Appropriate site selection for the astronomical observatory - Erzincan province sample application

,


Introduction
In order for observatories to operate at high efficiency, it is important that the site where they are built provide the most ideal observing conditions.Site selection is an evaluation process to create the most suitable conditions by combining many elements from different sources and consists of the steps of site selection and site testing, Hudson & Simstad (2010) and Koc-San et al. (2013).Site testing is a process that requires many steps, such as sky brightness, atmospheric visibility, precipitable water vapor, and atmospheric extinction coefficient calculation Koc-San et al. (2013).This process will expand in direct proportion to the number of alternatives.It is important to select the site in such a way that the number of alternatives is limited in order to reduce the time, energy, and cost required for the site testing process.The use of multi-criteria decision analysis (MCDA) in conjunction with geographic information systems (GIS) and remote sensing, and the ability to manage complex and voluminous data from multiple data sources, is a powerful tool that provides an efficient process for site selection.In recent years, the development of remote sensing resources for site selection, the use of MCDA and AHP, and their support by a GIS system have attracted the attention of researchers, Koc-San et al. (2013).The Analytic Hierarchy Process (AHP) has become one of the most widely used methods in the decision analysis since its development by Saaty (1988).AHP is a method of decision making in which the weights of criteria are determined by pairwise comparisons.The determination of weights by expert decision makers who evaluate the criteria increases the reliability of the method.There are many studies in various fields using decision analysis approaches along with GIS, AHP and the multi-criteria decision analysis (MCDA).Selection of the most suitable site for a biogas plant, Tulun et al. (2022), selection of sites for solid waste disposal Bilgilioglu et al. (2022), collection areas after earthquakes, Aman & Aytac (2022).The first site selection for an astronomical observatory in Turkey was made in 1989 by Aslan et al. when determining the site for the construction of the TUeBTAK National Observatory (TUG).In this study, field experiments, meteorological measurements and anthropogenic factors were given priority in determining the suitable site, Koc-San et al. (2013).Another study was conducted in 2013 by Koc-San Vd. using GIS, the MCDA method integrated with remote sensing methods and AHP to propose alternative sites for the TUeBTAK National Observatory.The importance of astronomy in the world the increasing growth of Erzincan Binali Yıldırım universty in Erzincan Province, the prediction of scientific studies in the field of astronomy in the future encouraged us to choose the appropriate place for the construction of an observatory within the boundaries of the province.The aim of our study is to identify the most suitable sites for the construction of an observatory in the province.A method combining MCDA with remote sensing technologies and geographic information systems (GIS) was used to select the suitable site, Koc-San et al. (2013).There are many parameters to be considered when selecting an astronomical observatory with the most efficient operating conditions, Jia et al. (2012).Important parameters affecting the selection of suitable sites for astronomical observatories have been published by Walker (1984), Cowles (1989) and Hudson & Simstad (2010).In addition, a study similar to our study was conducted for Antalya province.Considering these studies, the parameters affecting the selection of a suitable site for the observatory were determined by Koc-San et al. (2013).In this study, three main groups were formed as meteorological, geographical, and anthropogenic parameters, 10 factors were created among these main groups and recorded as grid layers in GIS, Koc-San et al. (2013).According to the AHP method, these layers were compared with each other and their weighting values were determined by the experts.Considering these weighting values, the most suitable site was selected for the astronomical observatory in Erzincan, Turkey.In this study, we propose the most suitable optical astronomical observatory for construction by mainly overlapping the maps that we created by digitizing in the system GIS as a result of the data analysis using remote sensing, MCDA and AHP methods for site selection.

Workspace and data used
The study area is located in the western part of the Eastern Anatolian region of Erzincan province in Turkey and covers 11.815 km 2 , Figure 1.From the data of the General Directorate of State Meteorological Affairs of Erzincan Province, the average number of clear days per year is 116.6.In winter, when the air mass coming from Siberia acts from the east, there may be very severe winter days.The average annual temperature in Erzincan is 10.9 C, with an average of -6.7 C in January, the coldest month, and an average of 31.4C in July, the warmest month Verileri (2010) and Kaya (2011).The region is surrounded by many mountains whose peaks are over 3000 meters high.High altitude, moisture-free sites are the most suitable for observatory construction, Koc-San et al. (2013).In this study, datasets from three main areas (meteorology, geography and anthropogenes) were used to determine the most suitable location for astronomical observing conditions.The datasets are explained in detail.

Meteorological datasets
In our study, the meteorological data set consists of cloud cover, precipitable water, and wind speed parameters.One of the most important factors affecting an astronomical observation is the cloud cover of the area to be observed, Jia et al. (2012).One of the most important criteria affecting atmospheric permeability is the amount of settable water.Cloud cover and precipitation water are the most important parameters in our study, Koc-San et al. (2013).Of the meteorological parameters, cloud cover and collapsing water data are products of Moderate Resolution Imaging Spectroradiometry (MODIS) data, which covers the entire Earth almost daily.MODIS data have resolutions of 250 m (2 bands) and 500 m (5 bands) in the VIS and SWIR bands, respectively, based on different methods, Barnes et al. (1998).In this study, the MOD10A1 products were used for cloud cover and MOD05L2 for stormwater.The MOD10A1 products were obtained from the National Snow and Ice Data Center (NSIDC) and the MOD05L2 products were obtained from the Atmosphere Archive and Distribution System (LAADS) (NASA-EOSDIS, 2022).The edge length of each MOD10A1/MYD10A1 granule is 1200 km (corresponding to about 10), and the projections are equidistant and sinusoidal.MOD10A1 provides a thematic map with HDF extension that includes data on snowpack fraction, snow, cloud mask, terrain, and water classes, Hall & Riggs (2007).MODIS cloud mask data is binary information about cloud cover, i.e., whether a cloud is present or not Strabala (2005).In this study, cloud cover data were obtained from the snow cover fraction layer of the MOD10A1 data, Hall & Riggs (2007).Sedimentable water content, another meteorological data set, is one of the important criteria affecting atmospheric permeability.Sensors have been developed to determine water vapor content (Advanced Very High-Resolution Radiometer (AVHRR), Sobrino et al. (2003), and MODIS, Gao & Goetz (1990) and Kaufman & Gao (1992).Another MODIS product, MOD05L2 (near-infrared and QA infrared), is used to determine water vapor content, Sobrino et al. (2003) and Kern et al. (2008).To obtain the map of the working area of water vapor, the infrared data sets of water vapor obtained from the MOD05L2 product were used to determine the water vapor content (NASA-EOSDIS, 2022).In this study, 4380 granules in HDF format were combined in the program GIS for the spatial information of each parameter cloud cover and precipitable water, and 1095 data in the study area were used (365 3 years), covering a 3-year period from January 1, 2020, to December 31, 2022.Wind speed data were obtained from the Global Wind Atlas 3.0 map, a free web-based application developed and operated by the Technical College of Denmark (DTU).The Global Wind Atlas 3.0 was published in collaboration with the World Bank Group using data provided by Vortex with funding from the Energy Sector Management Assistance Program (ESMAP) (more information: https://globalwindatlas).Our study used wind speed data from the GWA version 3.2 with a spatial resolution of 10 m to account for local variations in average wind speeds.The Global Wind Atlas (GWA) primarily supports wind energy development in the exploration phase and in the preliminary wind resource assessment phase prior to on-site meteorological surveys.The Global Wind Atlas provides wind source mapping at a horizontal grid of 250 m and wind source mapping at 10, 50, 100, 150, and 200 m above the ground level.Users can download GIS data for all layers at a point, in a specific area, or within a country.The speed at which air flows over the water surface affects the evaporation rate.Humidity depends on the amount of vapor in the air.Therefore, a higher wind speed results in minimal evaporation of water and lower humidity.Lower wind speed results in maximum evaporation of water and higher humidity, Ravi & D'Odorico (2005).Wind speed has an indirect effect on site selection for astronomical observatories.

Geographical datasets
Geographic Data Group; The Digital Elevation Model (DEM) consists of geology, landslides, active fault lines, and data.As altitude increases, sky transparency increases, Hudson & Simstad (2010), components that increase atmospheric scatter are reduced.Therefore, DEM is an important parameter.The Shuttle Radar Topography Mission (SRTM) has a spatial resolution of 90 m 90 m DEM represents the topography of the studied area.DEM Data used in the study were obtained from the USGS Earth Explorer.The data were processed in the ArcGIS 10.8 program, Bal et al. (2021).The construction areas of the buildings should be kept away from geological conditions, landslide zones, active fault lines, and floodplains.Considering that the building to be constructed is an observatory equipped with expensive equipment, it is important to choose the safest construction site.For this reason, maps of geology, landslides, and active fault lines were used in our study.Geology, landslide and active fault maps (for the study area) were drawn separately for each of the geology formation, landslide and active fault parameters using the drawing editor via the soil unit map viewer application of the General Directorate of Mineral Research and Exploration, Akbaş et al. (2011) and Üzer et al. (2013).In Erzincan province, it is divided into four basic tectonic-stratigraphic units of the pre-Eocene age, which have different environmental conditions and are tectonically connected.They are the Munzur Limestone, the Erzincan Nappe, the imen Mountain Nappe and the Kelkit-Rela autochthon from south to north.Of these nappes, the Kelkit autochthonous nappe lies on the imen Mountain nappe in the south and the Erzincan nappe in the north, and the imendaı nappe lies above the Erzincan nappe.The Munzur limestone napi is located below the Erzincan nappe, Yilmaz (1985).The data on the geological maps show that the rock classes of the study area are in seven different formations and are from the Cretaceous, Lower Miocene, Upper Miocene-Pliocene, Oligocene, Eocene, Miocene, Malm Cretaceous, Lower Cretaceous and Upper Cretaceous, Gürsoy (2012).Another criterion for selecting a suitable site is the landslide inventory data.Landslide maps from the General Directorate of Mineral Research and Exploration were used.According to the landslide maps, landslide activity consisted mainly of Miocene clastic neritic limestone, Mesozoic unconfined basic and ultrabasic rocks.In the landslide map prepared for the study area, all landslide activities were marked as areas unsuitable for construction.It is important to build as far as possible from active fault lines, which is another parameter.The urban area is surrounded by the Northeast Anatolian Fault Line and the Ovacık Fault Line, as well as the secondary fault lines associated with these two main fault lines, Barka & Eyidoan (1993).Maps of active faults produced by the MTA were digitized using the GIS program, and maps of active fault lines were produced.

Anthropogenic datasets
Anthropogenic datasets were created using city lights, distance to roads, and point density of mine sites.City lighting is important for people to continue their activities when they cannot benefit from sunlight.It is one of the most important parameters that negatively affect astronomical observations.The Operational Line Scanning System (OLS) is one of the Defense Meteorological Satellite Program (DMSP) sensors and was originally developed for lunar light cloud observations, Elvidge et al. (1997).Its ability to image in low light has made it the preferred accurate and effective data source for imaging anthropogenic lights on the Earth's surface in cloudless night conditions without moonlight, Zhang & Seto (2011).In this study, the DMSP/OLS data were used to evaluate the effects of urban lights in selecting suitable locations.It is important to determine the positions of astronomical observatories in relation to mine sites because mine sites cause dustiness and affect sky opacity positively and sky observations negatively Koc-San et al. (2013).The locations of the mining sites were taken from the General Directorate of Mining Affairs of Turkey.In order to reduce the construction cost of the astronomical observatory and to make it easy to reach, the distance to the roads was used as a criterion in our study.These comparative data were obtained from the OpenStreetMap.The data for Turkey was obtained using the overpass turbo, Ma et al. (2015).

Methodology
In this study, there are five stages: (i) determining the criteria/factors and storing them in the environment GIS, (ii) data analysis, (iii) map evaluation using the MCDA method, (iv) determining the weight of each criterion using the AHP method and (v) combining the criteria.In this study, after analyzing raster data layers and digital data, a model was created in Arc-GIS 10.8 software.The data processing and analysis in this model is shown in the following analysis, Figure 2. In this study, three main parameters consisting of anthropogenic, meteorological, and geographic datasets were used to select the appropriate site, Koc-San et al. ( 2013).These datasets were obtained from the drawing editor, remote sensing data or digital maps, and a map was created for each dataset using the program GIS.It was ensured that the projection systems (WGS84) of the maps created or used were identical.Each data set used in this study is represented by a layer.Each data layer is converted from a raster layer to a vector layer.Many GIS -based applications are used to analyze these layers.A common field was added to each layer, and the field values were set between 1 and 5, where 5 is the best score and 1 is the worst one.All data layers were converted from vector to raster again according to the generated field data.In the next step, the weights of each criterion were determined by pairwise comparisons of the experts' criteria (45 comparisons for 10 criteria), as required by the AHP model.Finally, the weighting values of each level were entered into the Weight Overly application in the GIS program and the most appropriate location within the boundaries of the study area was selected.effective in astronomical observations.The best observation conditions occur during periods of cloud-free weather (no clouds or very few clouds).According to the World Meteorological Organization, total cloud cover is defined as the part of the celestial dome covered by all visible clouds, Mekhaimr (2017).The celestial dome is divided into eight sections, each section is called an arrow, and an estimate is made on how many arrows the clouds cover.It is defined as clear weather or slightly cloudy (< 2 octas), partly cloudy (3-5 octas) and cloudy sky (≥ 6 octas), Mekhaimr (2017).Observation conditions according to cloud cover classifications, 2 octas and below are the most ideal conditions for outdoor observation, the 3-5 octas December is defined as partly cloudy skies and partly offers periods of observation.Between the years 2010-2022, the cloud closure data of the meteorological department of Erzincan province were interpreted according to the above-mentioned octa data classes, and the number of days that can be observed was estimated as 247.5 days per year.

Precipitable water data
Rainfall information was obtained from the MOD05L2 dataset using the water vapor infrared data channel.For the three-year period, the settable water infrared data were processed and converted into a single data layer using the program GIS, and the stormwater data were presented using this layer.The stormwater vapor data are scaled from 1 to 5. In the scaling, 1 represents more settleable water vapor and 5 represents less settleable water vapor (Figure 3(b)).

Wind speed data
Wind speed data are from the Global Wind Atlas database.The data have a spatial resolution of 250 m.The data for Turkey obtained from the Global Wind Atlas are arranged according to the boundaries of the study area.The wind speed data are scaled from 1 to 5. Regions with low wind speeds were scored as 1 and regions with high wind speeds were scored as 5 (Figure 3 (c)).

Geology data
In the study, the geological map of the study area was created in a geojson format using the drawing editor application from the content of the geoscientific map viewer of the General Directorate of Mineral Research and Exploration.The obtained map was digitized using the program GIS and used as a geological data layer.Alluvial fans, sand, dunes and debris cones are marked on the map as unsuitable areas for the construction site, while the other geological infrastructures are marked on the map as suitable areas.In this data layer, a value of 5 represents geologically suitable areas and a value of 1 represents unsuitable areas (Figure 3 (e)), Koc-San et al. (2013).

Landslide inventory data
The layer of landslides was created by the same processes as the layer of geology.There are many active landslides in the study area.Areas with landslides should be excluded when selecting a suitable site Koc-San et al. (2013).During the study, landslide-prone and landslide-free areas were scaled as 1 and 5, respectively (Figure 3 (f)).

Active fault lines data
In this study, active fault lines were identified as vectors using the geoscientific map viewer of the General Directorate of Mineral Research and Exploration.These vectors were converted to raster data, and locations 20000 m and farther from the fault lines were scaled as 5, and locations 5000 m and farther were scaled as 1.The distance map produced was used as active fault line data (Figure 3 (g)).

Digital elevation model
During the study, GTOPO30 data was used for the elevation information, two granules with the boundaries of Turkey were downloaded, these two granules were combined in the application GIS and the study area was cut out from the map.The data range in the study area is between 755 and 3412 m.Accordingly, these data were divided into five data groups on a scale of 1 to 5, with the highest points scaled as 5 and the lowest as 1. (Figure 3 (d)).

City lights data
Data from the Defense Meteorological Satellite Program (DMSP) Operational Linescan System (OLS) databases within the study area boundaries were used Elvidge et al. (2001).The data have a radiometric resolution of 32 bits.The data sets are scaled from 1 to 5. 1 represents the most illuminated areas, while 5 represents the least illuminated areas (Figure 3 (h)).

Mining activities data
The study area includes chrome, iron, manganese, copper-zinc-lead metal mines and industrial raw materials such as asbestos, barite, gypsum, magnesite, marble, perlite, bricks and sulfur.The dust layer around mining operations affects astronomical observations.For this reason, a grid map has been created to define mining areas.The regions where mining is carried out are divided into 5 scales in the range of 500-2000 m and above.Distances of 2000 m and more to mining sites are scaled by a value of 5, 500 m and less by a value of 1 (Figure 3(j)).

Road information data
Road proximity data were obtained from the OpenStreetMap application.The obtained data were converted from raster to vector data, and the road proximity parameters were defined and converted to raster data again.The raster data are divided into the distance ranges of 100 m and less and 500 m and more.Distances of 100 m and less are scaled by a value of 5, and distances of 500 m and more are scaled by a value of 1 (Figure 3(i)).

Analytic hierarchy process
The maps created for each of the 10 data in a raster data format were scaled between 1 and 5 and combined with the program GIS according to the weight ratios determined by the AHP method.The AHP method is based on comparing data layers with each other by numerical evaluation on an absolute number scale, Saaty (1980), and is a good way to address seemingly complex problems with a simple hierarchy, Cheng et al. (1999).While the individual criteria in the AHP are compared with each other, their importance levels are evaluated using odd numbers between 1 and 9, although even numbers may be preferred in cases where the determinant is ambiguous.The assigned criteria values appear in the AHP matrix (see tab. 1).The importance of the criteria was determined by expert decision-makers in the fields of astronomical observations, earth sciences, meteorology, and urban planning.When comparing the two criteria, the most important criterion determined by the experts was rated 9.It is rated 1 if the two criteria are equally important.The specific AHP procedures are: 1. Determining the importance relationship between the indices and creating the overall AHP structure.
2. Creating the following comparison matrix A with the weights determined by the experts based on the scale from 1 to 9 (see tab. 2), Saaty (1990), where a mn is the result of comparing the m'th factor with its importance relative to the nth factor, a mn > 0, a mn = 1/a nm , a mn = 1 (when m = n).
3. A is the comparison matrix, W and w i (i = 1, 2, ..., n) represent the corresponding eigenvector of λ max and the weighting value for ranking, respectively and λ max represents the largest eigenvalue of the matrix; in this study, the λ max value was determined to be 10.911.
4. The consistency of the pairwise comparison matrix should be checked using the consistency index (C.I).The consistency check is performed (C.I, Equation 4).In this study, the C.I value was calculated as 0.107, In addition, to check the consistency of the created matrix, Saaty (1980), the consistency ratio (C.R) value should be determined.The C.R formula is shown below, Where C.I is the consistency index and R.I is the random inconsistency, Saaty et al. (2008) and Saaty & Tran (2007).The critical value for the C.R value is 0.10.Values above this value are considered inconsistent, while values below this value are considered consistent, Saaty et al. (2008), Saaty & Tran (2007) and Koc-San et al. (2013).For n = 10, the value of the random

Results and discussions
The most suitable site within the study area was determined by multiplying and overlaying each of the criteria maps (10 maps) obtained after the steps described above with the criteria weights calculated using AHP (Figure 4 ).The three main criteria identified, meteorological, geographic, and anthropogenic, are shown in red in the most suitable areas (Figure 4) In this study, according to experts, the most important parameters are meteorological datasets that contain information about cloud cover and precipitation water.As a result of the analysis of 3-year cloud cover data, the maximum and minimum values of the number of cloud-free days in the study area are shown in Figure 5.The map showing the number of cloud-free days was formed as a, b, c (see Figure 5).Among these classes, the highest number of cloud-free days was determined in the pixels in class a (598-672 days) (see Figure 6, blue pixels).The lowest number of days (559-632 days) was observed in pixels in class c.Although the pixels in class a are ideal for the number of cloud-free days, they are the candidate regions closest to the city center and the university campus.The closest candidate pixel to the city center is at least 8.39 km.
Precipitation water and water vapor is an effective parameter especially for astronomical observatories where observation activities are carried out at in- frared wavelengths.Pérez-Jordán et al. (2015).As the water vapor gets closer to the earth, its relationship with the humidity value increases.Observations are not carried out when the humidity value exceeds 85% Solmaz et al. (2021).The annual average humidity of Erzincan Province was 64.26%, Kaya (2011).These average humidity values mean that observations can be made in general.
The smallest and maximum wind speed values within the provincial borders of Erzincan are shown in Figure 7. Wind speed is the limit value for 11 m/s observations, Solmaz et al. (2021).There are very few areas above the limit value within the provincial borders of Erzincan.However, wind speed values above the limit value are not among the candidate pixels (see Figure 7, red).

Conclusion and Recommendations
This study presents an exemplary application of the approach created by using MCDA, remote sensing and GIS technologies in selecting the most suitable site for the astronomical observatory as a result of analyzing complex criteria.This approach is a powerful method for site locating an astronomical observatory.This methodology, supported by field trials, will provide temporal, economic, and energy advantages in determining the location of the observatory.Applicable for areas where the future observatory construction is planned.

Figure 1 .
Figure 1.Location of the work area

Figure 2 .
Figure 2. Flow chart of the process steps followed in the study.

λ
max = 10.911,C.I = 0.107, R.I ( n = 10) = 1.49,C.R = 0.068 (A: distance to roads, B: geology, C: landslide, D: distance to active faults, E: wind speed F: point density of mining sites, G: city lights, H: rainy water, I: cloud cover and J: DEM).The 10 maps with data layers used in the study were overlaid in the program GIS in relation to the weighting ratios determined by the AHP method, and the most suitable areas (Figure 4, red areas) were proposed for the construction of the astronomical observatory.

Figure 4 .
Figure 4. Map of candidate sites for the astronomical observatory covered with DEM (marked in red)).
3-year image data (between 2020-2022) of these datasets were analyzed.Dealing with over 2000 images provides reliable information in the spatial domain, Koc-San et al. (2013).The 3-year meteorological dataset provides information reliability about the current meteorological condition of the study area and minimizes meteorological changes due to global warming.Statistically, 3 years of data (1095 data from the combination of 4380 granules) is satisfactory for this type of analysis, Koc-San et al. (2013).

Figure 5 .
Figure 5.The number of cloud-free days obtained as a result of the 3-year data analysis.

Figure 6 .
Figure 6.Maps of regions and the number of cloud-free days within the provincial borders of Erzincan.

Figure 7 .
Figure 7. Candidate areas (green) and areas with wind speeds greater than 11m/s (red) on the wind speed map.)

Table 2 .
Saaty & Tran (2007)mparison matrix and calculated weights.Saaty & Tran (2007).In our study, the C.R value was calculated as 0.068.The matrix obtained by the experts through pairwise comparisons (see tab. 2) is consistent and can be used in the analysis phase.