Dj) (difference is less than \epsilon), which implies the demand distribution should be "similar" as well.

In non-technical words: demand density should not be very different between no-disaster and small-disaster, between small-disaster and not-so-small-but-still-small-disaster, ..., and between big-disaster and a catastrophe. The density should only move "a little bit" every time the impact increases. When demand density is usually modelled by normal distribution, then using normal distribution for disaster case is certainly not a bad idea. At most, the density will probably skewed or fat-tailed.

The most straightforward answer to the specific question is: "No".

In all seriousness, the huge number of variables associated with disaster response make such projections incredibly challenging and almost universally incorrect to a frustrating degree. That is not to say that they are without value - some system must be used, so that it is possible to develop successively more accurate projections.

As an example - consider an extremely strong earthquake striking an island nation.

Initial Assessment

This would provide some basic information for an initial shipment of the most urgently needed resources, by expedited means, to support the response phase of operations. It will tend to be the highest cost for transportation, so shipments would be limited to high value, dense packed, urgently needed materials.

At the same time, planning should be undertaken to determine what materials will be needed as the response phase transitions into a response/initial recovery phase. Because of the greater shipping time for bulk materials, it would be important to project needs for several months, based on some of the following factors:

It is important for the logistics function to both serve the field operations by meeting the needs; but also serve by prompting field leadership to provide the information needed to serve them well. Logistics considerations are not always at the forefront of field leadership thinking - so logistics serves them by gently guiding them to decision.

Thank you very much for all your replies, this is exactly the dilemma, while some suggest estimating these distributions, others question its possibility. But like Christopher Paulus Imanto suggested, normal distribution would not harm when there is absolutely no other ways to estimate the nature of distribution.

Thanks again, this helps.