FINAL RESULTS PART 2: CREATING OF PREDICTION MODEL MAPS OF EARTHQUAKE OCCURRENCES WITH GEOGRAPHIC INFORMATION SYSTEMS (GIS) IN THE MARMARA SEA AND SURROUNDING, TURKEY

In this study, we investigated earthquake occurrences to created prediction model maps in the Marmara Sea and surrounding. So, we used Geographic Information Systems (GIS) for this maps. After, the statistical analysis estimated for catalogue and study area divided into 6 different seismic source zone based on some important information. We used deterministic methods to investigated earthquake occurrences and created probability model maps in GIS. These methods were respectively, Inverse Distance Weighting (IDW) was a quick deterministic interpolator that was exact, Global Polynomial (GPI) was a quick deterministic interpolator that is smooth (inexact), Radial Basis Functions (RBF) were moderately quick deterministic interpolators that are exact, Local Polynomial (LPI) was a moderately quick interpolator that is smooth (inexact). We calculated earthquake occurrences and compared prediction model maps of this methods. As a result, Marmara Sea zone will be occur an earthquake at 7.5 magnitude in the next 100- year, while İzmit -Sakarya zone will be occur will be occur an earthquake at 6 and above magnitude in the next 100-year to determined predictions using four different method. This investigates in the Marmara Sea can be used to predict of earthquake occurrences by some scientists.


INTRODUCTION
The 1500-km-long North Anatolian Fault Zone (NAFZ) divided two branches into three branches to the east of the Marmara Sea according to some scientists. These branches finished where the westward escape of the Anatolian Block convert anticlockwise rotational range in the northern Aegean Sea and Edremit Bay. These three branches indicated distinct kinematic and seismic aspect in the Marmara region. First of all, the NAFZ divided into two branches west of about 30.5E latitude into two branches. While the northern strand of the North Anatolian Fault expanded from Bolu to Izmit (Bozkurt, 2001), the second branch expanded southward from Bolu and divided into two branches once more in the Pamukova Plain at about 30E latitude (Koçyigit, 1988). The northern of these final branches was the middle strand of the North Anatolian Fault that expanded nearly west-east along Iznik Lake, Gemlik Bay and Bandırma Bay. It altered aspect round the western part of Bandırma Bay and rotational south-westward into a fault zone created of a lot of faults. The southern branch of the North Anatolian Fault was a northeast-southwest-trending fault extension from Pamukova. The NAFZ formed the Yenişehir pull-apart basin with other fault to the north.
An eastern portion of a 150-kilometer-long segment of the North Anatolian Fault broke beneath Turkey's Marmara Sea in 1999 year. Consequently, the Izmit earthquake occurred a 7.6 magnitude quake in Turkey and it was series magnitude and some cities and towns destroyed thousands of lives. After an earthquake, scientists investigated the main Marmara Fault to find out if the segment is a locked or creeping fault. In the course of locked segments, energy set up quietly over time, before it is delivered quickly in big earthquakes. However, creeping faults, on the contrary to, nonstop delivered energy outside significant seismic complexity.
The Marmara Sea concluded different basins (Tekirdağ, Çınarcık, Central and Kumburgaz). We plotted epicentre distribution in the Marmara Sea and surrounding (Fig. 1). Addition, we selected 6 different seismogenic zone (Table 1). We compared faults with earthquake occurrences, so we realized buffer analysis in 5 and 10 km (Fig.  2).

DETERMINISTIC METHODS FOR SPATIAL INTERPOLATION
Deterministic interpolation techniques formed surfaces from measured points, based on either the extent of similarity (inverse distance weighted) or the degree of smoothing (radial basis functions). Deterministic interpolation techniques occurred from two groups (global and local). While global techniques calculate estimated using the entire dataset, local techniques computed predictions from the measured points within neighborhoods. It are smaller spatial areas inside of the larger study area. Geostatistical Analyst included different technics as a global interpolator and inverse distance weighted, local polynomial, radial basis functions.

INVERSE DISTANCE WEIGHTED INTERPOLATION
Inverse distance weighted (IDW) interpolation clearly performed the guess that works that were immediate to one another were more similar than those that were further separate. According to predict a value for any unmeasured location, IDW applicationed the measured values surrounding the prediction location. The measured values immediately to the prediction location had more impact on the predicted value than those farther away. IDW supposes that every measured point had a local effect that reduced with space. It delivered greater weights to points immediately to the prediction location, and the weights decrease as a function of space, therefore the name inverse distance weighted. A surface calculated using IDW. So, it counted on the choice of the power value (p) and the exploration neighborhood strategy. IDW was an exact interpolator that the maximum and minimum values in the interpolated surface can only form at sample points. The output surface was susceptible to clustering and the existence of outliers. IDW methods that the phenomenon being modeled is application by local variation that can be conquered (modeled) by characterizing an enough exploration neighborhood.

RADIAL BASIS FUNCTIONS
Radial basis functions (RBFs) were a sequence of complete interpolation techniques also, the surface should crossed through each restrained sample value. Five distinct foundation functions available: •

GLOBAL POLYNOMIAL INTERPOLATION
Global polynomial interpolation fits a smooth surface that is defined by a mathematical function (a polynomial) to the input sample points. The global polynomial surface changes gradually and captures coarse-scale pattern in the data.

LOCAL POLYNOMIAL INTERPOLATION
In the meanwhile global polynomial interpolation complied a polynomial to the whole surface, local polynomial interpolation complied many polynomials, every within represented overlapping areas. The exploration areas must be described by using the size and shape, number of earthquakes and sector format. Addition, the Exploratory Trend Surface Analysis parameter can be used to simultaneously change the Bandwidth, Spatial Condition Number (if enabled), and exploration areas values. Figure 6. Local Polynomial Interpolation function maps plotted to 6 different seismic source zone in the Marmara Sea.

RESULTS and DISCUSSION
We explored earthquake occurrences to obtained prediction model maps using Geographic Information Systems (GIS) in the Marmara Sea and surrounding (between from 10 to 100 years). So, we used deterministic methods (Inverse Distance Weighting (IDW) was a quick deterministic interpolator that was exact, Global Polynomial (GPI) was a quick deterministic interpolator that is smooth (inexact), Radial Basis Functions (RBF) were moderately quick deterministic interpolators that are exact, Local Polynomial (LPI) was a moderately quick interpolator that is smooth (inexact)) for prediction magnitudes of earthquake occurrences. We calculated earthquake occurrences and compared prediction model maps of this methods. Whole prediction maps observed low magnitudes with blue colors, also observed medium magnitudes with yellow and orange colors, but observed high magnitudes with red colors. IDW function will be occur 4 and above earthquakes up to the next from 10 years to 40 years, but observed higher 5 and above earthquakes prediction up to 50 years and also, will be occur 6 and above earthquakes up to 100 years in the all zones. RBFs will be occur 4, 5, 6 and above all earthquakes up to from 10 years to 100 years in the all zones. While Local Polynomial Interpolation function won't be occur 4 up to from 10 to 40 years and above earthquake, will be occur 5 and 6 above earthquakes between in the next from 10 to 100 years. While Global Polynomial Interpolation function will be occur with higher prediction of 4 and 6 above earthquakes between 10, 20, 30, 40 and 100 years, will be occur with lower prediction of 5 and above earthquakes in the next 80 years. Consequently, Marmara Sea zone will be occur an earthquake at 7.5 magnitude with %0.5 probability in the next 100 years, while İzmit-Sakarya zone will be occur will be occur an earthquake at 6 and above magnitude with % 0.5 probability in the next 100 years to determined predictions using four different method.