Quantitative study of the impact of natural areas of the Markazi Province on pollination based on a distribution modeling approach

Extracting honey from beehives is one of the economic activities for local communities, which is effective in the direct and indirect employment of villagers and as a result, the sustainable development of these areas. Among the pollinating insects, bees play a much more prominent role, and usually the location of the hives can determine the extent of benefit from this ecosystem service. Among bees of Iran, Carnica hybrid (Apis mellifera meda) has a special place in honey production. Despite various studies conducted on this hybrid, so far none has investigated the suitable conditions for the placement of beehives of this species, so this study seeks to identify suitable areas for the establishment of beehives. Awareness of the areas that are prone to beehive placement can be one of the priorities of the planners in the field of agriculture and animal husbandry in Markazi Province.

In this study, in order to model the areas that are susceptible to pollination, variables such as spring density, height, topographic humidity, light shade, average spot size, land use/cover, landscape diversity edge density, distance from agricultural land, average wind speed up to a height of 10 meters, roughness of vegetation and vegetation density were used. Since it is difficult to access all the areas that are not suitable for the establishment of hives in the modeling process, alternative methods such as pseudo-absence methods were used. However, identifying suitable areas for recording pseudo-absence points can also result in errors. Therefore, first, using the output of presence-only models, pollination desirability was calculated. Then, by subtracting the desirable areas from the entire surface of the land, pseudo-absence points were randomly created in the remaining areas. After preparation of this group of points, the presence/pseudo-absence models were ready to be implemented. In order to evaluate distribution models, TPR variables and the Kappa index were used. TPR, which is also referred to as sensitivity, is a numerical value that identifies the percentage of presence points that are detected again after applying the presence point threshold. Also, a random forest model was used to calculate the impact of this data set on environmental changes.

The presence-only models in this study were implemented with adequate power. The value of AUC was calculated as 0.89, 0.90 and 0.76, respectively, for Bioclim, Domian and single-class support vector machine models. The results of the evaluation of the used models showed that all models have well predicted the presence of beehives in the areas of pseudo-absence of beehives. The Kappa index for this category of models was at least equal to 0.83. On the other hand, based on the TPR criterion, many of the hive points have been detected again after applying the hive threshold, which can indicate a good level of prediction of the used models. Also, the findings showed that the diversity of the landscape had a greater impact on the quality of pollination than the proximity to agricultural lands. The height of up to 1813 meters above sea level, as well as the wind speed of 3.47 meters per second, were the best conditions for the presence of beehives. Among the different cities, Arak, Farahan, Khandab, Shazand and Khomein had the highest value for pollination.

Planning for the protection of natural areas as well as areas prone to the establishment of beehives can be done with emphasize on the location of Arak, Farahan, Khandab, Shazand and Khomein cities. The findings of this study show that the use of species distribution models can be effective in identifying suitable areas for beehive establishment and pollination activity. On the other hand, combining the findings of this group of studies with other spatial data that determine the patterns of the landscape can provide a clear view of the influence of the landscape.