SIMMAP 2 - Universidad Politécnica de Madrid
Contents
1. colour corresponding to each of the classes is displayed Just press the button Change below Class colour and select the colour with which you want that class to be displayed or specify it in detail by its RGB components by pressing Define custom colours 3 3 Map linear dimension L Map linear dimension L is simply the length in pixels of the side of the square pattern to be obtained Thus the total number of pixels in the resultant patterns is LxL L 300 3 4 Minimum mapped unit m m is the minimum mapped unit size of the smallest patch to appear in the MRC patterns The default and minimum value is 1 patches comprised by a single pixel will exist Varying m you can simulate landscape patterns corresponding to different scales or different degrees of detail in the interpretation of remotely sensed images m 20 m 50 3 5 Neighbourhood criterion N This parameter is the last one you may want to vary once you previously understood and tested the influence of the previous ones In general there is no need to vary this parameter You only need to change the neighbourhood criterion if patterns with anisotropy with patches oriented in a certain dominant direction are to be obtained This parameter controls how patches clusters strictly speaking are built from initially random binary patterns see the paper by Saura and Martinez Millan 2000 for further details For each neighbourhood criterion a2. tested but they are very rarely used You may fix the font sizes in your computer in Windows gt Control Panel gt Display gt Settings gt Font Size 3 Simulation parameters There are five simulation parameters that influence the characteristics of the patterns generated by SIMMAP These are Initial probability p Number n and abundance of the classes Linear dimension of the pattern L Minimum mapped unit m Neighbourhood criteria N The controls for all these parameters are included in the Simulation parameters panel located in the upper part of the SIMMAP 2 0 window A very wide variety of spatial categorical patterns can be obtained by adequately varying these simulation parameters as described below First of all it is important to note that SIMMAP is based on a stochastic simulation method That is any number of patterns can be obtained for the same values of the simulation parameters which differ in the exact location of classes in the pattern each generated image is really unique but are similar in their overall spatial structure and appearance This is illustrated in the figure below SA ti Pi Mle ne M ak 1 erat E a f mha os Fa p a Par k a amp dou io z a Aa 3 1 Initial probability p This is the main simulation parameter p controls the degree of fragmentation of the obtained patterns When p is small patches are more numerous and smaller and t
3. this is opposite to what occurs in percolation maps 3 2 Number n colours and abundance of the classes Patterns with any number of classes n can be generated as shown in the figure below By clicking the button in the Simulation parameters panel you can modify the abundance of the classes as well as the display colour for each class 90 10 50 50 20 80 To modify the percent of total area to be occupied by each of the classes you have to input the adequate values for each of the classes in the Class weights edit box You may enter any value bigger than o SIMMAP internally normalises the weight values w that are introduced by the user and converts them to abundance probabilities a so that their sum equals 1 as follows where aj and wi are respectively the abundance probability and weight corresponding to class 1 and n is the total number of classes in the pattern For example if you introduce the weights of 25 and 50 in a two classes pattern SIMMAP will assign 0 33 and 0 67 as the abundance probabilities for each of the classes SIMMAP assigns classes to the patches in the pattern in a probabilistic manner that is aj is the probability that class i is assigned to a given patch in the pattern whatever its size In many cases the obtained class abundance actual percent of the pattern area occupied by each of the classes in the final image will be very close to 100 a the requested ab
4. LoS ONE 3 8 e2999 Hufkens K Bogaert J Dong Q H Lu L Huang C L Ma M G Che T Li X Veroustraete F Ceulemans R 2008 Impacts and uncertainties of upscaling of remote sensing data validation for a semi arid woodland Journal of Arid Environments 72 1490 1505 Millington J Romero Calcerrada R Wainwright J Perry G 2008 An agent based model of Mediterranean agricultural land use cover change for examining wildfire risk Journal of Artificial Societies and Social Simulation 11 4 La Sorte F A Hawkins B A 2007 Range maps and species richness patterns errors of commission and estimates of uncertainty Ecography 30 649 662 Li X Z He H S Bu R C Wen Q C Chang Y Hu Y M Li Y H 2005 The adequacy of different landscape metrics for various landscape patterns Pattern recognition 38 12 2626 2638 Shen W J Jenerette G D Wu J G Gardner R H 2004 Evaluating empirical scaling relations of pattern metrics with simulated landscapes Ecography 27 4 459 469 ZZ Li X Z He H S Wang X G Bu R C Hu Y M Chang Y 2004 Evaluating the effectiveness of neutral lanscape models to represent a real landscape Landscape and Urban Planning 69 1 137 148 Saura S 2002 Effects of minimum mapping unit on land cover data spatial configuration and composition International Journal of Remote Sensing 23 22 4853 4880 Saura S Mart nez Mill n J 2001 Sensitiv
5. SIMMAP 2 0 Landscape categorical spatial patterns simulation software USER S MANUAL May 2003 Santiago Saura ETSI Montes Universidad Polit cnica de Madrid Ciudad Universitaria s n 28040 Madrid Spain E mail santiago saura upm es This manual software and related information can be downloaded from http www2 montes upm es personales saura INDEX 1 About SIMMAP 2 2 System and display requirements 3 3 Simulation parameters 3 4 Display and save options 11 5 Spatial pattern indices 13 6 Limitations and known errors 19 7 Where has been SIMMAP used 21 1 About SIMMAP SIMMAP is the result of implementing the modified random clusters MRC hereafter simulation method This method provides more general and realist results than other commonly used landscape models as has been described in the paper by Saura and Martinez Millan 2000 which is referenced below The MRC method generates categorical thematic landscape spatial patterns in raster format grid based data SIMMAP is distributed without charge for non commercial use with the only condition of citing the following two references in any document or work in which SIMMAP is used Saura S and J Martinez Millan 2000 Landscape patterns simulation with a modified random clusters method Landscape Ecology 15 7 661 678 Saura S 1998 Simulaci n de mapas tem ticos mediante conglomerados aleatorios Proyecto fin de carrera Escuela T cnica Superior
6. comma should be set as your decimal separator in the regional configuration settings of your computer when using SIMMAP 2 System and display requirements SIMMAP runs on a PC Windows environment Windows 95 or newer It can be used in any standard PC with at least 16 MB of RAM memory More RAM memory is recommended if big patterns with big patches are to be generated A point and not a comma should be set as your decimal separator in the regional configuration settings of your computer when using SIMMAP SIMMAP will be better displayed in screens of 800x600 pixels or more In smaller screens e g 640x480 you may have problems to simultaneously see all the SIMMAP windows You can adjust the display size in Windows gt Control Panel gt Display gt Settings fixing the Desktop area to 800x600 pixels or more Also for an adequate colours display the colour palette in your computer should be set to 16 million colours or more If the icon in the upper left corner of SIMMAP windows is not shown in brown and green colours then you need to change the number of colours in your palette you are probably working with only 256 colours You can change it in Windows gt Control Panel gt Display gt Settings fixing the Colour palette to 16 million colours true colour 16 bits or more SIMMAP has been designed to run with small fonts 96 dpi Big fonts 120 dpi will also work fine Other font sizes out of this range 96 120 dpi have not been
7. d maximum PC 1 when every pixel is included in a single patch that fills the landscape 19 5 5 About the comparison of the indices values calculated with SIMMAP and FRAGSTATS Some users may wish to make the indices values given by SIMMAP equal to those provided by a commonly used software like FRAGSTATS If this is the case select the option Include inner edges in perimeter and deselect the option Perimeter as dependent variable Also compute the indices in the raster version of FRAGSTATS with the 4 neighbourhood rule i e do not use diagonals in patch finding This way the same values of NP LPI MSI AWMSI PAFD the initials for this index are DLFD in FRAGSTATS and PL the initials for this index are LAND in FRAGSTATS will be obtained with both programmes Some other indices are also comparable in SIMMAP and FRAGSTATS although they require some slight modifications according to the following expressions mps sme MES k pix PSSD A s PSSDOS pix IX ysm _ ELFO 10000 A EDS ED E L Aai 50 L 1 where ISMP and IFRC are respectively the values of the index J calculated by SIMMAP and FRAGSTATS and Apix is the area in hectares of the pixel which is used by FRAGSTATS to compute the values of those indices The rest of the indices that are calculated by SIMMAP are not computed by FRAGSTATS or vice versa 6 Limitations and known errors SIMMAP has been checked in detail in order to avoid errors Thus ho
8. de Ingenieros de Montes Universidad Polit cnica de Madrid Madrid Spain Users are asked to provide the author a brief description about the applications for which SIMMAP is used The objective of this manual is to briefly describe how to use SIMMAP The necessary information for understanding the effects of the simulation parameters on the MRC patterns is provided Also a concise description of the landscape pattern configuration indices that are used to quantify the spatial characteristics of the simulated patterns is included Further details about the MRC method and its results can be found in the paper by Saura and Martinez Millan 2000 and are out of the scope of this document SIMMAP simulations are very low computational time consuming In a standard PC at 333 MHz typical computational times are less than one second for 200x200 pixels patterns around 2 seconds for 400x400 images and around 4 seconds for 800x800 pixels landscapes SIMMAP patterns may be used for many purposes in different fields such as landscape ecology remote sensing spatial statistics simulation and computer graphics etc see section 7 The author hopes that this software is useful for your particular application and looks forward to hear about it SIMMAP is provided as is without warranty of any kind The user assumes all the responsibility for the accuracy and suitability of this program for a specific application Note that a point and not a
9. dex It is similar to MSI its minimum value is also 1 but uses patch area as a weighting factor because larger patches are assumed to have more importance for overall landscape structure i NP p i NP gja a Leva AWMSI gt r Sa 4 a i l i l PAFD Perimeter Area Fractal Dimension It derives from fractal theory It can be demonstrated that the perimeters p and areas a of a set of self similar shapes obey the following relation PAFD p k a 2 where k is a constant and PAFD is the Perimeter Area Fractal Dimension theoretically ranging from 1 to 2 of the set of similar shapes Assuming self similarity in the patches shapes and taking logarithms in both sides of this equation PAFD is estimated as twice the slope of the fitted line of perimeters p versus areas a of each of the patches of the class or landscape However the least squares regression can be done in two ways PAFD Inp k Ina 1 perimeter as dependent variable PAFD equals twice the slope of the regression line Ina k In PAFD 2 area as dependent variable PAFD equals twice the inverse of the slope of the regression line 18 Both expressions yield slightly different values for PAFD and there is not a special reason why one should be preferable to the other You may find PAFD calculated in any of these two ways depending on the author or the software used for its computation If the option Perimeter a
10. different value of the percolation threshold pec is obtained p 0 5928 for the default 4 neigbourhood criterion A neighbourhood criterion is defined by entering for each of the 8 neighbour cells the probabilities of being considered as belonging to the same cluster that the central pixel marked by an X in the figure below For example see below the definition of the 4 neighbourhood criteria the default used by SIMMAP Other symmetric neighbourhood criteria can be defined as is the case of the 8 neighbourhood criteria see below For this 8 neighbourhood criterion the percolation threshold pe is around 0 41 However changing to the 8 neighbourhood criteria does not provide a significant increase in the variety of the obtained patterns Saura 1998 However there are neighbourhood criteria that do provide different patterns than those that can be obtained with the 4 neighbourhood rule these are the asymmetrical neighbourhood criteria both 4 and 8 neighbourhood are symmetrical rules These asymmetric criteria generate patterns with anisotropy that is with patches oriented in a certain dominant direction e g the horizontal and the two diagonal criteria shown below 4 Neighbourhood 8 amp Neighbourhood Diagonal Diagonal Horizontal 10 When the neighbourhood criteria is changed also the percolation threshold pe changes and thus the value of p that generates a certain pattern fragmentation degree is also different In ge
11. ect the Window to pattern option the MRC patterns will be displayed in a window with the same number of pixels than the MRC patterns On the contrary the Pattern to window option will fit the MRC simulated patterns in a window of specified linear dimension which is entered in the edit box below the Pattern to window option This allows enlarging or reducing the original MRC patterns when they are displayed This may be also useful to display MRC patterns bigger than the screen size L 400 displayed in 400x400 pixels window 1 to 1 display scale L 400 displayed in 200x200 pixels window reduced WA a 12 L 200 displayed in 400x400 pixels window enlarged PS ere F F 1 T as L L 200 displayed in 200x200 pixels window 1 to 1 display scale he T aac oa If the Multiple windows option is enabled each new pattern that is generated is displayed in a new different display window Thus you get as many windows as simulations you make If your computer RAM memory is running out normally this should not be a problem or if you do not feel comfortable with too many windows disable this option each new pattern will be then displayed in the same display window and thus previous simulations will be lost 4 2 Saving the MRC patterns The MRC patterns generated by SIMMAP can be saved in image files bmp format which may be imported into other remote sensing GIS or image p
12. hus patterns are more fragmented As p increases the number of patches decreases and its mean and maximum size increase resulting in more aggregated patterns pik L h Te E RA a Ls Ly Tyla am wo As shown in the figure above the increase in pattern aggregation is not linear but more rapid as p is nearer a certain value the percolation threshold pe Dc 0 5928 for the default 4 neigbourhood criterion Other values of pe are obtained when different neighbourhood criteria are used however as will be described later for most simulations it is not necessary to change the neighbourhood criterion just use the default 4 neighbourhood In SIMMAP there is no need to use values of p bigger than pe All the variety of patterns can be obtained with p lt p by adequately fixing the simulation parameters values In fact when p gt pz a single patch tends to fill the entire pattern and then no control can be achieved about the spatial characteristics of the obtained patterns When p 0 a simple random map percolation map is obtained These patterns are not realistic representations of real world landscape patterns since they are much more fragmented than real patterns see figure above In general bigger values of p are those that will provide more realistic pattern simulations It is important to notice that in the MRC method the initial probability p is not related to the abundance of the classes but to their fragmentation degree
13. ing out of RAM memory if your computer is making an intensive use of the hard disk when generating the MRC patterns if enough RAM is available SIMMAP does not need at all to use your hard disk otherwise Windows may use the space in your hard disk to place the data that can not be located in your RAM memory This will cause simulations to slow down Several simulation and display parameters have limitations in their maximum values in SIMMAP 2 0 The limitations are the following Maximum pattern linear dimension L 2000 pixels Maximum minimum mapped unit m 99 pixels Maximum number of classes n 29 Maximum display window linear dimension 2000 pixels In what refers to the minimum mapped unit SIMMAP will be able to remove a maximum of 160 000 small patches patches smaller than the specified minimum mapped unit Only in some combinations of simulation parameters values which are not very reasonable this limitation may be exceeded For example if you simulate a pattern for p o0 and L 1500 and in addition you want to obtain m 90 there will be too many patches to remove from the original pattern in fact probably all the patches will be of size smaller than m and that s quite a lot If this happens SIMMAP will show a message box saying Too many patches smaller than m pixels to remove Requested pattern cannot be successfully simulated In this case it is suggested to decrease m or L and or increase p It is ex
14. ity of landscape pattern metrics to map spatial extent Photogrammetric Engineering and Remote Sensing 67 9 1027 1036
15. neral wider neighbourhood criteria produce more aggregated patterns for the same value of the initial probability p e g the 8 neighbourhood is a wider criteria than the 4 neighbourhood so pe is reduced and the range of p values interesting for the simulations is also shortened The opposite occurs when more restrictive criteria are used e g the Diagonal criterion in the figure above is more restrictive than the Diagonal one These effects are illustrated in the p values used in the simulated patterns in the figure below 4 neighbourhood p 0 55 8 neighbourhood p 0 37 To change the neighbourhood criterion click the N button in the SIMMAP 2 0 window You can just choose one of the common ones that are already predefined in SIMMAP the five criteria that are described and illustrated above by selecting the option Common criteria Or you can specify in detail the criterion by entering the neighbourhood probability values in each of the 8 adjacent cells selecting the option Other criteria 11 4 Display and save options 4 1 Displaying the MRC patterns These options located in the Display options panel of the SIMMAP 2 0 window control how the generated patterns are displayed in your computer screen The colours of the classes to be used for the display are selected in the Classes abundance and colours window after pressing the button in the Simulation parameters panel as described in section 3 2 If you sel
16. of each of the patches in the class or landscape 15 SMPS Squared Mean Patch Size expressed in pixels Is an alternative measure of central tendency that gives smaller weight to the smaller patches in the pattern SMPS where a is the size area in pixels of each of the patches in the class or landscape AWMPS Area Weighted Mean Patch Size expressed in pixels Like MPS or SMPS but giving even less weight to the small patches in the pattern when computing the mean size It is calculated according to the following expression i NP 2 a AWMPS Te PSSD Patch Size Standard Deviation The values of PSSD MPS and SMPS are related by the following simple expression SMPS MPS PSSD LPI Largest Patch Index Is the percent of the total class or landscape area occupied by the largest patch in the class or landscape NPI Number of Patches with Islands Islands are holes or inner patches embedded inside bigger ones NPI is just the total number of patches in the class or landscape which have islands inside them PPT Percent of Patches with Islands 16 PPI 100 NP IA Islands Area Is the sum of the areas of the island patches those that are embedded inside a bigger patch expressed as a percentage of the total pattern area At the class level JA sums the areas of the islands that are embedded in the patches of a given class not the areas of islands belonging to tha
17. pected that these maximum values are more than enough for the majority of the applications If not a version of SIMMAP with a higher maximum value for some of these parameters may be provided if possible to those interested 21 7 Where has been SUMMAP used SIMMAP and the modified random clusters method have been used as a key part of the analyses in the following papers published in international SCI journals Science Citation Index where you can find examples and ideas on the application of this software for various purposes Diaz Varela E R Marey P rez M F Alvarez Alvarez P 2009 Use of simulated and real data to identify heterogeneity domains in scale divergent forest landscapes Forest Ecology and Management 258 2490 2500 Peng J Wang Y Zhang Y Wu J Li W Li Y 2010 Evaluating the effectiveness of landscape metrics in quantifying spatial patterns Ecological Indicators 10 217 223 Shuangcheng L Qing C Jian P Yanglin W 2009 Indicating landscape fragmentation using L Z complexity Ecological Indicators 9 780 790 Hagen Zanker A 2009 An improved Fuzzy Kappa statistic that accounts for spatial autocorrelation International Journal of Geographical Information Science 23 61 73 Estrada Pe a A Acevedo P Ruiz Fons F Gort zar C de la Fuente J 2008 Evidence of the importance of host habitat use in predicting the dilution effect of wild boar for deer exposure to Anaplasma spp P
18. pefully no important problems should appear However it is possible that users find some bugs that were not detected before Help in reporting bugs is appreciated Some issues that may arise when using SIMMAP with non adequate display settings have been described in section 2 System and display requirements 20 Note that a point and not a comma should be set as your decimal separator in the regional configuration settings of your computer when using SIMMAP You may suffer from lack of RAM memory if you are generating very large patterns with large patches and your computer RAM is not very large e g only 16 MB If this is the case try not to use p values over the percolation threshold usually 0 5928 for the default 4 neighbourhood criterion if you are generating big MRC patterns big L especially considering that these values are not of particular interest in the MRC method Also if you generate too many patterns i e open too many display windows simultaneously you may have problems with your RAM memory even if the patterns are not really big however this should not happen frequently even in computers with only 16 MB RAM If so disable the option Multiple windows in the SIMMAP 2 0 main window this way only one display window will be presented in your screen showing the last pattern you generated and making less use of your RAM memory note that then previously generated patterns will be lost You can notice that you are runn
19. r vertically but not along the diagonals Many authors use this definition of patch although some others use the 8 neighbourhood rule indices values obtained with each of these two rules are not directly comparable The definition of patch affects many of the indices that are described below NP PD MPS SMPS AWMPS PSSD LPI NPI PPI IA MSI AWMSI PAFD PC although the calculation of some others does not require previous identification of patches on the pattern EL ED IEL IED Definition of perimeter SIMMAP defines perimeter as the length of the patch outer boundary So edges defined by small islands embedded inside the patch are not included in the definition of perimeter However many programmes that are commonly used for the computation of landscape indices e g FRAGSTATS do not differentiate between the inner edges and the true perimeter including both concepts in the computed perimeter Both definitions of perimeter provide different values of the perimeter 14 dependent indices that are described below in general higher values of MSI AWMSI PAFD and PC are obtained if inner edges are included Probably a more adequate measurement of pattern shapes is obtained when inner edges are not included in patch perimeters However for comparability purposes SIMMAP includes the possibility of obtaining the values of the indices corresponding to any of these two definitions Just click in the Include inner edges in perime
20. rocessing programmes if necessary Just press the button Save in the upper left part of the windows in which each of the generated patterns is displayed a save dialogue will appear where you can specify the location and name of the file where the corresponding pattern will be saved The option selected to display the MRC patterns has an influence in the characteristics of the saved image file When you save a pattern that has been generated with the Window to pattern option the resultant image file will have the same resolution number of pixels than the original MRC pattern This is highly recommended as you are 13 saving the real spatial information provided by the MRC method 1 1 scale On the contrary if the Pattern to window option is selected what you are saving is not the real MRC data but an enlargement or reduction as modified by the computer to show the pattern in a display window of a given size You should avoid this if further quantitative analysis not just for display or graphical purposes are to be done with the simulated MRC landscapes If you want to save in bmp files MRC patterns bigger than your screen size and still retain the original MRC spatial data just generate the simulations with the Window to pattern option selected you will not be able to see the entire obtained pattern in your screen but the Save button will always be visible in the upper left corner of the pattern window just press Save and you will s
21. s dependent variable in the Indices window is enabled PAFD is computed according to expression 1 if it is disabled then patch area is used as the dependent variable in the regression expression 2 The coefficient of determination of this regression R2 may be considered as an indicator of self similarity in the analysed pattern and is included in the indices window next to the PAFD value Values of R bigger than 0 9 are very frequently obtained in landscape patterns The value of R is not affected by which of the two previous expressions is used in the regression As estimated by regression techniques PAFD needs a sufficient number of patches in the pattern to be adequately estimated When values outside the theoretical range of variation for PAFD 1 to 2 or R 1 to 1 are obtained NV is displayed in the corresponding boxes in the Indices window 5 4 Other indices PC Patch Cohesion This index was developed by Nathan H Schumaker and according to the simulation model he developed it correlates better with animal populations dispersal success than any other of the commonly used landscape pattern indices It is calculated as i NP DP 1 PC 1 _ fis DP a i l where pi and a i are the perimeter and area of each of the patterns in the class or landscape and L is pattern linear dimension in pixels PC value is minimum PC 0 when all patches of habitat are confined to single isolated pixels an
22. t class embedded in other patches 5 2 Edge indices EL Edge Length expressed in pixel sides An edge is defined as a shared side between two pixels that belong to different classes Edges defined by map border are not included ED Edge Density Expressed as percentage of the maximum edge length that can appear in raster patterns of linear dimension L It is simple to demonstrate that 2 L L 1 is the maximum edge length that can appear in raster landscape data of linear dimension L EL ED 100 _ 2 L L 1 TEL Inner Edge Length expressed in pixel sides Inner edges are defined as those edges that are completely surrounded by pixels of the same class Thus it measures the presence of holes or islands in the patches in the pattern IED Inner Edge Density Like ED it is expressed as a percentage with respect to the maximum edge length that can appear in grid based data 2 L L 1 However EL gt TEL and in general values for IED much nearer to o than to 100 are clearly to be expected IED 00 E 2 L L 1 5 3 Shape indices The following three indices MSI AWMSI and PAFD intend to measure the complexity irregularity or elongation of the shapes in the pattern taking higher values the more convoluted and elongated the shapes are MSI Mean Shape Index Its minimum value for perfectly squared patches shapes is 1 17 i NP P us he P N AWMSI Area Weighted Mean Shape In
23. ter option in the Indices window to view the indices values according to the desired definition if this option is activated inner edges are also included in the patch perimeters All the indices are computed both at the class only patches that belong to a certain class are considered in the calculation of the indices and landscape level all patches in the pattern are considered independently of which class they belong to The indices values for each of the classes and for the entire simulated landscape are shown in the Indices window which is located by default at the right of the SIMMAP main window All these indices are calculated only if the option Indices located under the Simulate button in the lower right part of SIMMAP main window is activated One index is presented only at the class level this is PL percent of total pattern area occupied by a certain class 5 1 Number of patches and size indices NP Number of Patches in the class or landscape PD Patch Density Is a normalised way to express NP and is here calculated as PD 100 gt where NP is the number of patches in the class or landscape and L is the total number of pixels in the entire landscape L is also the maximum number of patches defined with the 4 neighbourhood rule that may appear in raster landscape data MPS Arithmetic Mean Patch Size expressed in pixels i NP a l MPS NP where ai is the size area in pixels
24. tore the entire generated pattern in the image file even if you can not see it completely in your screen 5 Spatial pattern indices Several landscape pattern configuration indices are computed on the MRC patterns This allows quantifying the spatial characteristics of the simulated landscapes as well as their comparison with real world pattern data It is not the purpose of the author to give a description of the background and behaviour of these pattern indices This is out of the scope of this manual and can be found in the landscape ecology and spatial pattern analysis literature where these indices are frequently used Only a brief description of the indices is provided so that no ambiguity exists about how they are calculated and what are they really measuring SIMMAP is itself a good tool to understand how these indices behave when varying pattern characteristics just change simulation parameters and see how indices vary Two important definitions affect how indices are calculated and thus the obtained values of the pattern indices Definition of patch A patch is defined in SIMMAP according to the 4 neighbourhood rule this has no relation with the neighbourhood criterion simulation parameter whatever the value of that simulation parameter the patches are always defined in the same way to compute the indices values This 4 neighbourhood rule considers as belonging to the same patch those pixels that are adjacent horizontal o
25. undance However if high values of the initial probability p are used big patches will be generated and it may be more problematic to obtain the desired classes abundances In particular if p gt pc a patch will tend to fill the landscape occupying for example 80 of total area In this case it will be impossible to obtain a 3 classes pattern with each of them occupying 33 of the area the class to which the big patch is assigned will occupy at least 80 Also if p is near to pe big patches will appear that may make difficult to obtain the desired abundances Consider for illustrative purposes a hypothetical pattern with 10 patches each of them occupying 10 of total area Suppose that the user wants class 1 to occupy 20 of total area Since classes are assigned randomly 1 10 or 2 20 or 3 30 patches etc may be assigned with a certain probability to class 1 and is not ensured that the desired 20 will be obtained in a particular MRC simulation These effects are less pronounced the smaller the percent of total area occupied by the individual patches in the pattern i e the lower p is and the bigger L is So if you need to obtain a more accurate class abundance increase L and or decrease p The actual class abundance that is obtained in the generated MRC patterns is shown in the Indices window in the index PL percent of pattern area occupied by each of the classes On the right of the Change abundances and colours window the
Download Pdf Manuals
Related Search