You are here: Home >> Rivers >> Flooding >> Stage 4

Stage 4: Data analysis

The risk of flooding

Using tables and bar charts

• For each land use category, count the total number of squares in which the land use was recorded as dominant.
• Produce a summary table of your results. In the example shown below, 11 land-use categories have been used, and classified as either on of off the flood plain, but there are many other ways in which you could organise your results.

Land Use Category	On Flood plain	Off Flood Plain	Total
Pre War Low Cost Housing	15	24	39
Pre War High Cost Housing	2	28	30
Post War Low Cost Housing	6	12	18
Post War High Cost Housing	11	32	43
Retail / Admin - High Order	5	7	12
Retail / Admin - Low Order	24	19	43
Industry	5	12	17
Transport	28	33	61
Car Parks	14	4	18
Open Space (inc. agriculture)	80	13	93
Community Services	10	16	26
Total	200	200	400

Construct a mirror graph of results. This may look similar to the example below:

Calculating flood risk and mapping results

If you have collected data on flood likelihood and flood severity, you can calculate the risk score for each survey square using the following equation.

Risk = likelihood x severity

You can then map the risk of flooding for the study area. The example below shows how this could be done using a spreadsheet program.

The management of flooding

One way of displaying the results of the bipolar analysis is to construct a modified bar chart with the bipolar score on the x-axis and the different flood management techniques shown on the y-axis. The example below shows what this might look like.

Statistical tests

Association tests (like the chi squared test) can be useful. For example, if you have collected data on the likelihood of flooding and on land use (as described on the Data collection page) you can test the significance of the association between them. Need more information about this test?

Worked example

The null hypothesis is that there is no significant association between flood likelihood and land use.

The observed data was collected from 200 squares in a mixed residential / town centre area, both on and off the flood plain of the River Avon.

Each square was categorised as being of either at high likelihood of flooding (more than or equal to 2.5) or low likelihood of flooding (less than 2.5), and was categorised in one of 5 pre-determined land-use categories.

	High likelihood	Low likelihood
Land use	OBSERVED DATA	OBSERVED DATA
High-value retail	0	12
High-value housing / low-value retail	4	20
Medium-value housing	24	18
Low-value housing	37	12
Open space	62	11

First calculate expected values.

Expected values = (row total x column total) ÷ (grand total)

	High likelihood		Low likelihood		TOTAL
Land use	OBSERVED	EXPECTED	OBSERVED	EXPECTED
High-value retail	0	7.6	12	4.4	12
High-value housing / low-value retail	4	15.2	20	8.8	24
Medium-value housing	24	26.7	18	15.3	42
Low-value housing	37	31.1	12	17.9	49
Open space	62	46.4	11	26.7	73
TOTAL	127		73		200

Now calculate the sum of (observed - expected)2 ÷ (expected)

In our example, chi squared = 70.4

Now compare the calculated value of chi squared against the critical values for (columns of observed values - 1) x (rows of observed values-1) degrees of freedom.

In our example, at the p=0.05 probability level where degrees of freedom = 3, the critical value of chi squared = 9.49.

Since 70.4 > 9.49, the null hypothesis is rejected at the p=0.05 level.

Therefore, we are 95% certain that there is an association between flood likelihood and land use.

GO TO NEXT SECTION: Review