3.1. Results of Analysis of Past LULC Change (1955-2015)
The analysis of past LULC change was made by comparing the change in areal size of 6 types of LULC such as agriculture, pasture, forest, open forest, settlement, and water. In total, seven LULC maps were created for each year data available (Figure 5). The LULC maps demonstrating the change around the lake where is most increase of settlements was observed are given in Figure 6.
Figure 5 about here
Figure 6 about here
The areal size of each LULC class determined by analysing of past LULC change is given in Table 2. As a typical characteristic of Black Sea Region in Turkey, forests cover largest area in the study area. The settlements are located dispersedly due to topography of the region. In other words, the locations where are suitable for construction activities and for agricultural activities are found scarcely any. The most suitable locations for the settlement and agriculture in terms of topography and in terms of getting required economic incomes for continuation of life of inhabitants could be found around the lake but have any other risks of natural hazards. In the region, in past, the around of the lake was preferred for settling by an agriculture-based population, who engaged in agriculture and/or forestry sector. In addition, there is no any industrial area in the region. While more than 84% of the population of Çaykara District had found in the villages by 2012, the population of the district and the villages decreased by 65% and 70%, respectively, due to the important migrations from the region. Even though migrations in the region settlements in the study showed an increasing trend. While settlement area covers 57.35 ha in 1955, it covers 108.38 ha in 2015. The settlement area increased 51.03 ha within 60 years by conversion fully from agriculture. However, agriculture area decreased by 96.41 ha, i.e. the remaining 43.21 ha of agriculture replaced to forest, while 2.17 ha replaced to LULC class "water". The forested areas also increased by 101.14 ha between 1955 and 2015, by the conversion mostly from pasture area (89.64 ha) and also from open forest (11.5 ha). The change speed of a LULC type from one to other could also be estimated. However, the evaluating the change trend of LULC in a way of single 60 year-period would have not been meaningful due to non-linearity of trend. Therefore, two sub-periods, one before 2004 and other one after 2004, was defined. The reason for the selection the year of 2004 as reference point in determining sub-periods was a sharp increase observed in the change trend of settlement. The increase speed was 0.24 ha/year between 1955 and 2004, whereas it was 3.57 ha/y between 2004 and 2015.
Table 2 about here
3.2. Results of Simulating Future LULC with Dyna-CLUE
Both simulated LULC of 2015 and reference LULC map of 2015 is given in Figure 7. After Dyna-CLUE was run with past LULC demands, a confusion matrix was created (Table 3). The total accuracy was 0.99, however, both user and producer accuracies of settlement and open forest were obtained relatively low. The low prediction accuracies of settlement and open forest were due to lower number of control points because they cover smaller area in the region.
Figure 7 about here
Table 3 about here
With future LULC simulations, four LULC maps, i.e., two for 2035 and two for 2050, were created (Figure 8). When simulated LULC maps of 2035 were compared, it was observed that almost all agriculture area around the lake is converting to settlements. In Scenario 2, due to area restricted to conversion around the lake, it was observed that settlement area was growth towards other locations where are suitable for settlement as expected (check the red circles in Figure 8). Accordingly, LULC maps of 2050 showed similar results. It is clearly observed that allocation of future LULC demands was made meaningfully by the developed model.
Figure 8 about here
3.3. Results of Evaluating Interactions of LULC Changes with Natural Hazards
An extended period (1955-2050) was evaluated to question how LULC changes could occur by 2050 and interact with natural hazards. Especially, settlement areas were focused (Table 4). While 16.25 ha of settlement area in 1955 was exposed to snow avalanche hazard, this grew up to 68 ha in 2015. It corresponds to 62.9% of total area of settlement in 2015. While 5.75 ha of settlement area in 1955 was exposed to rockfall hazard, 28 ha of settlement area in 2015 was exposed to rockfall hazard. In addition, 1.54 ha of settlement area in 1955 was exposed to both snow avalanche and rockfall hazard, this grew up to 10.12 ha in 2015.
Table 4 about here
The snow avalanche hazard indication map overlapped with future LULC maps are given in Figure 9. Scenario 1 resulted that settlement will grow towards areas exposed to snow avalanches, up to 97.5 ha in 2035, and 126.5 ha in 2050. In other words, the increasing in settlement area could be 43.4% and 85.7% by 2035 and 2050, respectively. Scenario 2 resulted that settlement exposed to snow avalanche hazard will grow up to 119.75 ha in 2035 and 137.75 ha in 2050. The increase in settlement area exposed to snow avalanche could be 76.1% and 102.6% by 2035 and 2050, respectively. In Scenario 2, settlement would grow more into the areas exposed to snow avalanche hazard.
Figure 9 about here
Scenario 1 resulted that 24 ha in 2035 and 28.25 ha of settlement area in 2050 were exposed to rockfall hazard. Scenario 2 resulted that 26.25 ha in 2035, 28.75 ha of settlement area in 2050 were exposed to rockfall hazard. A clear growing of the settlement towards areas exposed to rockfalls was not estimated for both scenarios. In total, 28 ha of settlement area was located exposed rockfall hazard in 2015, and simulations were mapped same areas with small differences. These differences were due to use of Dyna-CLUE with selected parameters, providing about 0.7 of prediction accuracy for the settlement (see Table 3). The rockfall propagation zone map overlapped with future LULC maps are given in Figure 10. In Scenario 2, more area of settlement was obtained as exposed to both hazards than in Scenario 1.
Figure 10 about here
3.4. Results of Analysis of Change in the Number of Buildings and Interactions with Natural Hazards
Remote sensing data provided to extract the information on how the number of buildings changed during the 60 year-period (Figure 11). In 2018, a UAV flight was also carried out for mapping actual situation of buildings only at close vicinity of the lake (Figure 12). Unfortunately, the number of buildings from 1960 aerial photography could not be obtained due to its low-quality resolution not allowing to detect buildings. Table 5 summarizes that how the number of buildings changed between 1955-2018 and how interacts with natural hazards. The number of buildings grew 2 times from 1955 to 2015, whereas it grew 2.7 times around the lake. The high ratio of increase in the number of buildings around the lake is clearly due to growing tourism activities. During the period covering from 1955 to 2015, the number of buildings exposed to snow avalanche hazard grew 2.6 times, whereas the number of buildings exposed to rockfall hazard grew 6.6 times.
Table 5 about here
Figure 11 about here
Figure 12 about here