Suhel Sen, providing more information about your data will greatly increase the chances of someone being able to help you.

Relying almost entirely on interneTelepathy and ESPss (credit to the late David Marso for coining those expressions), I suspect you are estimating a binary logistic regression model with occurrence of a landslide (Yes/No) as the outcome variable. If so, use the SAVE sub-command to save the predicted probability as a new variable in your dataset, and use it as input to the command that generates the ROC curve. But the best thing would be to provide a lot more info about what you are doing! HTH.

Thank you sir for your response. I shall be grateful if I get any video tutorial.