---

Change detection and prediction of land cover change in the overall vision for better management of natural resources, protection of agricultural land around urban areas is very effective long-term policies and strategies. one of the methods used to control the process of changes in land cover and land use planners, is modeling. The land use maps 2000 and 2014 BEHBAHAN city was prepared using maximum likelihood classification. It also uses a Markov chain and logistic regression was 2028 land use map. Mapping modeled after the detection of changes in land use map was made in 2014 and 2028 with the LCM. Assess the accuracy, land use maps in 2000 and 2014 to evaluate the accuracy Kappa coefficient and logistic regression with (ROC) Relative Operating Characteristic was calculated. The results of modeling with logistic regression model showed that in all sub model ROC statistic is more than 94/0 These results show a very good offer logistic regression model in the analysis of changes in land use change modeling The results showed the sharpest decline mainly include changes is destruction of rangelands and convert it to other land uses. The highest increase was seen in the area of agricultural land use. Rangeland degradation process can be alarming for managers and urban planners and natural resources. To prevent the decline of rangeland degradation, the need to provide appropriate and efficient management solutions in the fields of grazing and rangeland restoration and improvement programs.                                             

---

Abstract

Detecting land use, land cover changes and recognizing effective factors is necessary to prevent land use changes and better management. The aim of this study was detecting changes of Arasbaran forest cover in two periods of 12 years, modeling and predicting forest cover destruction in this region. At first, the multi temporal Landsat 5 images in 1990, ETM+ Landsat 7 in 2002 and OLI Landsat 8 in 2014 were provided and were classified in two categories including high dense forest, low dense forest. Forest changes were detected in three periods, 1990-2002, 2002-2014, and 1990-2014, also changes in forest cover were estimated in different classes of variables influencing changes. Forest area changes in the study period were modeled by logistic regression models and Geomod. In order to compare the performance of these two models in predicting land uses status by preparing maps in 2014 and validating by real map of that year. Results showed that in the period of 24 years, 992 and 1592 hectares of high and low dense forests were degraded during 1990-2014, respectively. The results of decreasing forest cover modeling showed that variables such as distances from roads and residential, elevation and slope has a direct relation with forest degradation. However, there is an inverse relation between forest degradation and distance from forest variables. The validation result of forest cover maps which is predicted in 2014 show total accuracy and kappa coefficient is 96.8 and 0.9342, for logistic regression map and 96.4 and 0.9269 for Geomod map respectively. These results indicated that model had a good performance in predicting of land use changes. Finally, using the logistic regression and Geomod, forest cover changes predicted for 2025. The result of predicting showed that the forest cover will degradeted 3.9% in the next 10 years.

---

At present, the question of identification, control, and prevention of human, economic, and social losses resulting from natural events such as earth tremors, floods, and earthquakes have attracted special interest in scientific-research communities and in responsible authorities of most countries around the world. In recent decades, considering the upward trend in losses and damages caused by natural events (especially landslides), prediction of damages and losses and introduction of solutions and methods for controlling and avoiding them have been addressed in earnest. This research evaluated the region around the Benaravan fault using logistic regression (logit regression or logit model) to determine regions that face slope instability hazard. Field surveys were made, previous research was reviewed, and the prevailing conditions in the region was studied first and, using the Idrisi software, it was found that the nine factors of altitude class, slope, dip direction, lithology, distance from the fault, distance from the waterway, distance from the road, land use, and vegetation influenced occurrence of slope instability. After performing slope instability hazard zonation, percentages of slope instability in each class were calculated. Results indicated that areas with high hazard zonation in the study region constituted the smallest part of it. In this research, which used logistic regression, the elevation factor with the highest coefficient was the best variable for predicting the occurrence probability of slope instability in the region. The highest incidence of slope instability occurred at high altitudes with slopes of 23-32 degrees because of gravity force.

---

This study aimed to identify factors leading to slope instability, Maps preparation, determine potential areas mass movements and risk zoning in the Upper Basin Komanaj Chay. That is one of the important basins in the northern city of Tabriz, by Using logistic regression models and artificial neural network done. This basin due to topography, tectonics, geology, stratigraphy, and the climate is prone to a variety of slope instability, this phenomenon always occurs. According to the study variables such as altitude, slope, aspect, type of formation, distance to fault, distance from the river, land use, distance from the road, as the independent variable And distribution of unstable slopes as the dependent variable using logistic regression models and artificial neural network was analyzed .The results showed that the most important factors in the occurrence of slope instability in the basin are as follows: Elevation, distance from the river, lithology, faults, slope and aspect More than 50 percent of instability range in height from 1850 to 1520 in the study area dip 32-17 degrees, at a distance of 200 meters from the canal and 500 meters from the fault occurred. According to the results of a very high percentage of areas the risk of neural network and logistic regression models respectively 5.6 and 8.3 percent is the mainly areas close to the drainage network which includes the lithology of these areas are located in areas with lower resistance. Statistical methods logistics showed a lot of reflects of faults and lithology in this areas is based Landslide. Evaluation ROC indicator showed that the model was assessed using logistic regression model is 0.894 and neural network models is 0.826. In fact, both models show a high value and suggest that mass movement and slope instabilities observed a strong relationship with probability values derived from logistic regression models and artificial neural network model. The results of this study can be useful risk management slope instabilities and control is deteriorating factors.

---

.Abstract
Landslides and slope instability are major hazards to human activity, which cause damage, this instability and hazard are potential and damaging phenomena, our country is mainly mountainous topography, tectonic and seismic activities. The high humidity, diverse geological and climatic conditions have created most of the natural conditions to create a wide range of landslides. The purpose of this study was to identify the effective factors in landslide generation and identify potential landslides in Alamutroud basin in Qazvin province using logistic regression method. In this study, landslide distribution map of the basin was prepared by studying aerial photos and field surveys using GPS and satellite imagery. Eight effective factors including slope, slope direction, sea level rise, distance from riverbed, land use, lithology, distance from fault and drainage network density were used and then processed using Matlab software. , ARC GIS, IDRISI and EXEL. According to the results, 58.26% of the catchment area is in the low and very low risk zone and 17.44% in the high and very high risk area. Also the factors of height and distance from the ranch have the highest scores in landslide creation. Also the model validation indices including pseudo ROC and R_square ChiSquare were calculated /957 -2511/283 and/ 3518  respectively which have acceptable validity.