Manual Conefor 26

Transcript

1 CONEFOR 2. 6 User manual ( Saura & Torné, A pril 2012 ) the importance of habitat patches and links for maintaining or Quantifying enhancing landscape connectivity through spatial graphs and habitat availability (reachability) metrics Table of contents ... . ... 1 1. What is Conefor? ... ... ... ... ... ... ... 2. Authors 2 anual ... ... ... ... 2 3. Scope of this m ... ... .. 2 4. Key new features in version 2.6 compared to version 2.2 ... ... 3 5. Installation, operating system and computer configuration ... 6. Conditions of use ... ... . 3 ... ... ... ... 3 ... 7. Questions or further information ... 4 8. An overview of the new features in the graphical user interface for Conefor 2.6 IIC or PC ( intra , flux 9. Calculating node importance values including the fractions of connector ) ... 5 , 10. ... ... ... 7 Generalized Betweenness Centrality metrics 11. Calculating only the overall index values for the entire landscape ... ... 10 12. Calculating the Equivalent Connectivity ( EC ) for the entire landscape (overall index values) ... 11 13. Calculating the importance only for the added nodes ... ... 11 ... ... 12 14. Calculating link importance values for different modalities (removal, improvement, change) ... ... 17 15. Precision of the calculations (high or standard) ... ... ... 17 16. Other comments related to Conefor 2.6 ... ... ... 17. GIS extensions ... . 17 ... 18. Applications ... ... ... ... ... 18 ... ... ... ... 18 19. Empirical support 20. References ... ... ... ... ... 18 1. What is C onefor? Conefor is a software package that allows quantifying the importance of habitat areas and links for the maintenance or improvement of landscape connectivity , as well as evaluating the impact of changes on connectivity . It is co nceived as a tool for decision - habitat and land use in making support conservation and landscape planning , through the identification and prioritization of critical sites for ecological connectivity. Conefor includes new functional connectivity indices (integral index of conne ctivity (IIC) , probability of connectivity (PC) ) that have been shown to present an improved performance compared to other existing indices and to be particularly suited for landscape conservation planning and change monitoring applications (Pascual - Hortal and Saura 2006, Saura and Pascual - Hortal 2007, Saura and Rubio 2010, Saura et al. 2011a) . These indices are based on spatial graphs and on the concept of measuring habitat availability (reachability) at the landscape scale . This concept consists in

2 Conefor 2.6 user manual considering a habitat patch itself as a space where connectivity occurs, integrating the connected resources existing within the patches (intrapatch connectivity) with the resources made available by (reachable through) the connections with other habitat p atches in the landscape (interpatch connectivity). In this way, connectivity is conceived (and measured) as th at property of the landscape that determines the amount of reachable habitat in the landscape, no matter if such reachable and/or high quality habitat patches themselves (intrapatch connectivity), habitat comes from big from strong connections between different patches (interpatch connectivity) or, more frequently, from a combination of both. Authors 2 . y Santiago Sa ura and Josep Torné at Universidad Politécnica de Conefor 2.6 has been d eveloped b as an evolution of the previous 2.2 version (Saura and Torné 2009). See for Madrid www.conefor.org and section 20 for appropriate references Conefor S ensinode 2.2 (released in June . further details University of Lleida. Funding for developing Conefor has 2007) was developed by the same authors at 07140 - been provided by the Spanish Ministry of Science and Innovation through projects AGL2009 and REN2003 - 01628. Sensinode 1 .0 (LandGraphs package) was developed by Dean Urban (Duke University). 3 Scope of this manual . user manual only explains those This new features that have been implemented in Conefor 2.6 and and the basic functioning and were not included in version 2.2. All the other functionalities instructions for this software package are the same as for version 2.2. S t he manual of version 2.2 ee (attached) how to use Conefor . Later on we plan to for the rest of (most of) the information on ual integrating th e contents of these two manuals. produce a single updated man heck the Conefor C website ( ) for updates. In that website you can also subscribe to an email list that www.conefor.org ws regarding Conefor . will automatically notify you of any relevant ne . K 4 version 2.6 compared to version 2.2 ey new features in connector The three fractions of the (see  ) are computed dPC and dIIC indices ( intra , flux , 2.2. section 9 / dIIC values already provided by version dPC This , in addition to the total ) allows separately evaluating the different ways in which habitat patches can contribute to habitat connectivity an d availability in the landscape, i.e. their different roles as connectivity s (Saura and Rubio 2010) . provider  nectivity of individual links (sensilink) can be computed, in addition to The importance for con that was Several provided by version 2.2. the importance of nodes (sensinode) already sensilink functionalities are included in version 2.6, as described 4 (link removal, in section 1 link improvement, link change). Given that Conefor is now not just performing analyses but related to the nodes (sensinode) also to the links (sensilink) in the network , the name of the software package has been shortened and simplified from Conefor Sens inode ” to “ “ Conefor ” in this version 2.6 and future ones. IIC  The values of the Equivalent Connectivity ( EC ) corresponding to the and PC metrics (Saura ) . et al. 2011 a, 2011b ) are also presented in the results for the overall indices (see section 1 2 2

3 Conefor 2.6 user manual as BC ) metrics can now be calculated and presented values of Betweenness Centrality (  The the results in the node importance file. This includes the classical BC metric as defined part of and, more importantly, the generalized and by Freeman B C(IIC) and improved versions BC(PC) to match with the , which provide more ecological realism to this metric and allow it requirements and desirable properties of and PC (see section 10 ) . All these metrics are IIC integrated in a single analytical framework with the same uni ts of measurement , as described in detail by Bodin and Saura ( 2010). . Installation 5 operating system and computer configuration , This new version 2.6 has no particular installation requirements. Simply copy the executable file ( Conefor 26. exe) anywhere in your computer and double click the file to run the software (you can even run the software directly from a USB stick) . As for version 2.2, Conefor only runs in computers with a Windows operating system (any Windows version is fine, including Windows XP, V ista, 7 and others). I t runs both in 32 - bit and 64 - bit architectures, although the current Conefor compilation is only using 32 bits (i.e. it cannot make use of more than 4 GB RAM in your computer). In the future we plan to produce Conefor compilations for 64 bits and also for other operating systems different email Conefor from Windows (check the Conefor website for updates and /or for subscrib ing to the list if you want to be automatically notified of ). relevant news future versions or other Note that in t he input files for Conefor (node file and connection file) the point (and not the comma) be set as the decimal symbol , and that no symbol should be used as a thousand separator should ,235.45 or 1.234.45 are (e.g. a correct number format is 1235.45, while 1.234,45 or 1235,45 or 1 incorrect). Conefor will expect all the numerical values in the input files in accordance with these specifications , and will write the results in the same format. If any numerical value in the input files is not written according t an error will occur when trying to run Conefo r using those input o this format, files. T his means that the regional configuration settings of your computer should be set accordingly before generating the input files for Conefor through any of the GIS extensions that have been developed specifically for this purpose (see section 1 7 ). O therwise, the GIS extensions will write the subsequent Conefor input files with the wrong numerical format and th ese files will not be usable for processing Conefor. in 6 Conditions of use . Conefor is freeware and has the same conditions of use as version 2.2. This means that it can b e used with no res trictions for non - commercial purposes with the only c ondition of citing the software, its website and the appropriate references ( see www.conefor.org a nd section 20 below ). 7 Questions or further information . For any question or further information please visit www.conefor.org , where you can find the most recent versions of the Conefor software package, the user manuals, GIS extensions, related papers, an overview of the applications in which Conefor has been applied around th e world, empirical support studies and other relevant information. You can also contact us at [email protected] . The authors would appreciate hearing about the applications in which Conefor is used, as well as help in will be reporting bugs and suggestions for improvement. However, very limited user support provided and only for specific questions regarding this software and the methods implemented in it 3

4 Conefor 2.6 user manual the manual s that cannot be solved by other means (e.g. by carefully reading and the different related papers). 8 . An overview of the new features in the graphical user interface for Conefor 2.6 Most of the new features in Conefor 2.6 are located within t he red boxes indicated in the image next to each red box indicate in which below. T he rest is exactly a s in version 2.2. The numbers See section of this user manual you can find the description of that particular option or functionality. also section 1 2 not indicated graphically in this image. for an additional new feature 10 1 4 9 1 1 1 3 5 1 4

5 Conefor 2.6 user manual . Calculating node importance values including the fractions of IIC or PC (intra, flux, connector) 9 the section 1 below), Conefor will calculate the Unless you specify the “only overall index” option (see 1 importance of every node as the decrease in the connectivity metric value caused by the removal of that individual node from the landscape (see manual for version 2.2 and related papers for further details) des the intra , flux and connector fractions for IIC or PC as described by Saura and . This inclu Rubio ( 2010) ; these fractions will be automatically calculated if any of these two metrics is selected for the analysis . Th e importance of a node (or link , see section 14 ) according to a given connectivity index (metric) M can be expressed in relative terms (“deltas” in Conefor 2.6, M ) or in absolute terms (“vars” in d Conefor 2.6, var M ): ( ) M is the overall connectivity index (metric) value when all the nodes are present in the Where ( initial, intact , undisturbed landscape) and i.e. the metric value for the M landscape is the overall after individual node from the landscape index value after the removal of that . Th erefore, dM and varM quantify the relative and absolute variation in the overall connectivity metric value for respectively the whole landscape ( M ) after the loss of a particular n ode /patch . The same equations apply for other changes related to the links 4 . between patches, as will be described later in section 1 You should select at least “Show deltas” or “Show vars” before running the software (unless the option “Only overall index” is activated, see below in section 11 ). If you select the option “Show values for each node and for d eltas” (this is the default), Conefor will calculate and show the dM each of the indices selected for analysis. If you select the option “Show vars”, Conefor will calculate and show the varM values for each node and for each of the indices selected for analysis. You can dM also select both “Show deltas” and “Show vars” and then Conefor will calculate and show both varM values. Note , however , that this may generate a large number of columns in the node and at the ranking of habitat patches (nodes) by their contribution to landscape importance file, and th connectivity is the same according to either dM or varM (since both variables are proportional, as given by varM =( dM · M )/100). Although in some cases dM values are easier to interp ret, this will d selected indices and user preferences. In some cases, if the landscape/network is very epend on the large, dM values can be very low (it is hard for a single node to have a large relative importance for connectivity if the network is made u p by many thousands of nodes) and hard to handle (eventually indistinguishable from zero). In these cases varM values might be a preferable option. The index M can correspond to any of the following connectivity metrics included in Conefor 2.6: NL , PC NC , H , LCP , IIC . , F , AWF , num num metrics the IIC , respectively, and PC in Conefor values (numerators of the IIC and PC Note that num num varIIC ) are used for computing the dIIC , see equations below , those and varPC values. This makes dPC unnecessary specifying a ny A Even value for calculating the node importances according to IIC or PC . L PC IIC and PC values for a particular A ) to calculate the node were used (instead of IIC if the or num L num importance values, dIIC and dPC would be exactly the same as those obtained using IIC PC or num num A is a constant that does not vary by the removal of any node or link respectively . This is because L 5

6 Conefor 2.6 user manual varIIC varPC values as they would resul t from using from the landscape. If you wish to obtain the or and PC values (for a given A value) you just need to divide the varIIC or varPC the values that are IIC L 2 A . calculated by Conefor 2.6 by L ∑ ∑ ) ( ∑ ∑ ∑ ∑ ∑ ∑ - Hortal and Saura 2006, Saura and See manual for version 2.2 and the two related papers (Pascual Pascual - Hortal 2007) for further det ails on these metrics and formulas. metrics can be PC As shown by Saura and Rubio (2010) the importance values for the IIC and partitioned in three different fractions ( , flux , connecto r) considering the different ways in which intra (node) overall habitat connectivity and availability in the a habitat patch or link can contribute to scape. Conefor 2.6 will automatically calculate for each node the values of these three fractions land d IIC separately whenever PC are selected for analysis. This applies t o both the M and var M and/or values calculated by Conefor as follows (the total dIIC , varIIC , dPC , and varPC values will also be presented, in addition to the partitioned : fraction values) The three fractions will be calculated both for the existing nodes that can be lost (standard analysis / node removal) and for the nodes that may be added in the landscape as a result of habitat version 2.2). While a habitat patch restoration measures (see option “nodes to add” in manual for through all these three fractions, link s (corridors, connectors) can only can contribute to connectivity fraction: see Saura and Rubio (2010) for details. connector contribute through the Baranyi et al. (2011) have sho wn, through cluster analysis and multidimensional scaling, that the , flux and connector fractions provide a non - redundant and complementary information on the intra they capture most of the variability importance of a patch in a landscape network, and that in patch conservation priorities that would derive from using a much larger set of connectivity metrics (see figure 5 in that study), perhaps complemented by a centrality metric (see section 10 below). e dM The only exception for th varM calculations i n the node importance file (as explained and above) are the values of the Betweenness Centrality metrics ( BC , BC(IIC) , BC(PC) ): these metrics are calculated in a different way (see section 10 below), and not according to the formulas shown above The Betweenness Centrality metrics included in Conefor 2.6 are calculated for each for d M or var M . node just according to their position in the intact (initial) landscape, and not following a node (patch) removal procedure as for the rest of the metrics. 6

7 Conefor 2.6 user manual uld be noted that the CCP index (Class Coincidence Probability) is no longer included in Finally, it sho . This is because CCP does not present good prioritization abilities compared to other Conefor 2.6 ( an excessive number of metrics ) ) and we intend to avoid metrics (see Pascual - 2006 Hortal and Saura it is quite similar CCP being calculated by Conefor if this is not necessary . On the other hand, analytically to LCP . Since LCP is included in Conefor 2.6, it is very easy to get the CCP values from the calculated by LCP ones see the formulas in Pascual - Hortal and Saura Conefor 2.6 if you wish to do so; . (2006) to understand how to make this simple calculation In the rare case that you still want to get .2 for that particular the CCP values directly, you can use the older version Conefor Sensinode 2 purpose (this older version will be available for download in the Conefor website). still 10 . Generalized Betweenness Centrality metrics .1. Which network centrality metrics are calculated by Conefor? 10 for each node Conefor 2.6 calculates : - BC , the classical Betweenness Centrality metric as originally defined by Freeman (1977; Sociometry – 41) 40: 35 . metric that were proposed by Bodin and Saura BC(IIC) and BC(PC) BC - , the improved versions of the in order to incorporat e some considerations that are important to increase the ecological (2010) IIC realism of this metric and to make it match with the characteristics and desirable properties of and PC In this way, BC(IIC) and BC(PC) are placed (integrated) within the same analyti cal framework . . All these metrics as the PC metrics and are expressed in the same units of measurement and can IIC values for each node are be directly compared. In particular, the equal normalized BC(IIC) and BC(PC) section 10 or fraction of the IIC and PC metrics respectively (see connector .5 below ), higher than the Bodin and Saura (2010). as shown by BC , BC(IIC) , and BC(PC) are calculated only for the nodes that exist in the landscape The values of for the links or the nodes to add. (standard node importance analysis), and not 10 .2. What do the Betweenness Centrality metrics measure and how are they calculated? BC , , BC(PC) ) is the same : The baseline concept for all these centrality metrics ( they measure BC(IIC) how much a specific node sits between all other pairs of nodes in a network, i.e. the se metrics capture the degree to which the shortest or optimal paths for movement between other habitat patches pass through a particular node . A node will be highly central (as quantified through thes e metrics) when it is involved in the shortest/optimal movement route s between many other pairs of patches by serving as an intermediate stepping stone patch . These Betweenness Centrality metrics differ from all the other metrics implemented in Conefor in that they can only be calculated at the level of individual nodes, i.e. these BC metrics do not provide an overall value characterizing the connectivity of the entire landscape . T here is no M landscape - you select these centrality level value (see equations in section 9 ) for these me trics. Therefore, when metrics for calculation their values will only appear in the node importance file and not in the results for the overall index values. 7

8 Conefor 2.6 user manual he values of the centrality metrics for individual nodes (as shown i n the node importance file) are T from the rest of the metrics. The centrality values for an individual node are calculated differently obtained taking into account the topological position of that node in the intact (initial) landscape, removal procedure (the formula for dM without performing a varM ny in section 9 above node or does not apply for these metrics). See Bodin and Saura (2010) for further details. Therefore, even if , BC(IIC) and BC(PC) are shown in the node importance file togeth the values of dM or BC er with the varM values for the rest of the metrics, it should be kept in mind that they are calculated using a for the others. /procedure different approach than BC(PC) metrics? 10 .3. Which is the difference between the classical BC and the generalized BC(IIC) and s the number of shortest paths between all pair of patches that go through a BC only consider particular node, regardless of the area (or attributes) of the patches being connected and of the BC(IIC) and length or strength of those paths. (PC) , however , take into account the area ( or any BC other attribute) of the patches that are being connected through a particular node, considering more central those nodes that serve as stepping stones between large and high quality patches than those that poor habitat resources. BC(IIC) also takes into account the sit in between patches with scarce or longer ; length (number of links) of the paths between patches in which a particular node is involved they are given ve movements, and therefore paths are considered less feasible for conducting effecti takes into account the probability of less weight in the centrality calculations. Similarly, BC(PC) * ( maximum product probability p , see Pascual Hortal and Saura movement through a particular path - ij * 2007) so that those paths with higher p ) , are given more weight in the centrality calculations. See ( ij Bodin and Saura (2010) for further details. Not BC(IIC) and e that are the most computationally intensive of all the metrics implemented BC(PC) in Conefor. You s hould therefore be cautious when trying to calculate these two metrics in large in very large BC(PC) landscape s with many nodes. It might be u nfeasible to compute BC(IIC) or networks due to the excessive computational time that would be ard BC metric required. The stand can however be calculated much faster. .4. How to select the BC, BC(IIC) and BC(PC) metrics for calculation in the Conefor interface ? 10 If you wish to calculate the classical metric, then you should s elect BC in the box for the binary BC connectivity indices in the Conefor interface. The resultant values for each node will be shown in one of the columns of the node importance file. The BC value calculated by Conefor for a particular node ) k correspond to the sum of all separate shortest pat hs between all pair s of patches ( different from k that go through , divided by the total number of shortest paths between all k pair s of patches (equation 5 in Bodin and Saura (2010)). No matter if you select “Show deltas” or “Show vars” the same result fo r will appear in the node importance file (even if you select both “Show deltas” and BC BC in the node importance file, with the values “Show vars”, only one column will be produced for calculated as described above) . BC(IIC) If you wish to calculate the etric, then you should select both BC and IIC in the box for the m binary connectivity indices in the Conefor interface. Conefor will understand that you want to calculate not only BC and IIC , but also BC(IIC) . This will result in the values for BC , IIC and BC(IIC) being ) will be shown in different columns in the node importance file. The three metrics ( , IIC and BC(IIC) BC calculated even if you are interested only in BC(IIC) , but this is not of much concern because BC(IIC) is BC(IIC) will consume most the most computationally intensi ve metric among these three. Calculating 8

9 Conefor 2.6 user manual BC of the total processing time. On the other hand, some of the calculations performed to get the values are also needed to obtain the results for . If you wish to calculate BC and IIC but and IIC BC(IIC) then you should run Conefor two times, one with only BC selected and the other with only not BC(IIC) selected. IIC in the box for the binary BC If you wish to calculate the BC(PC) metric, then you should select both in the box for the probabilistic indices. Conefor will understand that you want to indices and PC BC and calculate not only , but also BC(PC) . This will result in the values for BC , PC and BC(PC) being PC shown in different columns in the node importance file. The three metrics ( , PC and BC(PC) ) will be BC BC(PC) BC(PC) calculated even if you are interested only in is , but this is not of much concern because BC(PC) the most computationally intensive metric among these three. Calculating will consume most of the total processing time. On the other hand, some of the calculations performed to get the BC and PC values are also needed to obtain the results for BC(PC) . Note that if your input connection file is a distance file, when selecting BC PC you will need to provide a distance threshold that will be and BC (as for any other binary connectivity metric) and a pair of probability - distance used for calculating PC and BC(PC) . The distance threshold value specif ied in the values that will be used for calculating box for the binary indices does not affect at all the calculations for , which are just based on BC(PC) - the probability distance values specified for the probabilistic indices. In the same way, the BC calculations for he probability - distance values that are specified for the are not at all affected by t probabilistic indices. The same applies if your input connection file is a probability file. In this case a probability threshold will be requested for calculating BC but this will not affect at all the calculations for PC or BC(PC) , which will run using directly the probability values in the connection file without requesting any additional distance or probability value in the probabilistic indices box. See the manual for version 2.2 if you are not familiar with these latter considerations. If you wish to calculate BC and PC but not BC(PC) then you should run Conefor two times, one with only BC selected and the other with only PC selected. 10 .5. How do the generalized Betweenness Centrality metrics r elate to the connector fraction of IIC How should this relationship be interpreted? and PC? If you select “Show deltas”, the BC(IIC) and BC(PC) values will be scaled in the same way as for dIIC dPCconnector and ( dIICconnector and ), and the resultant values for each node will be shown in dPC two columns with names dBC(IIC) and dBC(PC) . In this case , dBC(IIC) ≥ dIICconnector and dBC(PC) ≥ that the names dPCconnector , however , Note dBC(IIC) and dBC(PC) just indicate that the values are . directly comparable with those for dIICconnector and dPCconnector . It does not mean that the dBC(IIC) and dBC(PC) values are calculated as the relative variation in any connectivity metric value following any patch removal procedure. . S e e Bodin and Saura (2010) for details and equations If you select “Show v ars”, the and BC(PC) values will be scaled in the same way as for varIIC BC(IIC) varPC varIICconnector and varPCconnector ), and the resultant values for each node and will be ( varBC(IIC varBC(IIC) and varBC(PC) . In this case shown in two columns with names ) ≥ varIICconnector and varBC(PC) ≥ varPCconnector . Note however that the names varBC(IIC) and varBC(PC) just indicate that the values are directly comparable with those for and varPCconnector . It does varIICconnector not mean that dBC(IIC) and dBC(PC) values are calculated as the absolute variation in any connectivity metric value following any patch removal procedure. If you select both “Show deltas” and “Show vars”, then the values for BC(IIC) and BC(PC) in the node importance file will be varBC(PC) . respective ly scaled and shown both as dBC(IIC) and dBC(PC) and as varBC(IIC) and 9

10 Conefor 2.6 user manual how the flows might The centrality metrics are calculated in the intact landscape and do not consider patch. Unlike the conne ctor fraction of the IIC and PC change as a consequence of losing a particular do not deliver estimates of the impacts of a patch removal in terms of metrics, they connectivity loss. k dBC(IIC) and values for a given node should be interpreted in the The pairs of dIICconnector s to the pairs of values varBC(IIC) and varIICconnector , dBC(PC) following way (exactly th e same applie , and dPCconnector and varPCconnector ) : dBC(IIC) quantifies the amount of fluxes that and varBC(PC) in the intact landscape (the undisturbed landscape in which no habitat are expected to go through k is part of the best (shortest or most probable) paths between other because patch has been lost), k does not quantify how much the connectivity betwe en other habitat patches. However, dBC(IIC) s on the presence of k habitat areas depend in the landscape, i.e. it does not measure how irreplaceable is as a connecting element between other habitat areas. This latter aspect is what is k dIICconnector . It might happen that even if k is considerably involved in the fluxes measured by ), k dBC(IIC) does not have a large impact in the the loss of occurring in the current landscape (high , much smaller than dB(IIC) ), dIICconnector connectivity between the other habitat areas (low the connect ivity that was provide d by k as a connecting element or stepping stone because has been largely or fully compensated by other patches and alternative paths in the available for movement k however , , is the only element sustaining the connectivity betw een other habitat landscape. If alternative patches or paths available to compensate for its loss, the areas, and there are no other k removal of will have a large impact in the connectivity of the remnant network ( dIICconnector will dBC(IIC) be in this case as high as . Therefore, these metrics capture how much of the current ) stepping stone (connecting) role played by a particular patch in the intact landscape ( dBC(IIC) ) is lost (2010) for See Bodin and Saura when that patch is removed from the landscape ( dIICconnector ). further details . on these metrics and their ecological interpretation or ) or can be considered as a “fourth fraction” of the IIC PC metrics, being measured BC(IIC BC(PC) int ra , with the same units and within the same analytical framework than the and connector flux fractions (a common currency for connectivity). These four metrics/fractions provide a multifaceted, complete and non - redundant view of the different ways in which habitat patches can be important as connectivity providers, as suppor ted both in analytical grounds (Saura and Rubio 2010, Bodin and Saura 2010) and by statistical assessments on their practical outcomes as compared to a wider set of connectivity metrics, as shown by figure 5 and the rest of the content in Baranyi et al. (2 011). values for the entire landscape 11 nly the overall index Calculating o . The option “Only overall index” (first option in the “Mode” box) should be selected when you are only interested in obtaining the overall index value s for the whole landscape ( M ) and not in the ). importance values for each individual node ( d M or varM or link This option allows the user to save a lot of processing time when the interest is only in the M value. is much faster to compute than the This is because t he overall index value M d M or varM values for every node . O M requires just one calculation of the metric value, while obtaining dM or btaining n varM additional calculations of the metric value (one for each node), where n is the total requires number of nodes in the landscape uch more calculations of the metric value (in general n ·( n - 1)/2) . M are required for obtaining dM or varM if any of the link importance analysi s options if selected ( see section 1 below) . 4 Obviously , if th (node or link importance nly o verall index” is selected the dM or varM values e “O will not be calculated. As stated in the previous section, the results for the overall index values files) 10

11 Conefor 2.6 user manual not contain any value for the etweenness C entrality metrics even if th ese have been selected will B for calculation. . 12 Calculating the Equivalent Connectivity (EC) for the entire landscape (overall index values) or and go for the “Results - > Overall index values”, you will notice that in this IIC If you compute PC addition to the IIC or PC (or IIC ) or PC new version, in values already provided by version 2.2, num num the (Equivalent Connectivity) values for these two indices : EC(IIC) and results also include the EC ECA . This corresponds to the Equivalent Connected Area ( ) index as described in Saura et al. EC(PC) ), where the patch habitat area was used as the node attr ibute. ECA (IIC) (2011 ECA(PC) a, 2011b and defined as the size of a single habitat patch (maximally connected) that would provide the same are the IIC value of PC metric (respectively) as the actual habitat pattern in the landscape. In a more and general case in which the attributes of the nodes might correspond to some other characteristic .) this index is better different from just habitat area (e.g., habitat quality, population size, etc (Equivalent Connectivity) . Therefore this more general name ( EC ) is the one used in renamed as EC Conefor 2.6. or ECA is just computed as the square root of the numerator of the IIC and PC indices ( IIC EC and num , ), yielding IIC) and EC(PC) EC( respectively. Although Saura et al. (2011a, 2011b) only used PC num / EC for the PC index ( EC(PC) ), the same approach and advantages (see below) apply for IIC ECA PC or IIC ( EC( IIC ) ). Both EC( PC ) and EC( IIC ) are calculated by this new version. You just need to select for analysis, and the results for the overall index values will also show the EC(IIC) and EC(PC) values, respectively. and EC(PC) maintain all the desirable properties and good prioritization abilities of IIC and PC EC(IIC) : (a) they avoid (as an overall index value) compared to IIC and PC but have the following advantages the very low metric values that might be obtained for IIC and PC when the amount of habitat is very small compared to the total extent of the analyzed landscape, since EC(IIC) and EC(PC) will not be smaller than the largest attribute (e.g. habitat area) for the patches in the landscape ; (b) EC(IIC) and have the same units as the attributes of the nodes, which facilitates the interpretation and EC(PC) understanding of the resultant values ; (c) EC(IIC) and EC(PC) avoid the need of specifying any A L for value, as is needed IIC be to and PC , which in some cases might not be straightforward and might some degree arbitrary ; and, more importantly (d) the relative variation in EC(IIC) or EC(PC) after a can be direc tly compared with the particular spatial change (or set of changes) in the landscape amount of habitat area in the landscape variation in the total (or any other considered node attribute) after the same change, with an easy interpretation and additional insights that can be gained from that comparison (Saura et al. 2011a, 2011b) . This latter property makes EC(IIC) and EC(PC) quite convenient for quantifying changes in landscape connectivity and comparing them with the changes in the amount of habitat in the landscape, as described in Saura et al. (2011 a, 2011b ). See figure 1 in Saura et al. (2011a) and figure 1 in Saura et al. (2011b). 1 3 . Calculating the importance o nly for the add ed nodes When th e option “Only added nodes” is selected , Conefor will make the calculations of the dM or values only for the potential nodes that may be added in the landscape (as a result of potential varM t for the nodes that already exist habitat restoration actions that may improve connectivity), and no 11

12 Conefor 2.6 user manual Selecting “Only nodes in it (the impact on connectivity of their potential loss will not be evaluated). to add” will also deactivate any of the options for link importance analysis. This option (“Only added nodes”) act ive only when the option “There are nodes to add” has been is previously selected in the box for the node file. When “There are nodes to add” has been selected (see section 4. 2 in the manual for version 2. 2) and Conefor will expect a node file with three columns calculate (a) the will or varM values representing the connectivity loss that would be in general dM dM caused by the removal of each of the nodes that exist in the current/initial landscape, and (b) the values representing the connectivity improvement that would occur thanks to the addition or varM of new nodes in the landscape as resulting from habitat restoration measures (see manual for version 2.2 for details). If the option “Only added nodes” is selected Conefor will just calculate (b), i.e. it will only evaluate how much each of the new potent ial habitat areas (nodes to add) would hey were added in the landscape. Since , in general contribute to improve connectivity if t the , number of candidate nodes to be added in the landscape (b) is much smaller than all the , this option can save a lot of processing time by just calculating the nodes/patches that exist in it (a) or varM values for the nodes to add and not for the existing nodes. Note that BC , BC(IIC) dM and metrics are not calculated for the nodes to add; the values for these metrics wi ll be equal to BC(PC) zero in the node importance file, but this just means that they have not been calculated . 1 . Calculating link importance values for different modalities (removal, improvement, change) 4 1 .1. What does “link importance” mean and which results are obtained through this analysis ? 4 Conefor 2.6 includes the possibility of calculating the contribution of individual links to maintain or improve overall landscape connectivity. This goes beyond the capabilities of previous Conefor ver sions, in which such type of analysis was possible only for nodes and not for links. Some recent in Conefor new (see the section on applications at the studies have benefited from this functionality Conefor website for the full references): Gurrutxaga et a Landscape and Urban Planning 101: ; l. (2011 - 320 ), Saura et al. (2011; Forest Ecology and Management 262: 150 - 160 ), Carranza et al. (2012; 310 - 290 ), Rubio et al. (in press ; Landscape Ecology 27: 281 Forest Systems). The importance of a link for maintaining or improving connectivity is calculated in the same way as for the nodes (see section ), i.e. as the relative ( dM ) or absolute ( varM ) variation in the value of a 9 given connectivity metric M after a certain change affecting one of the links in the landscape. Note that links can only contribute to habitat connectivity and availability (as measured by IIC and hat contains no PC ) through the connector fraction. Since a link is defined as a connecting element t contribute through the intra habitat area, a link cannot n, a link cannot . For the same reaso fraction be the final/ intra and the flux permanent destination for any dispersal flux. Therefore both the fractions will be by definition equal to zero for any link in the landscape. If a connecting element contains some habitat area, this should be modeled as a node in the graph, and its value as a connecting element or stepping stone will be anyway quantified through the connector fraction of or the I IC PC metrics, together with its potential contribution through the intra and flux fractions. The connector values for links can be directly compared with the connector values for nodes. See Saura and Rubio (2010) for details. In summary, even when Cone for will express the results of the link (depending on the metric and type of result selected importance values as dIIC , varIIC , dPC and varPC in the Conefor interface), the user should be aware that for links these values correspond in fact 12

13 Conefor 2.6 user manual the fraction, that is, they correspond to dIICconnector , varIICconnector , exclusively to connector and dPCconnector respectively, even if this is not written explicitly in the link varPCconnector importance result files. Note also that the link importance analysis can be much more time consuming than the node n +1 calculations of the metric importance analysis. While a node importance analysis will require ) in a graph with n nodes, a link importance analysis in the same graph will require n ( n - 1)/2 + value ( M ( 1 calculations of the metric value (since - 1)/2 is the number of links in an undirected complete n n , especially n nodes). This can be considerably slow and even unfeasible in large networks graph with intensive metrics like PC . Users should keep this in mind when trying to analyze for computationally 2.6 . Some o ptions that might allow users to large datasets through this new functionality in Conefor reduce the number of calculations and the required processing time are be ing evaluated and might be included in a future version of this software package (check for updates or www.conefor.org ). subscribe there to the Conefor email list if desired The link importance analysis in Conefor 2.6 does not include the calculation of an y of the Betweenness Centrality metrics (see section ), which are only calculated for individual nodes. 10 The link importance analysis can be performed under the following three different modalities that “Link importance” box in the Conefor interface can be selected in the link removal, link : ement and link change. improv 4 .2. Link removal: the impact of losing an existing link. 1 Link removal When the “ option is selected, Conefor will remove one by one each of the links ” existing in the landscape network (individual link removal, only one link removed at the same time) r that link loss on landscape connectivity according to dM and calculate the impact o of varM (where M can be any of the metrics implemented in Conefor except the Betweenness Centrality metrics , are the ones recommended for this and other although analyses). This will types of IIC and PC produce a dM or varM value for all the pairs of patche s/nodes in the landscape , each link being . represented by the IDs of the nodes it connects IIC ) those patches that are not linked in the In the binary connection model (the one that applies for initial/intact landscape will surely have =0 and varM =0 (th ere is obviously no impact in the dM potential loss of a link between two patches if such link does not actually exist in the initial landscape). For those pairs of patches that do have a link in the initial landscape, Conefor will , which will result in a the index value after removing that link ( M , see section 9 ) recalculate dM after varM that might be (although not necessarily) higher than zero. and For the same reason as above, i n the probabilistic connection model (the one that applies for PC ), those pairs of patches that have no direct connection between them ( p =0 =0) will surely have dM ij and =0. For the rest of the cases (pairs of patches with p varM >0), Conefor will change the p value ij ij initial for each pair of patches from its cape to p =0, and recalculate the value value in the intact lands ij of M after that change ( M ), which will result in a dM and varM that might be higher than zero. after of change they In the binary connection model all the link losses are comparable in the magnitude represent (a link that existed in the initial landscape is removed completely to evaluate the impact of its loss). However, in the probabilistic model the initial value of p for each link in the initial landscape ij might be very different among the different links. This introduces some complication when trying to 13

14 Conefor 2.6 user manual dM varM values resulting from this analysis for the different links in the directly compare the or ability of use of probabilistic model , since the actual magnitude of change in the strength or prob ) might be quite different for each link. For example, it is not the each link (as characterized by p ij dM p =0.01 in the initial landscape than obtaining the same same obtaining =10% for a link which had ij p dM =10% for another link that had a =0.9 in the initial landscape. This has to be taken into account ij when trying to interpret and summarize the results of this link removal analysis for the probabilistic connection model. If the “Link removal” opti there is the possibility to use the “Reduce calculations” on is selected, option (see the box for the link importance options in the Conefor interface). Such “Reduce calculations” option allows specifying a maximum distance (if the connection file is a distance file) or the connection file is a probability file) so that the calculations are a minimum probability (if performed only for the pairs of patches (links) with a distance not larger than the specified maximum p is case the rest of the not smalle r than the specified minimum. or with a probability Note that in th ij dM =0 and =0 because links (those weak links with large distances or small probabilities) will get varM would be the results they have not been evaluated in the analysis; this does not mean that tions” option would not have been selected. necessarily zero if the “Reduce calcula .3. Link improvement: the potential benefits of strengthening connections between habitat patches 1 4 ” is selected, Conefor will perform quite the opposite analysis to Link improvement When the option “ that of the “Link removal” option. In the binary connection model (the one that applies for ), Conefor will add one link (only one at a IIC time) to each of those pairs of patches that are n will ot directly linked in the initial landscape, and ). This may result in a connectivity gain as M value after that change /addition ( recalculate the M after measured by metric M , case in which dM <0 and varM <0. Note that the relative and absolute values negative ) will variations of the IIC and PC metric s ( dM and varM in this case because have varM M is higher than M (see the formula for dM and in section 9 ) , but in fact these negative after values mean that connectivity has improved. Obviously, for those pair s of patches that were already linked in the initial landscape, the connectivity cannot increase by adding a link that in fact already existed, and therefore dM =0 and varM =0 . In the probabilistic connection model (the one that applies for ), Conefor will calculate the PC potential (positive) impacts of improving as much as possible the direct connection (link) between each pair of patches (only one at a time). In this probabilistic model this translates in assigning a p =1 ij to each pair of patches, which means that the strength or frequency of use of the direct connection ) will be improved for all the pairs of patches except between i and j (as quantified by p those that for ij already had p =1 in the initial landscape. F or those pairs of patches with p =1 in the initial landscape ij ij dM =0 and varM the result will necessarily be p value for each link =0. For the rest, the increase in the ij may result (although not necessarily) in an improvement in connectivity as quantified by dM or varM (as said above, will correspond to negative dM and varM values) . such improvement In the binary connection model all the link gains are comparable in the magnitude of change they represent (a link that did not exist in the initial landscape is added to evaluate the potential benefits of its creation or restoration). However, in the probabilistic model the initial value of p for each link ij in the initial landscape might be very variable among the different links. This introduces some varM values resulting from this analysis for complication when trying to directly compare the dM or 14

15 Conefor 2.6 user manual in the probabilistic mode , since the actual magnitude of change in the strength or the different links l p ) might be quite different for each link. For probability of use of each link (as characterized by ij =10% for a link which had example, it is not the same obtaining dM =0.01 in the initial lands cape p ij dM =10% for another link that had a than obtaining the same =0.9 in the initial landscape. This has p ij link improvement to be taken into account when trying to interpret and summarize the results of this analysis for the probabilistic connection model. Note that not all the link changes that are set by the “Link improvement” option can be considered realistic. For example, if the dispersal abilities of your analyzed species are in the range of a few onsider that hundred meters, it would be certainly unrealistic to c can improve up to 1 for two p ij patches that are separated by hundreds of kilometers, no matter how much increase in the permeability or hospitability of the landscape matrix you may be able to achieve. However, the link hypothetical (and unrealistic) improvement in the strength or improvement option wil l evaluate such feasibility of use of that link (as characterized by ) in the same way as for all the other pairs of p ij patches. The resultant dM or varM for a values may be therefore only feasible to obtain in reality graph small subset of the total number of links (direct connections) in your . To overcome this issue and perform a more fine tuned and detailed analysis related to the potential changes in the links in - your landscape you might cons ider instead the “Link change” modality that is described in the next sub section . Finally, in the same way as described above for the “Link removal” analysis, i f the “Link the possibility to use the “Reduce calculations” option ” option is selected, there is improvement also (see the box for the link importance options in the Conefor interface). Such “Reduce calculations” option allows specifying a maximum distance (if the connection file is a distance file) or a minimum probability (if the connection file is a probability file) so that the calculations are performed only for the pairs of patches (links) with a distance not larger than the specified maximum or with a p not smaller than the specified minimum. Note that in this case the rest of the links probability ij dM =0 and varM =0 because they (those weak links with large distances or small probabilities) will get they would be zero if the “Reduce have not been evaluated in the analysis; this does not mean that calculation s” option would not have been selected. 1 4 .4. Link change: how user - defined changes in the links translate in connectivity gains or losses This is the most powerful and flexible modality for link importance analysis. It however requires additional input information to be provided by the user compared to the link removal and link improvement options. If the option “Link change” is selected, Conefor will require the connection file to have four columns s the only processing option in which the instead of three as in the usual case (in fact , this i connection file needs to have four instead of three columns). The first three columns are the same as in any connection file , and a value (typically some form of j for Conefor : ID of node i , ID of node distance or a probability p in ) characterizing the connection between nodes i direct dispersal and j ij the initial landscape (see manual for version 2.2 for further details). The fourth column required by nge” option the “Link cha will also contain a value characterizing the connection between nodes i and j , but this value will correspond to the new distance or direct dispersal probability between nodes i and j that would result in a given change scenario . This value in the fourth column will be in general probability in the fourth larger different than the value in the third column; a smaller distance or a than in the third column will correspond to the case in which the quality or strength of the link 15

16 Conefor 2.6 user manual s increases in a given change scenario ; and a higher distance or smaller between two patche probability in the fourth than in the third column will mean that the connection between those two ssible for each patches gets weaker. All types of combinations and different types of changes are po of the links in a given landscape and “Link change” analysis. For example, some connections may be improved, some others may decrease its quality or even disappear completely, and some other links may suffer no change at all in the same ana lysis, depending on the particular values for each link that are specified in the third and fourth columns of the connection file. Obviously, for those links with the same values in the third and fourth column of the connection file the result will necessa rily be varM =0, since that would mean that no change is foreseen (evaluated) for that particular dM =0 and It is also possible to use a link file as the connection file for the “Link change” option, although in link. C can be calculated, with the only possible changes for this case only binary metrics such as II individual links being the complete disappearance of an existing link or the addition of a link (direct connection) between patches that did not exist in the initial landscape. Therefore, i , both n this case the third and the fourth column of the connection file can only contain either 0 or 1 for each pair of patches, indicating if a link exists (1) or not (0) between those patches in the initial landscape (third and if it will be lost, gained o r will column) remain unchanged in a given scenario (as specified in the forth column of the connection file ). he “Link change” analysis In t ill replace the value for a particular pair of patches in the Conefor w the fourth column of that file, which in general third column of the connection file by t he value in p value that might either make a link appear or disappear (binary connection model) or change the ij is associated to that link (probabilistic connection model). Conefor will rec e the network alculat ity (as evaluated by the value of for the entire landscape) after that change, and M connectiv will after report the resultant or varM value. This change will be done only for individual links (one at a dM values correspond to the connectivi varM dM time), so that ty change (either gain or loss) that or would result from implementing that change in a given link while all the others remain unchanged. As said above for the other modalities of link importance analysis, the connectivity losses as dM PC will measured by correspond to positive or and varM values, while connectivity gains will IIC dM or varM values (see formulas in section 9 for these variables). correspond to negative might correspond to an increased effective distanc e or The changes evaluated through this option resistance due to the intensification of a certain part of the landscape matrix (e.g. the development or other type of barrier), to a n increase in the r hospitability of the permeab ility o of a highway landscape matrix , to many other types and inten sities of change that may be of interest in or particular case studies and a given co nnectivity analysis . regions in accordance with the objectives of Note that in fact the “Link removal” and “Link improvement” options are just particular cases of the “Link change” modality, but with the former two not requiring to include a fourth column in the (since they assume that all the evaluated links change exactly to the same state, connection file which is the disappearance of a link or the unlimited imp rovement of the quality of a link, respectively) . If you use the “Link change” option and your connection file is for example a distance file, then if you set the fourth column all to zero values, the “Link change” result will be the same as that for “Link improvement”. If in the same case you set all the values in the fourth column to arbitrarily large (infinite) values, then the “Link change” result will be the same as for “Link removal”. , then setting all the values in the fourth If your connection file is either a probabilities file or a link fi le column to zero will provide the same results as in the “Link removal” option, while setting all the values in the fourth column to one will provide the same results as in the “Link improvement” option. 16

17 Conefor 2.6 user manual duce calculations” option does not apply to th e The “Re “Link change” analysis (it only affects the “Link removal” and “Link improvement” modalities). 1 . Precision of the calculations (high or standard) 5 The can be set to precision in the calculation of the values that Conefor will produce as an output . either “High” or “Standard” was possible in that version) and that was In version 2.2 the standard precision was used (no choice en the three fine for the applications for which that version was intended to be used. However, wh PC intra flux , connector ) of IIC fractions ( , are to be computed separatel y (as provided in this new and version) a higher precision is advisable to provide more accurate results for these three fractions, especially in large graphs / habitat n etworks. Therefore the high precision is the default in this new version. The high precision will consume more RAM and require a slightly higher processing time than the standard p recision (depending on the size of the network on the computer characteristics), but and ot make a significant difference for most of the this should n s (unless your network is so large user and/or your RAM memory so small that you are forced to use the standard precision to make the processing feasible). In summary, you c an change to the standard precision only if either you do not need the separated fraction values (i.e. you will just use the overall index value M or the total dM or varM value for each node) or if your computer is not able to process the graph with the mo re demanding high - precision mode . 1 6 . Other comments related to Conefor 2.6 The option to save DBF files is disabled in this new version, but you may easily convert the text files r external software if othe any (the output format for the results in Conefor) into DBF format with ti needed. software package (GIS, spreadsheet, sta tis Almost any cal, etc . ) is able to directly read the text files as produced by Conefor. ArcGIS is a ble to directly open and work with them as tables , as long as each column has an appropriate header (some of the files produced by C already onefor include these headers for each column; for the others you can easily add them in a new (first) line of text with any text editor). In the particular case of ArcGIS, some characters like parenthesis are not admitted in the names of the column headers; this means that you should manually change the headers for the BC(IIC) or BC(PC) metrics in the node importance file (e.g. , if to BC_IIC and BC_PC) this metrics have been selected for computation, before you can open in ArcGIS . this result file setu - Finally, in the “PC >PC >More p” window in the Conefor interface , you may notice that a new - option “Alternative calculation mode” has been added. This is however a n implementation in progress andled with much care. We recommend users not to perform any that should be h calculations of the PC metric with this option activated. 1 7 . GIS extensions Several GIS extensions have been developed specifically for Conefor. These extensions allow generating from a spatial layer (in either vector or raster format) the files required as an input to perform the connectivity analyses in Conefor. These extensions generate these input files (node file e exact format required by Conefor; therefore, these files can be and connection file) directly in th 17

18 Conefor 2.6 user manual used as they result from the GIS extensions with no other changes or intermediate processing steps. The You can find more deta ils about extensions allow for batch processing of multiple files/layers. s extensions (including the web pages from where they can be downloaded) at e th e . http://www.conefor.org/gisextensions.html 1 8 . Applications Conefor and the new and PC metrics have bee n used in a wide variety of conservation and IIC management plans, scientific studies, and official reports on biodiversity indicators by the European - related Commission and the European Environmental Agency, among other connectivity applications. Such applications have been reported in numerous studies comprising different types of species and ecosystems and a large number of countries, from China to the USA and from Brazil to Finland. You can get more details on these applications (and perhaps in spiration on the way you can use Conefor for the purposes of your own case study) from the references listed in http://www.conefor.org/applications.html available here . , and in the map of Conefor applications 1 . Empirical support 9 Several studies have ev aluated and demonstrated the ability of the new habitat availability (reachability) metrics implemented in Conefor ( IIC , PC , etc.) to explain or predict ecological processes related to landscape connectivity, including species distributions, colonization e vents or genetic diversity patterns at the landscape scale. Most of these studies have been performed by other research groups and institutions different from the one that developed Conefor and the metrics there implemented, which provides an independent a ssessment and empirical support to these quantitative developments and metrics. Some of these studies have evaluated the ability of IIC or PC to explain ecological processes and have compared it with the performance of other existing have been shown to outperform the other PC connectivity metrics. When this has been done, IIC or analyzed connectivity metrics by presenting a stronger relationship with the analyzed ecological processes and empirical data. on this empirical support an d validation at You can find more details . http://www.conefor.org/empirical.html 20 . References IIC and metrics: For the PC - Pascual - Hortal, L. & Saura , S . Comparison and d evelopment of new graph - based 2006. landscape connectivity indices: towards the priorization of habitat patches and corridors for conservation. Landscape Ecology 21 (7): 959 - 967. Download - Saura, S. & Pascual - Hortal , L . 2007. A new habitat availabili ty index to integrate connectivity in landscape conservation planning: comparison with existing indices and application to a - 3): 91 - 103 . Download case study. Landscape and Urban Planning 83 (2 For the three fractions of the PC or IIC metrics ( intra , flux , connector ): - Saura, S. & Rubio , L . 2010. A common currency for the different ways in which patches and links can contribute to habitat availability and connectivity in the landscape. Ecography 33: load Down 523 - 537 . 18

19 Conefor 2.6 user manual EC ) or Equivalent Connected Area ( ) index and its use for For the Equivalent Connectivity ( ECA monitoring changes in landscape connectivity: - - Freire , M . 201 1 a . Network analysis to assess Saura, S., Estreguil, C., Mouton, C. & Rodríguez ivity trends: application to European forests (1990 - Ecological landscape connect 2000). 11: 407 416. Download Indicators - Saura, S., González - Ávila, S. & Elena - Rosselló , R . 2011 b . Evaluación de los cambios en la - conectividad de los bosques: el índice del área conexa equivalente y su aplicación a los Montes, Revist - 21 (only available in bosques de Castilla y León. a de Ámbito Forestal 106: 15 Download 1 (short version as published on paper). Spanish). load 2 (expanded version Down as available only online at http://www.revistamontes.net/ ). BC(PC) BC(IIC) For the generalized Betweenness Centrality metrics : and - Bodin, Ö. & Saura, S. 2010. Ranking individual habitat patches as connectivity providers: g network analysis and patch removal experiments. Ecological Modelling 221: integratin 2393 2405. Download - For the complete, non - redundant and multifaceted view provided by the IIC or PC - r elated fractions or metrics: - Baranyi, G., Saura, S., Podani, J. & Jordán, F. 2011. Contribution of habitat patches to network connectivity: redundancy and uniqueness of topological indices. Ecological 1310. Indicators 11: 1301 - Download For the software: - Saura, S. & Torné , J . 2009. Conefor Sensinode 2.2: a software package for quantifying the importance of habitat patches for landscape connectivity. Enviro nmental Modelling & Software 24: 135 - 139. Download We prefer any of the references above for citing the Conefor software package or the metrics and however, for some reason, you need to cite this manual for any methods implemented in it. If content that is not available in the previous papers, please use the following reference: - Saura, S. & Torné , J . 2012. Conefor 2.6 user manual (April 2012). Universidad Politécnica de . Madrid. A vailable at www.conefor.org 19

Related documents