The influence of study area selection and landslide inventory practices on landslides spatial distribution: an example from Northern Morocco

Abstract

Numerous studies focused on the technical limitations of Landslides Susceptibility Maps (LSM). They were concerned with the impact of LSM technique selection, conditioning factor combinations, and/or Landslides Inventory Map (LIM) practices on LSM sensitivity. However, no previous papers focused on study area selection and its influence on the output. In fact, most authors subdivide their study area into administrative/political territories, which may be useful for decision makers but is not very informative from a pure scientific stand point. Therefore, 3 territories of Northern Morocco were investigated in this study: the first corresponds to the 1:50 000 Tetouan topographic map (cartographic), the second covers Martil watershed (geomorphological) and the third is Tetouan province (political). The latter study area is of capital importance given its two contrasted geological and morphotectonic domains (Internal and External Rif), which may produce errors in the output. The input LIM datasets for the purpose of this study are: new-active LIM, Inactive-young LIM, Relict LIM, and all landslides LIM. We used two conventional LSM algorithms (Logistic Regression and Artificial Neural Networks) in order to avoid technique-specific biases. Our results show that study area selection is not as important as LIM with regard to the output LSMs, but remains very relevant in determining LSM distribution and accuracy for Tetouan map and Martil watershed study areas. As for Tetouan province, the model is unchanged using the same LIM in the External Rif but changes significantly in the Internal Rif. Our LSM analyses also revealed the link between landslides age and elevation in the External Rif domain where relict processes are mostly concentrated in mid-slopes while new-active ones occur in lower slopes. This is not observed in the Internal Rif, which further exhibits the importance of study area selection based on naturally delimited geomorphological units rather than political or cartographic boundaries.