The habitation Model Trend Calculation (MTC): A new effective tool for predictive modeling in archaeology

ABSTRACT The aim of this paper is to create and present a new archaeological predictive model via GIS, incorporating what archaeologists consider the most important criterion absent of similar past models, that of critical thinking. The new model suggested in this paper is named habitation Model Trend Calculation (MTC) and is not only integrates the archaeological questions with a critical view, but it can be easily adjusted, according to the conditions or the questions concerning the archaeological community. Furthermore, it uses new topographical and geomorphological indexes such as Topographical Index (TPI), Hillslope and Landform Classification that give a new sense of the topographical and geomorphological characteristics of the examined area; therefore this model is a more powerful tool compared to older models that did not use new topographical and geomorphological indexes. The success of the created model is checked as a case study in the region of Messenia, Greece during the Mycenaean era. The region of Messenia is considered as one of the most important Mycenaean regions of Greece due to the great number and the importance of Mycenaean sites identified. For the present paper, 140 habitation sites were divided into four hierarchical categories (centers, large villages, villages, and farms) based on the extent and the plurality of the tholos tombs that exist in the broader region and according to the hierarchical categorization used by the archaeologists who have studied the area. The new predictive model presented in this work can assist in solving a series of criticisms that have been expressed in the previous studies regarding such models. Additionally, in the case of Mycenaean Messenia, the model shows excellent results in relation to the habitats of the time.


Introduction
Over the past decades, predictive models have been thoroughly applied in many and diverse scientific fields.The improvement of the models applied specifically in the field of archaeology has been significant in recent years.The main reasons of this progress are summarized below: • There is an increasing availability of digital information.Environmental maps with specific indexes of interest or precise topographic maps are much easier to access than they were in the past.• Modern high-resolution satellite images combined with powerful computing systems are available today, helping the formation of new research directions, including predictive models.• The existence of multiple and diverse open source softwares allows easy access and management of information to any interested user.• Finally, the adoption of the Geographic Information Systems of the archaeological field is an important parameter for the development, the adoption and often the dispute of such models.
Predictive models are projections, which constitute an expectation of where is more likely to find unregistered locations and interaction between human and landscape.Excellent publications with references to the use of GIS and the new trends concerning the modeling have been published in the recent time (Sarris and Dederix 2012).Many examples of predictive models have been used in regions all around the world (Graves 2011;Fernandes et al. 2011;Mpalla et al. 2012;Dorshow 2012;Alexakis et al. 2013;Howey 2015;Oonk και Spijker 2015;Demján andand Dreslerová 2016).
Predictive models have at times been criticized for the limitations they present.Verhagen (2007) summarized the various criticisms, grouping them in 5 significant categories and suggested that an improvement in the predictive method is needed.
(1) The archaeological data are often not adequate to provide a meaningful picture of the past landscape use.For this reason a more complete set of data should be created.(2) As for the environmental and climatic data, the contemporary conditions usually differ to those of the historic period of interest.
Therefore, the results produced demonstrate significant heterogeneity.
(3) The statistical tools used, especially those related to the correlations of factors, should be improved; many of the techniques used nowadays are outdated, therefore affecting the existing predictive models.(4) Most predictive models do not include sociocultural information, leading to biases in favor of the environmental aspects of humanlandscape interaction.
(5) Predictive models should be applied more strictly and more often; the current common practice is to only apply a predictive model once and in one specific case study.
Thus, it is obvious that predictive models do not take into account the complexity of human decisionmaking, despite the fact that there is a large and powerful set of theoretical and empirical data regarding the relation between global environmental characteristics and models of human settlements (Binford 1980(Binford , 1982;;Smith and Winterhalder 1992;Bird and O'Connell 2006).
In this paper, for the first time an archaeological Predictive Model has been created, which incorporates the critical view and perception, thus responding to the archaeological questions and covering a void of contention that exist even today among the "old archaeology" and the new technologies.
The purpose of this model is twofold, as the first objective is to "fill" this void of questioning and the second one is to guide the future archaeological Predictive Models by harmonizing and integrating the archaeological question into them.

Creation of habitation model trend calculation based on Popper's Three-world theory
The methodology proposed in this paper is based on the main knowledge of the predictive models that we use today, while trying to evolve them by refuting a number of criticisms that have been expressed in the past and to propose an overall new way of thinking about predictive models in the future.These new models should take into consideration an additional important factor: that of the insertion of critical thinking in the decision making process of each model.The new model suggested in this work is named habitation Model Trend Calculation (MTC), and is based on Popper's theory of the three worlds, as analyzed below.
In order to define the new model (MTC) in a decision-making framework and simplify it in terms of future use by researchers, a theorem was developed based on the Popper's Three-world theory, since the idea of critical thinking runs through the entire philosophical work (Popper 1978).
Popper was an Austrian philosopher and professor, and one of the most important philosophers of science in the 20th century.The main concern of Popper was the development of science.His major scientific work, entitled "The logic of the scientific discovery", was a revised and complete translation of an original work of 1934.It is a search for the internal logic of a continuous process of successive discoveries that constitute the development of scientific knowledge (Popper 1963).Many Nobel award scientists, among others John Eccles, Peter Mentauer, Hermann Bodi, Conrad Lawrence, Jacques Monod, Friedrich Hayek, and others, studied and positively commented Popper's work.
In 1978 Popper supported the theory of the three worlds giving value to the multidimensionalpluralistic view of the universe we are moving.He suggested a view of the whole (universe) that recognizes at least three different but interacting subsets.These subsets were referred to as "worlds", because they act differently but interact with each other.
The three worlds identified are: the world of physical objects (world 1), the world of the perception of the subjects, that of the subjective critical thinking (world 2), and the world of the overlap of all the derivatives of the human mind, i.e. the objective knowledge and theories (world 3).
The first world is represented by the spatial distribution of natural objects.It is the part of a world that can be measured quantitatively and qualitatively in a comparative way.The second world is related to the perception of the observer of the physical objects: each element or process is perceived in a different way from any possible subject and observer (here comes the concept of subjectivity).In the third world of objective knowledge, Popper touches all the theoretical framework of the derivatives of the human mind and its interactions through theories and theorems, stories and fiction, including any kind of information that is impersonal, stored, recorded in paintings, books, computers, art works, etc.All of them can be easily distinguished in different subsets within the world 3, for example, the world of science can be distinguished from that of the fiction world or the world of music from the world of art.
In order to make the concept of the separation of the three worlds clearer and to harmonize it in a new Universe (Predicting Model of Archaeological Positions) based on Popper's theory, which examines all the three worlds by confirming the factors of each sub-groups, the following (simplified) example related to archaeological work is indicative: • In World 3, a theory is suggested that the natural objects of the residential locations are located on hills to allow the supervision of the wider possible area (Chadwick 1976;Simpson 1981;Vermuele 1985, 125;Simpson 2014).
• In World 2, it is observed that the natural object of the settlement is located on a hill, not only to allow observation but also for defense purposes, as the hill presents steep slopes and protects from invasions of exogenous factors.• In World 1, the physical objects are examined in relation to those fixed parameters that exist as given observations of space (e.g. the settlement is built on alluvial rocks or is built on a hill).This is considered a constant observation factor, since the physical object always has been in the same space.It can therefore be used to analyze both qualitative and quantitative characteristics of the object and to reconstruct its landscape.
We can observe that the three worlds not only interact and complement each another, but also validate the observer's results and help to generate new theories by re-launching the model from the beginning."The Theory of the Universe and the Three Worlds" acts as the real image of the Universe, with the "Worlds" being in a constant motion and holding the equivalence of matter and energy of the whole (Figure 1).

Methodology advantages versus older habitation models
The central idea of a Predictive model using Popper's Three-worlds theory has tremendous advantages over previous predictive models, including the most recent publications (Argyriou, Teeuw, and Sarris 2017;Carleton et al. 2017;Oğuz-Kırca and Liritzis 2018;Eftimoski, Ross, and Sobotkova 2017;Cohen et al. 2017;Lane 2017;Healey et al. 2017) and theoretically it can constitute a guide for future models of Archaeological Sites Prediction.
The immense significance of this model, unlike all other predictive models, is that the concept of causality, i.e. the causal relationship of two situations (cause and effect), is introduced when the second situation emerges with certainty from the first.With the application of this critical thinking, the researcher not only knows the potential outcomes of the model, but also knows the exact usage of each indicator (index) used.Moreover, through the model itself, the indicators can be revalidated using the physical observations of the examination objects (in this essay settlements and monuments-World 1), through the viewpoint of the critical sociological factors of the observer (World 2) and the examined case (World 3).Thus, a direct cause-and-effect relationship is observed.
In order to further elaborate on this point, an example regarding the question of the altitude parameter, which is commonly found in archaeological publications and predictive models, will be discussed.When we examine a position in terms of its geomorphological data, the altitude quantification is usually considered sufficient.However, the elevation parameter alone does not provide with a complete image about the examined location, as an elevation site located at 100 meters could be at the top of a small hill, therefore serving as an observatory, or at the edges of a large mountain range, in which case the archeological perception is different.It could even be found in the middle of a spacious plain or between narrow valleys, so again the perceived significance of the location changes.
In our model, the precise position is fully specified based on all these geomorphological factors, in order to examine with the greatest degree of accuracy all the elements that will give the complete geomorphology of the location of the site.The exact elevation is combined with the Hillslope Classification index (which examines all likely possibilities for a particular location, from being at the top of a hill to successive subcategories until the ground level).In order to determine the accuracy of the location with regard to the wider geomorphology of the site, the above result uses the Topographical Index (TPI), which answers the question whether the sites are located in valley or reef areas, and at what level of terrain inclinations.Finally, in order to determine with absolute accuracy the position under consideration the method of Landform Classification is used, indicating whether a site is located on a hill, mountain, canyon or plain grounds.
In this way, the question of habitation position with regard to the geomorphological factor examined by the predictive model is fully answered; moreover, not only a numerical determination of the position provided, but also new questions can be generated, creating thoughts and discussions in the archaeological community.
Regarding the climatic factor, by examining the soil reports it was observed that the majority of the 140 habitation sites that examining in this paper (Malaperdas and Zacharias 2018) are located in generally southern orientations, resulting in conditions of maximum sunshine; these orientations also ensure protection from the cold, strong north winds.For the verification of the results, the wind intensity parameter prevailing in the locations was examined in correlation to the parameters of solar radiation and thermal load, for both the directness of the solar input entering the site and the cumulative effect that it may have regarding the orientation (McCune 2007).
Similar conclusions are also drawn from the geological factor and the relationship between geology, soil moisture and water availability as a function of the residential locations (Malaperdas 2019).Also, this methodology facilitates an interdisciplinary approach on research, as to some extent it standardizes factors and theorems and enables a better evaluation of the results.

Analytical Hierarchy Process (AHP) with weighted criteria and introduction of critical thinking
AHP is a popular qualitative access method with the aim of introducing objectivity in weight assignment (Barredo et al. 2000;Ayalew et al. 2005;Akgun and Bulut 2007;Ladas, Fountoulis, and Mariolakos 2007).In AHP, all factors are compared by the intensity of significance, by using a continuous scale from 1 to 9 (presented in Table 1).This scale, which is used for comparisons, enables the decision maker to incorporate experience and knowledge (Table 2).
The main aim of geographical analysis applications is to optimize decision-making by examining variables related to the phenomenon under study.Various alternative combinations of variables are examined in order to identify which ones can offer the best results.The development of Geographic Information Systems (GIS) has helped a lot to solve decision-making problems related to space planning and management, since GIS allows the combination of a large volume of geographic information.As was presented in detail in the previous chapter, one of the most popular methods of this kind is the multi-criteria analysis; a widespread version of this method is its application in combination with the method of AHP.
The AHP method is characterized by three main properties: (1) It is analytical, including mathematical and logical justification for decision-making.Thus, it helps to analyze the problem on a logical basis and to transform the thoughts and intuitions of the decision-maker into numerical evaluations.
(2) It structures the problem in a hierarchic order in order to reduce its complexity through disintegration into sub-problems.(3) It formulates an explicit decision-making process.
The experience of the expert (or the experts) is integrated into the decision-making process, which is developed on a scientific basis, making it easier for collaborative decision making.
AHP is a computational theory that uses both productive and inductive logic, taking into account at the same time many different factors, while it can be applied effectively in both the natural and the social sciences.This method is characterized by some important advantages, with the following being the most important: • Simple documentation.
• Multicriterion character.These features are particularly important in space analysis applications.According to Saaty (1987Saaty ( , 1996) ) the stages of implementation of the AHP method are the following: • Deconstruction of the studied problem in a hierarchical (or network) model, which is composed of its basic components allowing for pair wise comparisons.• Comparative assessment of each componentcriterion/sub-criterion.• Synthesis of the evaluated criteria in order to produce the final results.• Finding optimal/desired option.
The AHP methodology, despite its widespread use has been criticized, mainly because of its inherent inability to manage the uncertainty of the decisions of experts who are involved in its implementation.In the classic AHP approach, human opinions are represented as absolute values within a strict numerical framework.However, at a practical level, it is difficult for people's preferences to be expressed with absolute clarity with just one number on a specific scale (Chalkias 2015).
The software Super Decisions V.2.8, which was developed by Saaty and his team, was used in the present study.To maximize the simulation of our predictive model and to ensure that it is in accordance to the three-world theory, it was decided to obtain multiple criteria and decisions and to include them jointly into a scoreboard that examines all possible correlations between the environment and the selected habitat.This is done by creating a model that answers individual queries and criteria, thus providing the theoretical framework required to create a highpredictive model.Therefore, the weight of the factors is examined in relation to those conditions of the theoretical framework that have a direct influence on the choice of the place of residence.
The reason for this task is to achieve the highest possible accuracy of the predictive model and to strengthen it with the incorporation of critical perception.For example, when examining the likely function of settlements as observatories, it makes sense to primarily focus on the geomorphological examination parameters and more specifically to those parameters that determine if the settlements are on hills (Landform Classification), or if there is an open surveillance horizon or closed mountainous ecosystems that inhibit visibility (TPI), and of course if each settlement is on the top of the hill or at the foothills (Hillslope Classification).On the other hand, when we examine the settlements in terms of the surrounding cultivation land the most significant criteria are related to the geological factors and more specifically those geological and soil formations that help cultivation to grow (for example soil moisture, etc.) rather than the geomorphological factors of the topographic location of the area.
Therefore, tables with different criteria depending on the archaeological question were created.As a result, the diversity of the criteria according to what we examine each time is highlighted.
More specifically 12 tables were created for the evaluation of the significance of the various criteria, each of them referring to a specific archaeological question that has occasionally been addressed by the archaeological community.The significance of the criteria is examined in respect to the following parameters: (A) the defensive function of the settlements; (B) their position as observatories; (C) their position as environmental shelters; (D) their position in respect to sunshine; (E) the suitability and development of cultivation; (F) proximity to suitable building materials for necessary constructions; (G) proximity to water sources; (H) their position as a sign of power for the whole area; (I) the set of general criteria used in similar studies; (J) the overall presence of geological factors examined by this particular model; (K) the overall presence of geomorphological factors examined by this particular model; (L) the overall presence of climatic factors examined by this particular model.
Adding up the values of the weight of each criterion and dividing with the number of tables, a model is developed that does not have a one-dimensional view like most of the predictive models, but creates a complex analysis of the weighted criteria incorporating useful information of the theoretical framework.The main advantage is that it can be easily adjusted depending on the archaeological question of any future case.
Thus, for the creation and testing of the MTC various different approaches of archeological interest of the weighted criteria were tested.The individual archaeological questions of interest of the MTC regarding the place of habitation of the Mycenaean's were incorporated to the weighting table of our Archaeomodel and can be expressed by: where WM means the weighting factors of criteria; n is the number of Tables; Obs = Observatories; Def = Defence; Shl = Environmental Shelter; Sun = Sunlight; Cul = Cultivation; Str = Structures; Hyd = Access to Water; Mon = their position as a sign of power for the whole area; Glg = Geological Factors; Clm = Climatic Factors; Gmr = Geomorphological Factor; Use = General use;

Criteria weighed factors for creation of habitation MTC
By using the equation of weight calculation for the archaeological questions of the Predictive Model (Equation 1) the table of the weighted factors of the criteria for the habitation Model Trend Calculation was created.12 basic archaeological questions were used as criteria; for each of them a corresponding table of the weighted factors was developed through the "Super Decisions V.2.8" software (see Table A1-A12 shown in the Appendix), resulting finally in the creation of the table of evaluation and examination of all the conditions that influence the selected habitation and the details of everyday life (Table 3).
For the correct criterion of weighting in AHP, the primary requirement is that the sum of all weights is equal to 1.Then, after the scoring is done in pairs, the Consistency Ratio (CR) is used as the consistency used to build the matrix.The CR depends on the number of parameters and is automatically calculated by the software and is indicative of the possibility to create random matrixes (Ladas, Fountoulis, and Mariolakos 2007).Saaty (1980) stated that the CR should be less than 0.1 to accept the calculated weights; otherwise the values should be reassessed.In our case, the CR in each of the tables is significantly lower than 0.1 (Table 4) indicating a high degree of correlation for the parameters of the final table 0.0369 ≈ 0.04 (Table 5).Table 4 shows the rating results per query and the total twelve queries are as follows: Query 1: Did settlements function as observatories overlooking over the whole area?Are they located on a hill?Query 2: Did settlements have a defensive role?Query 3: Did sunshine affect the settlements?Query 4: Did settlements have a role as environmental shelters?Query 5: Were the surrounding areas suitable for cultivation?Query 6: Did settlements present proximity to structural materials?Query 7: Did settlements present proximity to the hydrographic network?Query 8: Monument function as a sign of power for the whole area.Query 9: Analysis of the general criteria commonly applied in similar predictive models.Query 10: Analysis of the geomorphological parameters.Thus, after the creation of the Final Table of Importance Criteria for the habitation Model Trend Calculation, the method of Weighted Linear Combination (WLC) was used, which is one of the most frequently used methods based on multi-criteria analyses (Malczewski 2000;Ayalew, Yamagishi, and Ugana 2004).The whole process is done in order to introduce the weighting factor of each criterion for the creation of the final predictive map based on the MTC (Ladas, Fountoulis, and Mariolakos 2007).
In the multi-criteria analysis process using the WLC method it is required not only that the sum of the weighting factors equals to one, as mentioned before, but also that the values of the weighting factors are standardized using specific numeric clusters (Table 6).

Indicators of geomorphological factors
The geomorphology of the area in regards to the habitation sites is of great interest.In older predictive models there was a tendency to use the factors of Elevation and Slope in order to describe the geomorphological parameters of a site.Contrary to that, the MTC combines a number of indicators, each with its own particular significance, to capture the full picture of the residential locations and represent it with the highest possible detail in the final predictive model.
In order to better describe elevation, all the indexes of the Hillslope Classification were used, therefore providing the exact topography of the site and also whether the site is at a higher elevation point.In order to examine whether sites are situated in low or high mountain ranges, canyons or hills, the Soil Formation Method (Landform Classification) was used.
In a similar way, slopes are not described in a simple metric method but with the use of the Topographic Position Index (TPI) or Inclination Parameter, therefore demonstrating whether the positions are in ridges or valleys and what is the level of their inclination.
Summarizing with respect to Geomorphological Factors, the factors taken into account are those of Elevation, Slope, Hillslope Classification, Landform Classification, and TPI (Εquation (2)).
where FMr = GeoMorphological Factor; HCl index = Hillslope Classification Index; Elv index = Elevation Index; LFr index = Landform Classification Index; TPI index = Topographic Position Index; Slp index = Slope Index; W i = Weight of Index.

Indicators of climatological factors
With regards to the climatological conditions it was observed that the majority of the sites are located in southern directions (Malaperdas and Panagiotidis 2018), so that maximum sunshine conditions prevailed and the sites were protected from the strong cold north winds.
For full verification, the Wind Intensity, the Solar Radiation and the Heat Load Indexes were examined, for both the directness of the solar input entering the site and the cumulative effect that it might have depending on the orientation (McCune 2007).Equation (3) refers to the calculation of the Climatological Factor (FCl).
where FCl = Factor Climate; Asp index = Aspect Index; Sol index = Solar Radiation Index; Htl index = Heat Load Index; Wnd index = Wind Intensity Index; W i = Weight of Index.

Indicators of geological factors
There are two main reasons for examining and integrating the geological factors.The first one is to examine the flexibility of the soils and their likely usage as building materials for all kinds of construction by the ancient societies.The moisture of the soil has an important role here because as the moisture decreases the permanent deformation of the soil increases; it is therefore not easy to use the soil when the ground is cracked (Tsafou and Chatziharistou 2007).
The second reason is to examine the fertility of the cultivated.It is known that the appearance of certain chemical elements, such as salts etc., is associated with the existence of specific geological formations of the subsoil.
Taking everything into consideration, the geological factors of the habitation sites are examined based on the Geological Formation, the Wetness and the Hydrographic approximate Indexes: Hyd ndex (4) where FGl = Factor Geology; Glf index = Geological Formation Index; Wet index = Wetness Index; Hyd index = Hydrographic approximate Index; W i = Weight of Index.

Creating the habitation MTC
The location of the landscapes on the earth and the residential planning of the surrounding area are reflected on the experimentation of innovative analytical techniques designed to understand the particularities of different urban and suburban conditions, even for ancient environments and societies.Nowadays the available analytical techniques can even lead to the complete and detailed reconstruction of a landscape (Agouris and Croitoru 2005).Using GIS to investigate available spatial information, it is possible to calculate sophisticated landscape indicators, which can be measured even on individual pixels (Paolillo, Baresi, and Bisceglie 2015).
For the creation of the MTC, a series of physical parameters of the environment were determined in order to define the potential of the area to support human activity.The MTC is described by Εq. (5), adapted to the needs of the study, and described as where MTC = Predictive Model; R i = the gravitational coefficient of parameter i;W i = the importance of parameter i; n = the number of parameters.
In our case and for the creation of the final map of the MTC, all the parameters analyzed in the previous sections were used, with the corresponding weighting factor.The three factors, therefore, were combined for the final result shown in Figure 2.
where MTC = Predictive Model; FMr = Factor GeoMorphological; FCl = Factor Climate; FGl = Factor Geology; Elv index = Elevation Index; Slp index = Slope Index; HCl index = Hillslope Classification Index; LFr index = Landform Classification Index; TPI index = Topographic Position Index; Asp index = Aspect Index; Sol index = Solar Radiation Index; Htl index = Heat Load Index; Wnd index = Wind Intensity Index; Glf index = Geological Formation Index; Wet index = Wetness Index; Hyd index = Hydrographic approximate Index.
In this way and on the basis of the criteria defined, this model can be used as a predictive model in areas that meet the necessary criteria-factors.The likelihood of habitation was classified into 5 categories, starting with those with the lowest fulfillment of the criteria of the model and moving towards those with the highest.The latter are the areas with the highest probability of habitation based on this model.

Predictive map of the MTC model
The final predictive map therefore shows the habitation suitability index of the model (Figure 3).In order to check the model, archaeological data was first used, which was placed on the final map divided by residential category.As shown in Table 7, to in order accept that the model operates properly, the first two hierarchical categories of centers and large villages must be included in the 5th category (High Probability) and the 4th category (Moderate to High Probability), that is the categories that present the highest habitation probability from the Mycenaeans.For the better understanding of each calibrated category, a column was created with the percentages of the total surface occupied by each category in the whole geographic surface of the study area.

Checking the MTC predictive model based on archaeological data
For the control of the model, all settlements were analyzed divided by category (Table 8, Table 9, Table 10, Table 11).The model proved to work in a satisfying way, as the vast majority of all positions (137 out of 140) are clustered within the higher probability categories of the model.That is true even of the sites which belong to hierarchically inferior categories (Categories 4 and 5).
More specifically, in a total of 140 residential locations, 64 of them (45%) belong to Category 5 -High Probability (including all sites of the first residential category, that of the "Centers").Considering that Category 5 sites occupy only a 9% of the total surface of the map, we can easily understand the high prediction success of the model.
In the hierarchically second category of the predictive Model, i.e.Category 4 -Medium to High Probability, 76 settlements are included (53% of the total number of settlements).
In Category 3 -Moderate Probability, only 2 settlements appear: the "Village" Filiatra-Stomio and the "Farm" Chalvatsou-Castro, where as in Category 2 -Low to Moderate Probability, there is only the "Village" Misorrahi in Flesiada.
It is worth noting that, although the first category of the model with the lowest probability occupies 22% of the total surface of the study area, no settlement was attributed to it.
For the simple representation of the results a polar coordinate system was used to illustrate both the qualitative information (probability of event) and the quantitative information (density of the result).The circle on the polar coordinate system, shown in Diagram 4, demonstrates the direction of the event, thus "showing" at each end the probability based on the calibration of Table 7.At the same time it shows in which category most positions (sites) are observed depending on the intensity of the color (the darker areas correspond to a higher density of sites and the lighter areas to a lower density).Combined with the results of descriptive statistics, we can see important aspects such as the mean direction, which can be extremely useful if we want to see the exact rotation trend of the results in Figure 4.
For each residential habitation category the following results are presented.
• Category of Centers Habitation Sites (Figure 5) • Category of Large Villages Habitation Sites (Figure 6).

Conclusions
For the development of this model, each habitation site was specified and a series of factors were used to provide the highest possible accuracy of Geomorphological and Geological characteristics, as well as Climatic conditions for each test site.An important advantage of the model is that based on the specific methodology developed, access to interdisciplinary of the queries is easier and  faster, since as a matter of fact, the factors have already been standardized, thus allowing for a better evaluation of the results.Also, with the creation of MTC model, presented in this paper, we aspire to create a new class of predictive archaeological models, which include and incorporate the critical thinking within them, having among other things, the great advantage of alternation.In this way, the MTC model aims at the easy adaption and integration of the archaeological query into different civilizations and chronological periods.
To sum up the MTC Predictive Model can significantly assist in solving a series of questions and criticisms that have been expressed in the past regarding such models.
• In addition to the statistical parameters commonly used (such as the deviation or average value), which can often be difficult to interpret by the reader, the MTC model highlights the attractive characteristics of landscape variables (such as the multiple facets of the relief), making them more practical in human perception.
Table 11.• This methodology follows variations of landscape variables instead of being based on estimations or series of points, as is usually the case in such models.
• The data continually revalidate both the theory and the relationships between them.
• A specific theoretical prediction for human decisions on previous land usage is being captured.
• It can be easily adapted to different local variations, depending on the prevailing conditions in each landscape.This allows the adaptation of the method to any kind of archaeological research.
• Simplifies the archaeological question.The search for causality gives new research dimensions to the archaeological community.
• It uses a new pluralistic approach of integrating the theoretical and critical thinking with the observed natural characteristics, while responding to the cultural and social questions that arise.
• Significantly facilitates researchers to determine which parameters to examine in a local context, taking into account the function of GIS.The potential to use standardized procedures is particularly important and is in constant demand for best practice determination (Boskovic 2015).
Apart from that, we observe that in the case of Mycenaean Messenia, the model is showing excellent results in relation to the habitats of the time.More specifically:  • It presents exceptional accuracy as it recognizes 137 out of 140 total residential locations in the highest habitation probability (Categories 4 and 5).
• It is especially effective for the first and hierarchically most important category of settlements, that of the Centers, as it clusters all positions in Category 5.
• It can be easily adapted to any specific question that interests the researcher, incorporating and creating a more complex, critical thinking in the final predictive model.

Figure 1 .
Figure 1.Flowchart of integration Popper's theory in the predictive model of archaeological sites (Malaperdas 2019).

Figure 2 .
Figure 2. Creation of the final MTC predictive map.

Figure 3 .
Figure 3. Predictive map of the MTC model.

Figure 5 .
Figure 5. Polar plot diagram and model predictive map for the centers category.

Figure 6 .
Figure 6.Polar plot diagram and model predictive map for the large villages category.

Figure 7 .
Figure 7. Polar plot diagram and model predictive map for the villages category.

Figure 8 .
Figure 8. Polar plot diagram and model predictive map for the farms category.

Table 1 .
The basic scale for decision-making.

Table 2 .
The weighting factors of various criteria.

Table 3 .
Final weighted matrix of MTC predictive model.

Table 4 .
Rating results per query.Analysis of the geological parameters.Query 12: Analysis of the climatic parameters.

Table 5 .
Final criteria weighted matrix of MTC.

Table 7 .
Calibration of MTC predictive model rates.

Table 8 .
Table of center sitespredictability.

Table 9 .
Table of large villages sitespredictability.
Table of farms sitespredictability.

Table A2 .
George Malaperdas carried out his PhD at the University of The Peloponnese.He is a member of the Laboratory of Archaeometry in Kalamata focusing on the development of new technologies (GIS-spatial analysis, topographical surveys, data bases and data entry, remote sensing and photogrammetry).Since 2013 he has tutored in undergraduate courses of Archaeometry, as well as the MSc in Cultural Heritage Materials and Technologies.Nikolaos Zacharias is a professor and the chairman of the Department of History, Archaeology and Cultural Resources Management at the University of The Peloponnese.His field is the use of luminescence for dating and authenticity testing of archaeological and geological materials, the analytical characterization of glasses and pottery and environmental dosimetry studies.Examination of the likely defense function of sites (WM_Def).

Table A3 .
Examination of the Sunshine of sites (WM_Sun).

Table A4 .
Examination of the sites as environmental shelters (WM_Shl).

Table A1 .
Examination of the likely function of sites as observatories (WM_Obs).

Table A5 .
Examination of the sites on Cultivation (WM_Cul).

Table A6 .
Examination of proximity of the sites to structural materials (WM_Str).

Table A7 .
Examination of proximity of the sites to water supplies (WM_Hyd).

Table A8 .
Examination of the sites as a sign of power for the whole area (WM_Mon).

Table A9 .
Examination of the sites in regards to general criteria commonly used in other predictive models (WM_Use).

Table A10 .
Examination of the Geomorphological Factors of the sites (WM_Gmr).

Table A11 .
Examination of the geological factors of the sites (WM_Glg).

Table A12 .
Examination of the climatological factors of the sites (WM_Clm).