I am doing a prediction using 2 images of 1992 and 2010. However, the output of my LCM comes out fine except for a two stripes in the middle with values of 2 different land covers. These stripes get bigger and further includes speckels with increase in the prediction year. I have reinstalled Idrisi on my machine and still having the same problem. This problem was not there when i used a smaller section of the image ealier. What have i done wrongly?
Thank you in advance!
I am facing the same problem. Soft predictions look alright but the hard ones have stripes. These stripes are usually present in the upper half of the image and the major class in the stripe is the class to which most transitions are expected to occur. For example, in the transition matrix, if there is a high probability of transition from forest to agriculture, the output has a stripe on top in which all forest is converted to agriculture and below that the forest remains intact. So as I increase the prediction year, the stripe becomes wider, all forest disappears within the stripe and below it the forest is intact. Based on past landscape change, i don't expect forest cover to change in such systematic, north to south direction. Heres a link to one of the output images (https://drive.google.com/file/d/0B30fTbRLC0_GeEUxNFBpUE1SREE/view?usp=sharing). Please let me know what could be the problem. Thanks.
Most likely the issue has to do with over allocation. That is, you are trying to allocate more area (quantity) than you have available for that transition. In which case, it will assign randomly, areas of background, or other land cover classes. Review your transition area matrix in your working folder and the number of years you are allocating to. It is a linear allocation from time 2 to time 3.
James Toledano
That seems likely. But then is there a way to control that or incorporate non linear trends in any way? I have multiple maps, say for time 1,2,3,4 and I am trying to predict 5. So I can model and predict a trend and may be able to generate a transition area matrix outside IDRISI. But can that be then incorporated to get change allocation and hard classified maps using LCM?
It is a linear process and assumes that the transitions you are modeling undergo the same process during all time periods, and into the future. You can use your own matrix, you load this as an external model on the first panel under the Change Prediction panel. The format of this matrix is in the Help. Although it is a linear process, you do have the option to insert planned infrastructure changes that can alter some of the driver variables in the future.
James
Dear
I am making a model of change of use and land cover, it has 21 transactions, the model is executed correctly but at the time of making the prediction for any year it throws in the hard prediction a line that crosses the entire landscape, I have identified that this occurs with a certain transition, and I have examined the variable files of this transition, as well as the probabilities of change, and the coverage of the initial years and I cannot find this line anywhere, his transition is not the largest that I execute in the model,
How do I solve the over allocation problem within the transition area matrix?
