Integration of NDVI Imagery and Crop Coverage Registration System for Apiary Schedule

The study was carried out in Turkey's province of Konya (38,873 km²) including its thirty-one districts (Fig. 1). The study area is located between latitudes 36°41′59.6′ and 39°16′03.6″, and longitudes 31°14′18.0″ and 34°25′37.3″. The altitude of the city is 1020 meters above sea level and the area has high mountains over 3000 meters. Because the topography of Konya includes both high mountains and plains with valuable agricultural lands, it provides a good conceptual model for an apiary schedule. Additionally, Konya has the largest agricultural lands in Turkey where much agricultural activity is conducted, and natural plant and agricultural apiary activities which enhance apiary schedule application can be determined. The study area is one of the largest mono-floral honey regions in Turkey and this characteristic enables a more accurate apiary schedule to be made.

Fig. 1

Generating apiary schedule maps requires including suitable locations for beekeeping, NDVI images, CCRS records, criteria specification and verification of apiary schedule stages. The implementation model of the study is shown in Fig. 2.

Fig. 2

Generating apiary schedules requires NDVI maps and CCRS records along with criteria maps for beekeeping suitability analysis. In total, fifteen NDVI maps were derived from MODIS images. NDVI calculation has a simple formula that uses near-infrared radiation and visible radiation (Eq. 1). The results range between +1 and −1, where values close to zero refer to non-vegetated areas (Hall et al., 2002). (1)NDVI=(NIR−Red)(NIR+Red){\rm{NDVI}} = {{({\rm{NIR}} - {\rm{Red}})} \over {({\rm{NIR}} + {\rm{Red}})}}

The MODIS images were retrieved over a 10-day period and the NDVI data (Fig. 3) were generated by using ArcGIS 10.5 software.

Fig. 3

In addition, parcel-based CCRS records were retrieved for the study area to generate a crop type map. For the purpose of defining the boundaries of the apiary schedule, the beekeeping suitability areas are used to determine suitability for apiaries and date interval. This is because every green area in the NDVI images might not refer to suitable locations and similar to this, every non-planted area might not refer to unsuitable locations, such as agricultural areas for foraging, and excluding unsuitable locations from the apiary schedule provides a more detailed and precise schedule map.

The criteria used to generate beekeeping suitability maps were specified by considering the beekeeping requirements in the field of topographic, climatic and environmental perspectives and twelve criteria were specified and included in the suitability analyses.

Elevation, aspect and slope criteria were included in the suitability analysis according to previous studies by Maris et al. (2008); Estoque & Murayama (2010); Amiri & Shariff (2012); Abou-Shaara et al. (2013); Camargo et al. (2014); Fernandez et al. (2016) and Zoccali et al. (2017). Slope and aspect data were derived from ASTER GDEM elevation data at thirty -meter resolution via spatial analyst tools in ArcGIS 10.5 software.

Precipitation was included in the suitability analysis according to Amiri & Shariff (2012), Camargo et al. (2014) and Fernandez et al. (2016) due to the importance of this parameter for beekeeping. The precipitation rate is expected to be between 1275 and 1800 mm annual rainfall to ensure optimal beekeeping (FAO, 1976; Maris et al., 2008). The precipitation data were retrieved from the database of the Turkish State Meteorological Service Institute observation stations. In total, the precipitation data of ten observation stations were used and the precipitation map was generated through density analysis in ArcGIS 10.5 software.

Physical environmental factors such as roads, railways, buildings, settlements and power lines indirectly affect beekeeping activities. Locations close to these zones have disadvantages for beekeeping, as human-related pollution, air and noise pollution, greenhouse gases, exhaust emissions and intensive traffic flow negatively affect both flora and beekeeping activities. Thus, these criteria were included in the suitability analysis to consider and avoid the negative effects of artificial structures. The criteria data were retrieved from the Open Street Database in vector format and converted to buffer zones to determine the effect zones via spatial analyst tools in ArcGIS 10.5 software.

Water resources are also important for supplying water to apiaries and for the richness of flora. The study area has 2127 square-meter water resources and can provide clean water for apiaries. Natural disasters indicate the locations that are a threat to apiaries. Areas with a high potential for floods, sinkholes and landslides were included in this criterion to avoid loss of or damage to apiaries. The natural disaster data were retrieved from the Ministry of the Interior's Disaster and Emergency Management Presidency in vector format and converted to buffer zones in ArcGIS 10.5 software.

The flora criterion is of vital importance for beekeeping activities. The flora are the decisive factor of pollen resources and define the productivity and yield. This criterion also includes the agricultural lands because most of the migratory apiaries are located in agricultural lands for clover, sunflower and sainfoin. Urban settlements and industrial areas were not included to avoid the disadvantages and adverse effects of urbanization on honey production. The flora map was derived from CORINE 2018 land cover data and classified according to the land use type considering beekeeping requirements. Because biodiversity and flower data are unavailable for the study area, land use map classifications were used in this study.

The Analytic Hierarchy Process (AHP), proposed by Saaty (1977; 1980), is a complex decisionmaking tool that considers the importance of each criterion. For this purpose, the AHP establishes an importance scale from 1 to 9 (1=Equal, 3=Moderate, 5=Strong, 7=Very Strong, 9=Extreme). The AHP is based on a normalized matrix and calculates weights with an average sum of the normalized matrix. Moreover, the AHP provides a consistency rate concept to be able to calculate the consistency of overall weights and priorities (Saaty, 1994). Finally, the consistency ratio (CR) is calculated, which should be less than 0.1 to prove that the weights and priorities are consistent (Saaty & Vargas, 1991). The AHP method was used to generate the final suitability map for beekeeping in the study area.

Beekeeping suitability map generation requires determining the weights of each criterion via the AHP method with a comparison matrix. The importance of each criterion and preference values were specified considering previous studies in literature and the decisions of thirty expert beekeepers located in the study area. The preference values are given to the criteria considering the importance (from 1 to 9) of each criterion compared to another. The comparison matrix and preference values are given in Tab. 2.

The CR of the comparison matrix was calculated as 0.081, which means that the weight calculation is consistent. Due to the high importance of the criteria of flora and distance to water resources, the weights were calculated as 34% and 13%. After weight calculation, the Beekeeping Suitability (BS) map was generated by the contribution of each criterion with its weight. The BS was calculated with the following equation:

In the CCRS records, all the parcels with sunflower crops were exported as a new layer. These parcels were then intersected with all NDVI images and merged. The result data included both planted areas outside agricultural lands and sunflower-harvested agricultural lands. A sample result of the NDVI and CCRS record map for 1^st – 10^th August is given in Fig. 4.

Fig. 4

Field studies included both specifying apiary location coordinates and apiary establishment dates, which refer to the beginning (when they are located) and end (when they are left) of the apiaries to verify both the suitability and apiary schedule. In total, 176 apiary location coordinates were collected and apiary production dates were recorded. Considering the apiary production dates, three clusters were detected, which refer to the three different seasons. The dates revealed that eighty of the apiaries were located in natural plant areas and ninety-six of the apiaries were located in agricultural lands. Moreover, in natural plant areas, the existence of two seasons was revealed with forty-five apiaries in early season and thirty-five apiaries in late season. The average beginning and end of the apiaries were calculated for each cluster. The records of five apiaries as a random sample out of 176 apiaries are given in Tab. 3. The dates retrieved from the field studies also represent the flowering dates, when the beekeepers locate their apiaries.

The apiary schedule are verified by comparing the flowering season retrieved from field studies and the existing apiary locations. Existing apiary locations are accepted as suitable because the beekeepers have decided on the locations after a long period of experience.

The apiary schedule is verified with zonal statistics by determining the vegetation index which honey bees each within a three-kilometre buffer zone. Although the forage area gives valuable information about the suitability and reliability of the apiary schedule, it is also the most appropriate data to prove the schedule's accuracy. Thus, three-kilometre buffer zones were determined for each of the 176 apiary locations, and zonal statistics were calculated to determine the total vegetation availability within all honey season dates for each buffer zone. Every pixel of the resultant NDVI maps was counted and summarized with a mean value to compare each buffer zone to another. The results for three apiary schedule zones were evaluated by comparing forage zone statistics and date intervals retrieved from the field studies. The mean values were converted to the 1-to-9 scale in which 9 represented the most intensive planted forage zones and 1 the non-planted forage zones (Fig. 5).

Fig. 5

The results indicated that 10.8% and 42.5% of the study areas were determined as highly suitable and suitable, respectively, for beekeeping activities. A large percentage of the highly suitable areas included agricultural lands. Approximately 60% of the suitability specified was accounted for by the high weight values of the flora, distance to water resources and aspect. The beekeeping suitability map of the study area is shown in Fig. 6.

Fig. 6

All the NDVI and CCRS combination maps were intersected with the suitability map, and nonsuitable areas were excluded from the apiary schedule. Thus, the generated apiary schedule maps refer to the suitable and high-intensive planted crop lands which can be used for beekeeping activities (Fig. 7).

Fig. 7

The apiary schedule map revealed the presence of two main zones which vary distinctly by seasons. The first season is for natural floral areas and the second season is for agricultural crop areas. The first season, the earliest season of the study area, starts on 1^st April and ends on 10^th May and covers the vicinity of Lake Beyşehir in the west of the study area due to the warmer climate of the region. Lake Eğirdir, which is not included in the study area but is quite close to Lake Beyşehir, also affects the warmer climate and early season of the region. The region has a large number of natural floral areas and large moors located between lakes and mountains.

The latest season of the study area starts on 1^st July and ends on 30^th August and covers the agricultural lands and represents the sunflower honey production in the southeast of the study area and south of Salt Lake. This region has the largest sunflower cultivation and most valuable agricultural lands in the study area and includes migratory apiary locations. The topography of the region is in the form of a plain and has very intensive agricultural activities.

According to the field studies, the average dates revealed that ninety-six of the apiary locations were located in agricultural lands for sunflower and eighty in natural floral areas (Tab. 4).

On the other hand, eighty apiary locations were clustered in two groups as early and late seasons, of which forty-five apiaries were located in early-season regions and thirty-five in late season regions. The apiary locations according to the groups are given in Fig. 8.

Fig. 8

Early season apiaries were clustered into three groups and located on plains. The average elevation of the apiaries varies from 1018 meters above sea level to 1046 meters, where the climatic and topographic conditions can be accepted as similar. The plant density values vary from minimum 2.0 to maximum 7.58 in the forage zone statistics. For the forage zone statistics, the maximum plant density started on 10^th April and continued until the 10^th May. For the field study results, the average start and end dates (red lines) were determined as 6^th April and 12^th May, which represents a good overlap (Fig. 9).

Fig. 9

Late season apiaries were clustered into two groups and located mostly in mountainous areas. The average elevation of the apiaries varies between 1158 and 1346 meters, where climatic and topographic conditions are quite different compared to the other apiary locations. Due to the climatic conditions in higher elevations, the flowering seasons of the plants start later than in lower elevations. The difference in elevation leads to the constitution of a second season, which is an advantage for beekeepers. The plant density values vary from minimum 2.3 and maximum 6.4 in the forage zone statistics. An evaluation the forage zone statistics shows that the maximum flowering plant density started on 20^th May and continued until 20^th June (Fig. 10). The field study results show that the average start and end dates were determined as 10^th May and 10^th June, which represents a weaker overlap than for the early season.

Fig. 10

In total, ninety-six apiaries were located in agricultural zones within two main clusters, fifty-two of them were located in the Çumra-Karapınar districts and thirty-eight in the Altınekin district. These regions have similar the topography and climatic conditions and the same agricultural crops. Plant density varies from minimum 2.1 to maximum 7.82 within the apiary schedule. The graphics revealed that the maximum plant density started on 1^st July and started to decrease on 20^th August. The field study results show that average start and end dates were determined as 8^th July and 24^th August. The plant density and apiary schedules are given in Fig. 11. The red lines represent the start and end times of the season which were retrieved from the field studies.

Fig. 11

The apiary locations and suitability values were examined to verify the suitability map, and the results showed that fourteen apiary locations intersected with low suitability and forty-two apiary locations intersected with highly suitable locations. Thus, 144 of the existing beekeeping areas intersected with suitability values higher than 6 and provided 82% intersection with existing apiary locations. The graphics of suitability values according to the existing apiary locations are given in Fig. 12., and the optimal production start suitability threshold value (6) for apiaries is emphasized with a red line. The blue line indicates the average suitability of all 176 existing apiary suitability values. The accuracy and intersection rates show that the suitability map can be used for both apiary schedules and locating apiaries.

Fig. 12

As a result, all the field studies and determined apiary schedule dates are given together in Fig. 13. The most suitable results were obtained for apiaries in agricultural lands via integrated CCRS records with regard to the increased resolution of NDVI images in agricultural lands. For the evaluation of natural plant apiaries, the early season apiaries were determined one week later and one week earlier than the field study dates. This result is also related to the NDVI imagery resolution because sparse and small plants could not be specified with this resolution. The one-week date differences from the start and end are quite significant because plant intensities are the same before and after the season, which the NDVI resolution is not sufficient to detect. Similarly, late season apiaries were calculated two weeks later than in the field study results. This result indicates the difficulty of differentiating the forest cover and natural plants in mountainous areas. However, these difficulties can be overcome through the use of high spatial resolution NDVI imagery.

Fig. 13

As seen in Fig. 8, some early season apiaries were detected inside agricultural lands, and upon examination they were revealed to be located in mountainous and natural floral areas surrounded by agricultural lands. The overlapping of these apiaries with the apiary schedule was generated between 10^th April and 10^th May. Late season apiaries were located in mountanious lands where climatic and topographic conditions were quite harsh.

Although early season apiaries in agricultural lands intersected more noticeably with the apiary schedule zones, the late season apiaries were not specified and did not intersect with the apiary schedule zones. One reason for this result is related to the resolution of the NDVI images. The late season apiaries were located on forested mountains with very small natural plants which NDVI resolution was not sufficient to detect. Another reason is that forests were excluded from the apiary schedule via the suitability map to distinguish the plant and forest covers. Thus, together with the resolution effect, these zones could not be determined in the apiary schedule. However, the resolution of the NDVI images was sufficient to detect the large moors and planted areas. Thus, early season apiaries were determined accurately because non-forest areas and large natural plant areas could be identified with NDVI imagery.