bayesian averaging model

Qualitative modeling is one of the methodologies employed for assessment. However, skepticism about the outcomes of such models has led to a reduction in their consideration. Consequently, this study employs the qualitative model approach of PAP/RAC to predict and determine the status of water erosion risk in the Qarnaveh watershed. This approach is juxtaposed with the RUSLE model, and influential factors are identified using Bayesian Model Averaging (BMA) to enhance evaluation, delineate vulnerable zones more effectively, and propose viable water and soil conservation strategies. Research findings revealed a 20.56 Km2 difference in predictions between the two models. Nonetheless, a meticulous spatial examination of water erosion risk prediction maps from both PAP/RAC and RUSLE models demonstrates a spatial congruence of 67.74%, predominantly within the "very high" category. Furthermore, the analysis of influential factors (slope, lithology, land use, and vegetation cover density) on the erosion risk prediction map, employing the Bayesian averaging approach, underscores slope as the most significant determinant in the developed superior patterns. In the PAP/RAC model, it exhibits adverse effects in the "very low" category but exerts positive influence in the "moderate," "high," and "very high" categories. In view of the significant influence of slope across all categories, it is therefore necessary to carefully classify and assess regional characteristics for refined predictions. Consequently, based on research findings, the PAP/RAC model emerges as the most efficacious and straightforward qualitative model for estimating water erosion risk in data-deficient watersheds, with precise consideration of its influential factors aligned with Iranian context. This methodological approach, which integrates descriptive and integrative techniques, not only facilitates precise prediction, but also facilitates the formulation of effective water and soil conservation strategies, thereby contributing to the development of a comprehensive strategy for sustainable watershed management.

It is necessary to evaluate, identify and prioritise areas prone to erosion. However, the nature of soil erosion and its processes have necessitated effective water and soil conservation strategies, which require hazard assessment at various temporal and spatial scales. Therefore, the current research utilizes the PAP/RAC qualitative model to predict and determine the water erosion hazard status of the Qharnaveh watershed. It also investigates the uncertainty of the factors affecting the erosion hazard map obtained from this model using Bayesian Model Averaging (BMA). This is done to enhance the evaluation, identification, and prioritization of points at risk of water erosion, thereby providing effective strategies for water and soil conservation to mitigate the trend and development of soil erosion.

The Priority Actions Program/Regional Activity Centre (PAP/RAC), established in 1977, is one of the six Regional Action Centers of the Mediterranean Action Program (MAP), which is itself part of the United Nations Environment Program (UNEP). The MAP includes 21 Mediterranean countries and the European Union, with a common objective of creating a healthier Mediterranean environment on the basis of sustainable development principles. The main area of PAP/RAC is to support Mediterranean countries in improving the sustainable management of their coasts, in particular through the adoption and implementation of the Integrated Management Protocol in the Mediterranean Coastal Zones. In 1998, the PAP/RAC model and approach were presented and employed as a PAP/RAC. According to the executive structure of the PAP/RAC model, in the first stage (predictive approach), in the first step, a Watershed slope map was prepared from a DEM at a height of 30 m, and based on the valuation manual, it was divided into five classes. In the second step, the lithological map was prepared from 1:250,000 geological maps of Golestan province, which was divided into five classes based on the valuation manual. In the third step, the slope erodibility map, which is the result of integration of the rock unit map and the slope map, was evaluated and prepared according to the instructions. In the fourth step, a land use map was prepared based on the classification of the Sentinel 2 satellite images of August 2023 in the Google Earth Engine along with the existing regional information and divided into four categories: residential, forest, pasture and agriculture and evaluated according to the instructions. In the fifth step, the vegetation density map for August 2023 was calculated using the Normalized Difference Vegetation Index (NDVI) and evaluated according to the instructions. In the sixth step, integrating the land use map and vegetation density map according to the instructions, the soil conservation map was created. In the seventh step, a predicted erosion hazard map was prepared by integrating the erodibility map (integration of slope and lithology) and the soil conservation map, considering the influence of slope factors, lithology, land use, and land cover. After creating the erosion hazard map of the region, field survey and aerial photos were used to assess the erosion states, facilitating the implementation of a descriptive approach to the model. Finally, by overlaying the erosion states on the erosion hazard prediction map, the model integration approach was executed. For the second approach of this research, the area of the watershed (each hectare) was considered as the size of the statistical population. Cochran's formula was utilized to determine the number of random points. Using the 'Create Random Points' command corresponding to the boundary of the Qharnaveh watershed in the ArcGIS software, random points were placed. Based on the value of each point in the model factors (slope, geology, vegetation, and land use) using the BMA method, the impact of each factor on the water erosion hazard map was determined in the R program environment.

The results of the PAP/RAC qualitative model showed that 0.19, 3.34, 34.94, 227.28, and 514.53 km2 of the watershed area were classified as very low, low, moderate, high, and very high erosion hazard, respectively. In other words, 741.81 km2 of the watershed area exhibited high and very high erosion hazard. Generally, according to the factors of the PAP/RAC qualitative model, the results indicate the high potential of this watershed for erosion and soil loss. Further, examination of uncertainty with the BMA reveals the significant importance of slope and lithology in the model output.

Considering the importance of evaluating erosion-prone points, it is recommended that the assessment of PAP/RAC model factors in the field should be conducted more meticulously. Additionally, if feasible, model guidelines tailored to the conditions of Iran should be developed to devise more effective water and soil conservation strategies based on the model results, aiming to mitigate the trend and development of soil erosion.