Forecasting and Predictions

Meteorologists have made excellent progress in the past few decades to improve our abilities to forecast when rain events might occur over the next week or so, which facilitates the short-term forecasting of floods and droughts discussed in the paragraph above. However, complex atmospheric dynamics prevent forecasts beyond more than a few weeks in advance. Nevertheless, we must have some basis for making decisions about development, infrastructure and agriculture (e.g., How big should we build a culvert under a road? What size retention basin is needed next to a new housing development? Which agricultural fields will require artificial drainage and which will require irrigation?).

For these longer-term predictions we can use statistics to determine how likely it is that a given location will experience, for example, more than 10 cm of rain in a day, or less than 5 cm of rain during a given month. Many million- and billion-dollar decisions about development and infrastructure are based on such predictions.

To make these predictions, hydrologists synthesize historical data and use a statistics-based approach to determine the likelihood that a given event might occur. While Figure 1 highlights the ‘messiness’ of precipitation events over time, reorganization of the data provides useful information. Figure 2 shows a histogram of the precipitation data presented in Figure 1. A histogram is a plot showing the number of events that fall within chosen data groups or “bins” (shown on the x axis). From these data you can quickly determine that Kingston, NY experiences no rain about 2 out of every 3 days (731 out of the total 1096 days in this record). Only 10 days in the record had rainfall that exceeded 6 cm, so from these data alone you would expect such large rainfall events to happen 10 days out of 1096, or about 1% of the time. On 210 days during this time period the amount of rainfall was between the minimum measurable (typically 0.025 cm or 0.01 inch) and 1 cm (0.4 inches).

Hydrologists tend to use the term ‘forecast’ when referring to a future projection for which we have a lot of information (and therefore relatively high certainty of when an event might occur and what magnitude it might be). In contrast, hydrologists use the term ‘prediction’ for future projections for which less information is available, and therefore uncertainty is greater.

Figure 2. Histogram showing the number of events (specifically, total rainfall measured for each day from October 1, 2010 to September 30, 2013) in each of the bins indicated on the x axis. Bin labels are simplified for illustration purposes. For example, the bin labeled 0-1 includes all 210 days where precipitation was between 0.025 cm (minimum measurable) and 0.999 cm.

Text description of Figure 2

Daily precipitation categorized by precipitation amount.

Daily Precipitation (cm)	Number of events
0	731
0-1	210
1-2	70
2-3	43
3-4	14
4-5	8
5-6	10
6-7	2
7-8	3
8-9	2
9-10	1
10-11	1
11-12	0
>12	1

Source: Patrick Belmont, data from NOAA National Climatic Data Center, as used in Figure 1 on the previous page.