Data preparation for Land Change Modeler is very format specific. Users that are currently using IDRISI Andes and Taiga will find this discussion most helpful, while users with later versions have the benefit of the Harmonize tab in LCM. The Harmonize tab resolves issues related to the input earlier and later land cover maps specified in the LCM Project Parameters panel in Land Change Modeler. Land Change Modeler requires that the input land cover maps meet specific requirements in terms of their spatial characteristics, backgrounds, and legend categories.
The formatting requirements for the input land cover maps include:
- Consistent spatial extents
- Consistent references systems
- Consistent pixel resolution
- Identical legend categories
- A background value of zero
- Identical background area
How to create legends in IDRISI to comply with the requirement of the Land Change Modeler:
To create the legends, please go to the TerrSet Explorer pane and select the Files tab. You should now see all of the files either in your Working Folder or in a Resource Folder. Highlight one of the land cover maps for which you wish to update the legend. Once highlighted, scroll through the Metadata below until you come to a row titled “Categories”. Click in the adjacent cell and a pick list button (…) will appear. Click the pick list button – this is where you will fill out your legend. Make sure to include a name for every legend category, and do not skip any categories. The code is the value as it appears in your map. For example, if you have four categories (1, 2, 3, 4) representing agriculture, forest, built, and water, the table should be filled out as follows:
Press OK and then click the save icon in the lower left hand corner of the Idrisi Explorer to save your edits. To update a second map with the same legend, highlight the file and again navigate to the Categories window. This time, in the lower left-hand corner of the window there will be a button that reads, “Copy from…”. Click this button and select the land cover map that you just updated. This will automatically fill in the legend categories for you. Save your edits. You should now be able to input the map files into LCM.
Thanks for the useful information. But, what about the driving factor layers. For example, I have a distance to roads layer, do I have to make categories for this layer also. Again, if I have a soil layer with 5 classes as driving factor, should I divide the raster to 5 different rasters or just categorizing the legend will do? Thanks in advance for your help.
I also have a similar question as Chitrini. Should soil types be assigned a unique value, similar to how land classes are assigned values?
Land cover maps should have a background value of 0, but what about driver variables? Sometimes 0 is significant in a driver variable, such a slope. What do you recommend for background, nodata values for driver variables?
Also, to help with Chitrini's first question (distance to roads) I have been using the values 0-x, with 0 being the feature and x being the farthest distance from the feature in meters. No need to use categorical values for distance drivers.
This is only for land cover maps.
Yes, I haven't been able to find a similar article for driver variables though. I am hoping to learn more about formatting for driver variables and have my questions answered. If you know of a resource for specific formatting guidelines for drivers please post a link!
The format for Driver variables:
Thanks! And what about background values? Sometimes 0 can be significant in driver variables (like slope or elevation). Will using 0 for a background value affect the results?
Also, how do you handle variables like soil or other non-distance from based driver variables? Do you just use a unique value for each soil type?
You must not use zero as null-value (or background) because there are a lot factors related to proximity. I recommend you use as null-data = -9999. Really, this decision depends of your data. You should specify this value. Null-value or background value must be the same in all factors. Finally,you should transform the categoric factors by using "Variable Transformation Utility / Variable Transformation Utility - Evidence likelihood transformation"
If you have land cover maps with background, you do not have to be concerned with the background value of the driver variables for those areas the coincide with the land cover background. When training during the Transition Potential stage, samples will be take from area of valid land cover only, not from background areas.
Per Oli's comment, you will find in the tutorial a working example for creating evidence likelihoods for your soil data.
Using a Landsat scene to carry out supervised classification.
The background on Landsat scene appears as a feature (class) in my classification.
Initially has a value of zero but still appears
Any idea how to solve this?
Even though I have filled the new ID properly, I got a error message "Please fill all the New IDs and indicate 0 for the class you want to remove". I have maps wit 30 LULC classes, is this a problem?
What about the message "Unexpected error condition encountered. Please check the integrity of the operation performed, the data files used and the amount of disk space"
I import arc ascii into idrisi and I set the geographic projection as "mexccl"
Any suggestions to fix my problem?
Why do I get a message that my spatial reference system is incompatible with IDRISI reference system. My land cover images have WGS_1984_UTM_Zone_49S. How do I fix this problem?
@Roshan Sharma. In Idrisi, referene system should be planar. At the LHS panel of the idrisi explorer, check the properties of your LULC images and make it planar. I hope it should your solve your problem.
@Akanksha. Thanks much. However, I don't have IDRISI. I use LCM extension in ArcMap. Can't find a way to fix that. Still stuck in the harmonize phase. :(
and also isn't UTM a planar system?
I get this error message.
Well, I don't have LCM extension in ArcMap. So, cant help much!
But in Idrisi, what happens is, .img or other raster formats need to be converted to idrisi raster format i.e. .rst. And when it gets converted to .rst, during harmonising the data, it asks you to mention spatial reference system as 'Plane' - this is what is compatible with IDRISI reference system (I have read it somewhere long back).
So, when i was working on IDRISI, my raster images were in UTM WGS 84 43N. And when converted to .rst, in reference system tab in metadata, I wrote 'Plane' and units were Meters. And then it worked well for me.
Though I dnt know about LCM extension in ArcMap, but I think it should work the same way. I would suggest you to check the metadata of raster image in LCM extension and see if you can change the reference sys to 'Plane'; and then see, if it works fine.
All the best!
@Akanksha. It worked. I had to change the metadata to plane manually. But Now I have a problem with the driver variable. It shows that the -" Driver Variable 'roads' dimension does not match with the earlier and later images".
For the roads raster I have the same spatial extent, coordinate systems and resolution. But still I get this program. I am wondering why there is not much support from Clarklabs in this forum. While other forums for eg. InVEST have such a quick response from the developers.
@Roshan. Does roads as driver variable has the same no. of rows and columns as your LULC images? This is the only thing that is coming to mind as of now. Do you have roads as a separate LULC class in your LULC image? Asking just for my information as I am thinking why it is showing error related to two images of different time periods?
@Akanksha ,I have the same issue or error. Do the number of rows and columns for my driver variables (eg dist to roads) have to be the same as my land cover layer column and rows? Can I specify this when exporting data from ArcMap (ie change rows and columns).
Okay I have checked the extent and number of rows and columns for land cover and driver variables, they are the same so im not sure whats the issue?
what is the error?
At the time I run LCM, after doing harmonize. Then I click continue, the program runs well up to bring up the graph data change analisys but when I will running "create maps" on Tab "change maps" error appears in the form of "attribute value documentation file: \ D ... \ id not found"
Can I get suggestions regarding the issue. Is there a process error?
I would like to find out the land cover change in my study area during the period of 2001 - 2010. I classified the land use in the study area by using Erdas imagine. The obtained result shows that the urban coverage in the area during 2010 is higher than 2001. But the net change analysis results show that the net change in urban area between 2001 - 2010 as a negative value ( quantification). Can I resolve that problem?
I would like to generate Markov transition area files and probability matrix in IDRISI with 8 LULC classes. All the land cover data (inputs) have LULC 8 classes. However, the model stretched them into 15 classes which is not real in the ground? What could be the possible reasons for having 15 classes and how can be resolved? Thanks.
HI, At what step did the land cover maps change from 8 to 15 classes?
I tried to use the land change modeler comparing 2 raster images of landcover. When I try to process it, it tells me "background areas do not match or background code is not 0", and asks me if I want to resolve it. I click yes, but then in the harmonize background window, category value(s) to assign to 0 for both images are already set on 0...
So I click Run and this error message comes up:
Attribute values documentation file D:\C.H\Memoire\Final\TerrSet\idrtmp_bool.ADC not found.
I have no idea what this message means, can you please help me?
hi..i tried to validate LU 2015 from (2003 and 2009 data), but the result was "error 0" how to fix it ?