Bridging theory and implementation – Testing an abstract classification system for practical mapping by field survey and 3D aerial photographic interpretation

Figures & data

Table 1. Aerial photographic interpretation and field survey by parameters most often used to define an applicable method for a given project

Defining parameter	API (aerial photographic interpretation)	FS (field survey)
Purpose of mapping (end users’ needs)	Overview of land cover content or detailed knowledge of local area, monitoring of land cover distribution and changes	Detailed knowledge of species distribution and land cover state of local area
Available classification systems (ecological resolution and information complexity)	Typically coarser systems that focus on spectral properties, vertical structure and texture indirectly assessing species composition	Detailed systems with focus on indicator species and species composition
Spatial resolution (the intended map scale and minimum mapping area, mma)	Coarser mapping (scales 1:5000–1:50,000, mma from 0.01–4 ha)	Detailed mapping (scales 1:500–1:25,000, mma from 0.0001–2 ha)
Available economic resources for mapping (budget)	Small budget (given available photographs and equipment) or large area	Large budget or small area
The time schedule versus size of the area intended for mapping	Fast mapping progress	Slow mapping progress
Human resources and competence	Experience of API and geographic information system (GIS) methods, good computer skills, and basic photogrammetry skills Knowledge of land cover, land use, and vegetation in region Knowledge of species ecology improves interpretation	Experience of field survey methods Knowledge of species, ecology, edaphic conditions, and land use in the study area
Limiting factors of the study area (context-dependent variation)	Shadow effects in steep areas Tree-coverage, mosaic features	Accessibility, infrastructure
Available technical solutions and material (equipment, software and material)	Aerial photographs 3D photogrammetric computer station, digitizing software	Orthophotos in GIS on a field computer or printed orthophotos and digitizing software
Scientific point of departure	Geosciences, resource management, and planning	Biosciences, nature management and conservation

Fig. 1. Study area with four rectangular study sites on Hvaler Municipality, south-east Norway (Maps from Geonorge (N2000 maps and AR5 area cover), WGS 1984 UTM Zone 32N, rotation -5.1 degrees)

Table 2. FragStats metrics used to describe differences between the vegetation maps, and the meaning of each metric

Metric (FragStats code)	Explanation
Number of polygons (NP)	Count of number of polygons
Edge density (ED)	Metres of edge per hectare (m/ha)
Mean polygon area (A_MN)	Mean polygon size (ha)
Median polygon area (A_MD)	Median polygon size (ha)
Standard deviation of area (A_SD)	Standard deviation of polygon area (ha)
Area of largest polygon (LPI)	Percentage of map area covered by the largest polygon in each map
Connectance index, threshold distance set to 10 m (CONNECT)	Functional connection on a scale from 0 to 100, where 100 means that all patches are connected
Shannon’s diversity index (SHDI)	Diversity of ecosystem units quantified by the probability of encountering new units in the map, recorded on a scale from 0 to infinite. The probability of encountering new ecosystem units increases with larger numbers.

Table 3. Criteria used for partitioning plot and map relations into three categories

Category	Criteria
Match	Plots where the map and the plot had corresponding ecosystem units primary ecosystem unit in a mosaic polygon matched mosaics matched but the order of ecosystem units was switched
Match with small delineation variation	plots closer than 5m to delineation and that matched the ecosystem unit on the other side of the polygon border plots that matched with a secondary ecosystem unit in a mosaic
Different classification	Plots with different classification (DC) than given by the polygon

Fig. 2. The ecosystem units for forest in NiN 2.0 as defined by the gradients ‘Risk of severe drought’ and ‘Basicity’; two examples of ‘different classification’ with ecological distances of two and five

Table 4. Grouping of ecosystem units present in study areas, with ecosystem unit codes and names

Binary group	Unit codes	Unit names
Basicity	T32-C3−5	Intermediate and moderately calcareous grasslands with low and moderate management intensity
	T4-C2	Sparse low-herb forest
	T4-C6	Sparse low-herb heather-bilberry forest
	V2-C2	Intermediate wetland forest
Acidity	All other units
Agriculture	T32-C3−5	Intermediate and moderately calcareous grasslands with low and moderate management intensity
	T44-C1	Ploughed field
No agriculture	All other units

Fig. 3. Ecosystem unit maps for sites API1 and FS1N; with different sites (coordinates WGS 1984 UTM Zone 32N, rotation -5.1 degrees)

Table 5. Landscape metrics given as average results for the two aerial photographic interpretation (API) sites and the two field study (FS) sites

Metric and FragStats code	API	Field study
Number of polygons (NP)	148	209
Edge density (ED)	609	831
Mean polygon area (A_MN)	0.34	0.24
Median polygon area (A_MD)	0.11	0.09
Standard deviation of area (A_SD)	0.96	0.52
Area of largest polygon (LPI)	18.1	7.8
Connectance index (CONNECT)	2.09	2.37
Shannon diversity index (SHDI) value	1.73	1.86

Table 6. Area covered by the primary ecosystem units, given for units evaluated by a minimum of one plot, including the number of plots for each ecosystem unit, the percentage accuracy in plot and map comparisons (producer’s accuracy), and the percentage accuracy of map and plot comparison (user’s accuracy)

Ecosystem unit		API (aerial photographic interpretation)					FS (field survey)
Name	Code	Land cover (%)	Number of plots	Plot & map accuracy	Number of plots on map	Map & plot accuracy	Land cover (%)	Number of plots	Plot & map accuracy	Number of plots on map	Map & plot accuracy
Drought-prone acidic rock	T1-C2	45	17	100.0	30	90.0	29.5	24	100.0	27	100.0
Heather-bilberry forest	T4-C5	12.7	13	61.5	18	38.9	14.3	13	53.8	9	77.8
Heather forest	T4-C9	23.5	21	81.0	19	68.4	31.5	22	100.0	23	73.9
Open acidic shallow-soil heath	T2-C1	-	5	80.0	-	-	6.2	4	75.0	5	80.0
Bilberry forest	T4-C1	0.1	9	11.1	-	-	1.7	3	0.0	-	-
Acidic wetland forest	V2-C1	0.5	4	25.0	1	100.0	2.4	5	80.0	4	50.0
Intermediate grasslands with low management intensity	T32-C3	0.8	1	0.0	-	-	0.1	1	0.0	-	-
Arable field	T44-C1	0.3	2	50.0	1	100.0	0.6	1	0.0	-	-
Sparse low-herb forest	T4-C2	1.3	2	0.0	2	50.0	0.1	-	-	-	-
Sparse low-herb heather-bilberry forest	T4-C6	0.4	1	0.0	1	0.0	0.5	-	-	2	100.0
Lichen forest	T4-C13	2.9	-	-	1	0.0	0.8	-	-	-	-
Upper rock/gravel beaches with pioneer vegetation	T29-C1	0.1	-	-	-	-	0.3	1	0.0	-	-
Upper rock/gravel beaches with established vegetation	T29-C2	-	-	-	-	-	0.6	-	-	1	0.0
Intermediate grasslands, moderate management intensity	T32-C4	1.8	-	-	5	0.0	0.6	1	100.0	2	0.0
Moderately calcareous grasslands, low management intensity	T32-C5	-	1	0.0	-	-	-	-	-	-	-
Plantation forest	T38-C1	-	-	-	-	-	0.2	1	100.0	1	100.0
Stone heap	T39-C1	-	-	-	-	-	4.2	4	100.0	5	100.0
Meadow-like ploughed field	T41-C1	1.0	-	-	-	-	0.4	-	-	1	100.0
Lawns, parks, etc.	T43-C1	1.2	1	100.0	1	0.0	1.3	-	-	-	-
Very acidic mire	V1-C1	-	1	0.0	-	-	0.2	-	-	-	-
Fairly acidic mire	V1-C2	0.7	-	-	1	0.0	-	-	-	-	-
Intermediate wetland forest	V2-C2	-	2	0.0	-	-	0.1	-	-	-	-
Total		87.7	80	62.5	80	62.5	93.8	80	82.5	80	82.5
Count selection		15	14		11		20	15		11
Count total	41	27	14				34	12

Fig. 4. Evaluation results in three categories, with the sum of ‘Match’ and ‘Match with small delineation variation’ used as a measure for classification accuracy

Fig. 5. The mapping process and use of aerial photographic interpretation (API) (grey boxes) and field survey (FS) (green boxes) both as single methods tested in the study and as combinations suggested by the study