SPB Academic Publishing bv, The Hague

Landscape indices describing a Dutch landscape R. Maureen Hulshoff Landscape and Environmental Research Group, University of Amsterdam, Nieuwe Prinsengracht 130, 1018 VZ Amsterdam, The Netherlands

Keywords: pattern indices, land use development, quantification of changes in landscape pattern, dominance index, shape index, rate of change Abstract The data set of a human modified Dutch landscape was used to evaluate whether landscape pattern indices developed in the United States are fit to describe a Dutch landscape. The grid based data set contains the development of land use over the period 1845- 1982. The indices were divided in two groups: pattern indices and change indices. In the first group the proportion of each land use type (P), patch number (N), mean patch size (A) and two indices of patch shape a(B1102 It may therefore be stated that understanding the ecological principles of changing pattern is impor- tant for nature conservation and landscape plan- ning. The first step to understand the ecological prin- cipals is quantification of change in landscape pat- tern. This quantification can be done in several ways: first, by describing the change in words, secondly, by means of maps and graphic illustrations, and thirdly, be means of indices. This last method has re- cently developed and gives opportunity to treat the data statistically and to compare patterns on differ- ent scales. Several indices are already developed O’Neill et

(e.g. , proportion of each legend type), other indices give information on the shape of the mosaic (e.g., rate of change). Pattern indices

Five pattern indices are considered: 1. P = the proportion of each legend type, 2. N = number of patches, 3. A = mean size of patches, 4. S = mean patch shape index, 5. D = dominance index. Changes of P in time give information on the in- crease or decrease in area of a legend type. The index gives no information on changes in geographical position. So, when the position of a legend type is changed entirely but the proportion of the legend type has remained equal, P will have the same value. Number of patches (N) and mean size of patches (A) are direct information on the landscape pattern. Dominance index (D) measures the extent to which one or a few legend types dominate the land- scape C (Pk In Pk) where c is the total number of legend categories and Pk is the proportion of the landscape in category k. The index gives no information on which legend type is dominant. Large values of D indicate land- scapes that are dominated by one or a few land use types in a further heterogeneous landscape. A value close to 0 indicates the land use types cover equal surfaces. Shape index can be expressed in a meaningful way by the area and perimeter of a patch since communi- ty and population characteristics differ between patch interior and patch edge (Fahrig and Merriam 1985; Forman and Godron 1986; (e.g., Davis 1986; Forman and Godron 1986; Iverson The first shape index (S 1) is calculated as the aver- age of the perimeter-to-area ratio (Iverson 1988) of all patches of a legend type: 1/Nil C i. The index measures mean perimeter-to-area ratio of a legend type. A high value of (Forman and Godron 1986). Shape index S2 is calculated as the deviation of a registrated patch from an isodiametric patch with the same area (Forman and Godron 1986). So this shape index is calculated as the average of this deviation for all the patches of a legend type: S2 = li the perimeter and .OO. The more S2 deviates from 1 .OO the more the patches deviate from a isodiametric shape. Shape index 2 is comparable with the particular fractal dimension used by n is the total surface area in square km of the study area. The index (C) indicates the rate of increase or decrease in area of that legend type and gives no in- 103 formation on the rate of change in geographical position. Study area and development of land use The study area is a well documented area in the northern part of the Veluwe region, in the central Netherlands (Fig. 1). This part of the Netherlands is geomorphologically characterized by ice-pushed ridges and fluvio-glacial deposits, partly covered by Pleistocene and Holocene aeolian sands (Koster 1978). Nutrient-poor sandy soils, thick humous plaggen soils and humous-rich podzolic soils are the main soil types to be distinguished. Originally, the landscape of the Veluwe was cov- ered by mixed deciduous forest. Due to occupation and reclamation of forest for agricultural purposes in the period AD 800- 1100 the forest area gradual- ly decreased to make place for heathland (Vervloet 1977). At some places even aeolian drift sand areas originated, due to the combined effect of 1. The asterisk resp. the fine pointed area illustrate the posi- tion of the sample area in the Netherlands. 104 (llth- 14th century, cf. Lamb 1984). After deforestation the agricultural system changed from a stock-farming system into a system of arable farming known as ‘potstal-system’ (Pape 1970). This is a mixed farming system characterized by small fields and extensive areas of common heathlands. To fertilize the originally humus-poor podzolic soild the fields were almost yearly supplied with a mixture of manure and sods, litter or sand. Manure was collected from cattle and sheep grazing the extensive areas of common heathlands. The heathland was also used to gather sods, litter and sand (Vervloet 1977). Due to intensive use of the heathlands vegetation cover locally disappeared and extensive drift-sand areas arose. Although already reported in the 10th and 11th century, the problem of drifting sand is reported more frequently since the 15th century. For long times the techniques used for stabilizing blown sands were hardly successful. At the end of the 19th century the Dutch State became concerned with the defence against drifting sands. By planting (mainly) pine-forest and drift-sand areas decreased in size and number (Vervloet 1977). With the introduction of artificial fertilizers and cheap wool from Australia at about the kmz of woodland (1 percent of the na- tions area) was left (Harms and Opdam 1990). Large heathland re-afforestation programs were started resulting in the present landscape of the Veluwe which is covered by coniferous plantations and has small relict-patches of heathland and 1:50,000 scale were studied. The period between the editions is about 20 years. In an area of 10 x 10 km (Fig. 1) land use pattern is quantified. To do so a lattice with square gridcells of 200 x 200 (e.g., geomorphological orientation, wind direc- tion). The result is a total number of 2175 gridcells. The maps are sampled using the method of mo- dal sampling (Webster CIS program. For the shape indices size and perimeter of the patches were used, where a patch is defined as a group of contiguous, adjacent cells of the same land use type (adjacent = horizontally, vertically and diagonally). The second shape index (S2) is based on a formu- la given by W2dain). Because this formula is based on polygonal format in this case study the formula is adjusted for gridcell data. So the average of this shape index (S2) is cal- culated by: S2 = (li/4dai) where Nil is the number of patches of land use category i in map 1, ai the size of each patch in that category. Correlations between the indices are calculated with Spearman’s correlation coefficient with a = 0.05. Results First, the results of the separate indices will be dis- cussed, then combinations of indices will be made to evaluate the additional value of the combination. 105 Table U = urban.

1845 1875 18 1913 1931 1957 1972 1982 PROPORTION OF LAND USE TYPES Forest Heathland Drift-Sand Agricultural Urban 0.1 7 0.66 0.08 0.09 0.00 0.31 0.56 0.03 0.10 0.00 0.34 0.42 0.12 0.12 0.00 0.41 0.38 0.08 0.13 0.01 0.54 0.25 0.06 0.15 0.01 0.55 0.22 0.04 0.16 0.03 0. 0.15 0.03 0.14 0.04 0. 0.12 0.04 0.14 0.06 TRANSITIONS (%) F -> F F -> S Fr> U H -> H 1.2 F .o 17.9 81.7 0.1 0.3 12.6 68.6 15.5 3.3 12.8 80.2 3.7 38.5 49.1 2.3 25.1 67.5 0.9 0.4 34.0 61.3 0.4 2.0 12.9 74.2 10.2 H H -> A -> F S s -> u 29.4 60.3 0.0 4.0 3.7 4.3 0.0 82.5 9.5 0.0 10.5 29.1 59.3 1.2 0.0 13.4 0.9 0.0 74.4 11.4 3.5 5.3 91.2 0.0 0.0 3.3 0.0 0.0 86.6 10.0 A -> F A -> S A -> A A -> U U -> H 0.0 0.0 0.0 u U -> A u s

.o 0.0 0.0 0.0 35.7 71.0 74.2 Pattern indices Large changes in cover (P) are recorded for forest, heathland, drift-sand and urban, little change is recorded for agricultural land (Table 1). Forest and urban area increase, the area of heathland and drift-sand decreases. The increase of the urban area starts after 1931. Figure 2 shows an increase of the average number of patches (from 35 to 78) and a decrease of the mean size of all patches (from 249 to 112 ha). The increase respectively decrease of mean size of forest 106 Table 2. Pattern indices for each land use type (standard deviation between brackets). 1845 1875 18 1913 1931 1957 1972 1982 NUMBER OF PATCHES Forest Heathland Drift-Sand Agricultural Urban 16 5 3 11 13 11 1 12 0 14 16 9 15 0 5 21 9 13 1 14 31 9 13 6 7 35 6 10 19 6 19 6 14 14 7 20 15 19 17 0 MEAN SIZE OF PATCHES (ha) 1 147.2 (1 583.6) Drift-Sand 224.0 (226.7) Agricultural 74.5 (1 04.2) 0.0 Urban Heathland 437.5 (874.6) 276.0 (1 00.1 76.9 (97.0) 84.6 (1 04.2) 8.0 SHAPE INDEX 1 127.9 (59.0) 11 6.4 (60.9) 88.8 (47.9) 124.6 (66.0) 0.0 130.9 (63.3) 123.9 (65.1) 113.0 (66.5) 103.3 (54.8) 150.0 (37.5) SHAPE INDEX 2 (m/4dha) 1.86 (1 103.9 (60.9) 140.0 (56.9) 114.1 (68.4) 97.5 (52.6) 165.5 (39.3) 106.1 (71.3) 118.7 (55.8) 56.0 57.3 (77.1) (54.9) 100.3 140.4 144.1) (226.0) 9.3 13.1 (6.0) (14.4) (0.0) 130.1 (66.6) 0.0 1.21 (0.30) 1.32 (0.44) 107 IV v VII

I +=mean size 1 a, 8

0 3 aI a= 4 a: a=17 1~18 a=9 a1:8 4 =

m I ' I' I 1845 1913 1982 ' time Fig. 2. Number and mean size of all patches. and of heathland and drift-sand patches is enor- mous, whereas for agricultural patches little fluctu- ation in mean size is recorded (Table 2). Remark- able is the high standard deviation of this index (Table 2) due to the great differences in size of the patches (a few very large patches and several very small patches). Comparing the two shape indices (Fig. 4) large differences in results are observed and no correla- tion between these indices is found. Shape index 1, calculated as the perimeter-to-area ratio, shows a general trend of decrease until 193 (e.g., Fig. 4, agricultural land). These results are surprising. Where they should theoretically indicate the same trend they do the op- posite in some periods. An explanation for this phe- nomenon is the size dependency of shape index 1 and the size independence of shape index 2. In Fig. 3 the differences between shape index 1 and 2 are shown by means of an arbitrary patch that changes in size and shape through time. In period T v) b

c 81 N 10 Fig. 3. Development of an arbitrary patch (a) and its cor- responding shape indices (b). decreases and shape index 2 increases. In period IV-V shape index 1 remains stable whereas the complexity of the perimeter decreases. In period V-VI, when the patch size remains stable and patch shape develops from complex to a square shape, both indices decrease. On the strength of this exercise it seems justified to conclude that the in- dices have a different ecological meaning: changes in shape index 1 indicate changes in the patch in- terior whereas shape index 2 indicates changes in the complexity of the perimeter. Like all shape indices mentioned in literature each of these indices covers just one of the aspects of the ecological value of a patch, (e.g., drift-sand) are relict patches. Therefore, the shape of these patches is fixed by their human modified neighbour patches. So, particularly in a human modified landscape like the Dutch landscape no difference will be observed between the shape of semi-natural and human modified patches. Iverson (1988) also found that 108 200 FOREST 150 I -’- -2.0 -1.8 U .& Q 0) 5c C *=s2 *= * I -0 3- a, -1 -.

(D -1.4 N I 1 -1.2 1.0 200 AG R F aQ , v) e*= =s1 0 s2 1845 1913 1982 time mean deviation from an isodiametric patch. forest perimeters are often straight lines adjacent to agricultural land. His conclusion on that point is: the complex ‘natural’ boundary does not exist /.iihnrn hiim-n -otixTitTT I DRIFT-SAND I -.- -2.0 -1.8 3- (D -1 -. Q 1.0 1 URBAN .8 U .6 (1 .4 .2 S1 is mean perimeter-to-area and S2 is For the dominance index (D) the results of the analysis are not very clear (Fig. 5). Where the possi- ble range of this index for this case study has value Al-0 hptwppn 0.70 a, 0.60 c a, 0.50 II 1845 1913 1982 070, respectively. So, the trend indicated by the proportion of each land use type (P) is not reflected in the dominance index. This may be caused by the fine-grained pattern of the study area, where a variety of land use types is always present. This statement is confirmed by the fact that no significant correlation is found between dominance index and proportion of each land use I1 and the economic recession before and after World-War 11. For the particular land use types the index shows large fluctuations and no sig- nificant correlation is found between the index of the land use types separately. This is in contradic- tion to the change in proportions of the land use types (P), which are significantly correlated for 109 1.20 1 1845 1875 18 1913 1931 1957 1972 1982 period Fig. 6. Total rate of change in different periods between (e.g., forest and heathland). Obviously, the rate of increase or decrease of a land use type is determined by the transition of several land use types into that type, whereas sometimes the increase of the proportion of a land use type is cor- related with the decrease of one other land use type. Combination of indices

Proportion of each land use type (P) alone gives no sufficient information on the way land use develop- ment take place. First, the index gives no informa- tion on the change in geographical position of the land use types (same cover of a land use type gives the same P value). Secondly, the index does not in- dicate how the land use types are replaced by each other. The first problem can be dealt with by in- cluding information derived from the GIS data set. Using the utilities of the GIS for each land use type, two different proportion-indices can be calculated: a. the proportion of gridcells which have the same land use type as in a previous map and b. the pro- portion of new gridcells of the land use type. The way a land use type is replaced by other land use types is to be reconstructed from the combina- tion of proportion of land use type (P) and the tran- sition percentages (Table 1). This together with the rate of change makes the information on the land use development more complete. For the present study the combination of the indices P and C and the transition percentages gives the following 110 of patches gives a good indication of the pattern development. Nevertheless continued research will be necessary to develop a useful method how to quantify the change in landscape pattern. New indices giving in- formation on the change in geographical position of the patches has to be formulated, a new shape in- dex including the aspect of patch interior as well as the aspect of complexity of the perimeter has to be developed, and last but not least an ecological meaning has to be given to the index value in rela- tion to the process of changing pattern. Acknowledgements I gratefully thank Josef Fanta, Frans & Sons, New York. Fahrig, L. and Merriam, G. 1985. Habitat patch connectivity and population survival. Ecology 66: Biol. Conserv. 36: 115-141. Forman, R.T.T. and Godron, M. 1986. Landscape Ecology. John Wiley 65(5): 1585-1601. Iverson, L.R. 1988. Land-use change in Illinois, USA: The in- 111 fluence of landscape attributes on current and historic land use. Landscape Ecology 2: 45-61. Koster, E.A. 1978. De stuifzanden van de Veluwe: een fysisch geografische studie. Thesis, University of Amsterdam. Krummel, J.R., Gardner, R.H., Sugihara, G., Reidel Publishing Company, Dordrecht. DeAngelis, D.L., Genetische verarming: de problematiek van het beheer van kleine populaties. De Levende Natuur 1: 7-13. Pape, J.C. 1970. Plaggen soils in the Netherlands. Geoderma 4: Rafferty, J., Mor- ganstein, D. 1985. Statworks, statistics with graphics for the man- dominated landscapes of eastern United States. In Perspec- tives in Landscape Ecology. 1987a. Landscape heterogeneity and distur- bance. Springer-Verlag, New York. Turner, M.G. (ed.) 241- 251. Van den Berg, A,, Lentjes, P.G., Van Lith, J. and Roos, J. 1985. Handleiding van MAP2 versie 1.0. De Dorschkamp, Wageningen. Van Dorp, D. and Opdam, P.F.M. 1987. Effects of patch size, isolation and regional abundance on forest bird communities. Landscape Ecology Cultuurhistorie. In Rapport van het Veluwe-onderzoek. pp. 67-80. Edited by S.M. ten Houte de Lange. Pudoc, Wageningen. Webster, R. 1977. Quantitative and numerical methods for soil survey and classification. Oxford University Press, Oxford.