Population map: dasymetric disaggregation

The text on this page is taken from an equivalent page of the IEHIAS-project.

Figure 1. 1 km dasymetric population density map: EU strategy 1

Figure 2. 1 km population density based on LAU2 regions

Four different modelling strategies were applied to estimate population, reflecting different aggregations of the original land cover classes. Strategy 1 was the most detailed, with strategy 4 being the most simplified.

All the modelling strategies performed relatively well at both national and EU level, with R2 between modelled and actual population at NUTS 5 (~ LAU-2) level generally between 0.8 and 0.9, and SEE in the range of 2–5000 people. Differences between strategies were slight, with overall performance decreasing slightly as the models were simplified. To minimise the effect of step changes at country borders, the EU models were preferred for European mapping.

A map for the EU strategy 1 model (R² of 0.82; SEE of 4518 people) in Figure 1 is contrasted with a population density map based on the NUTS 5 census data (Figure 2). The weights for the EU strategy 1 model are given in Table 1 below.

Table 1. EU strategy 1 model weights

Land Cover Class	Lit area	Unlit Area	Unlit Area
1: Continuous urban	187.0	142.3	0.023
2: Discontinuous urban	98.9	85.4	0.007
3: Industrial / commercial	141.2	95.7
4-6: Transport	54.0
7-9: Waste, extraction and construction sites	13.3		0.009
10: Institutions	128.1		0.023
11: Urban green space	62.6	19.7
12-17,19: Arable and permanent crops	0.4	0.2	0.005
18: Pasture	1.6	0.6
20: Complex cultivation patterns	1.0		0.013
21: Land principally occupied by agriculture			0.017
22,24,25: Other woodland		0.0
23: Broad-leaved woodland	8.8	0.2	0.005
26-29: Unimproved grass and scrub			0.009
30-38: Other uncultivated		0.4
49: Unclassified surfaces	19.1	1.0