Spatial-temporal trends analysis of flood events in the Republic of Armenia in the context of climate change

The primary goal of the study is to analyze the spatial-temporal trends and distribution of flood events in the context of climate change in Armenia. For that purpose, some meteorological parameters, physical-geographical factors and the flood events data were studied for the 1994–2019 period. The linear trends demonstrate an increasing tendency of air temperature and precipitation. Those trends expressed increased flood occurrences, especially for the 2000s, whereas the Mann– Kendall (MK) trend test reveals no significant change. The number of flood events reaches its maximum in 2011 with its peak in May. Out of 191 flood events, half of the occurrences are recorded in the flat areas and southern aspects of the mountains (22% of the country’s territory). There is a certain clustering of flood events in the areas with up to 5 slopes (66% of flood events). The most flood vulnerable areas were analyzed and mapped via Geographical Information System (GIS). The GIS-based mapping shows the location of flood vulnerable areas in the central, northern parts of the country, and the coastal areas of Lake Sevan. Our methodological approach elaborates the localization of flood-prone sites. It can be used for the flood hazard assessment mapping and risk management.


INTRODUCTION
Natural hazards are physical events that can cause significant damage to the natural and human environment.
Endogenic or exogenic processes such as active tectonics and climate changes are capable of changing landforms and triggering natural hazards, which in some cases, control human activities (Bathrellos et al. ). One issue worth considering is the fact that climate change is likely to alter the frequency and severity of extreme events. Changes in extreme events could cause more significant consequences in terms of short-term impacts on the ecosystems and human livelihoods than changes do in average values (Vastila et al. ). is known for its frequent flood cases that result in material loss, destructions, mud runoffs, landslides and unfortunately, death (Vardanyan , ). The number of various natural hazard types that occur on the territory of the RA exceeds 15, the most common of which are hails, strong winds, earthquakes, snowfall, floods, mud-flows debris and rockfalls, and landslides. Floods constitute up to 9% of all natural hazard events occurring during 1997-2003. This impact needs to be considered thoroughly (Hovhannisyan ). Therefore, floods are considered as one of the most severe and most frequent water-induced natural disasters, causing major damage to habitat, infrastructures, and properties worldwideregardless of geographical or hydrological locations and having direct economic impacts Many complex factors influence the dynamics and impacts of a flood formation, which must be addressed by a multidisciplinary approach and considerable holistic efforts. Some physical factors include basin geography and steepness, soil characteristics, vegetation cover and makeup, land use, and location of urban features such as roads, buildings, and houses. The basin geography is important, especially for the mountainous regions where rainfall is easily transferred and channeled down to streams and rivers thanks to gravity. This setting can cause water levels to rapidly increase, thus creating a higher risk for flooding (Pollak ).
According to Balasubramanian () ). Flood events threaten more than 30% of the country's area, and the average annual damage to the population and socio-economic system is about 4.6 million US$.
During the spring rivers flooded agricultural lands, settlements, roads, railways, destroying bridges and coastal barriers (Hovhannisyan  A catalogue of flood events of the last three decades  has been analyzed to support our study․ These data, along with meteorological and geomorphic parameters, such as slopes and aspects, were evaluated to record the temporal and spatial distribution of flood events for the study area. The GIS tool was used to perform complex analysis, processing, evaluation of the meteorological data, along with the geomorphic parameters and flood events. In recent years, flood susceptibility and hazard mapping have been done through extensive application of remote sensing (RS) data and GIS tools (Chapi et al. ).
Thereby, the study of flood formation factors and the most flood-prone areas in Armenia are considered as the main objective of this work, taking into account the actuality of the issue and the referenced findings above. Although there are many studies on flood formation of rivers and their risk assessments for RA, still complex analyses in combination with thematic mapping are missing. Moreover, climate change studies, that are rather comprehensive in Armenia, are not somehow linked to transformations of a flood phenomenon. Therefore, the specific objectives of this study are (1) to explore and evaluate air temperature and atmospheric precipitation trends, (2) to analyze temporal dynamics and spatial distribution of flood events, (3) to classify the flooding zones according to the observed number of floods in each river basin, and (4) to identify the impact of slopes and aspects on flood events in the river basins.

STUDY AREA
The RA is a typical mountainous country in the South Caucasus located in a northeastern part of the Armenian Highlands. The high mountains and deep gorges, broad river valleys and volcanic plateaus are following each other in the country. The average height above sea level (a.s.l.) is 1,830 m ( Figure 1). The highest altitude of the country is the volcanic peak of Mount Aragats (4,090 m a.s.l.). The lowest altitude is in the lower reaches of the Debed river -380 m. About 90% of the country is located up to 1,000 m a.s.l., approximately 75%up to 1,500 and over (Baghdasaryan ).
The RA has been adopted as a study area thanks to the following factors: first of all, the negative impact of global climate change is studied rather comprehensively in the RA (Armenia's Third National Communication on Climate Change ; Galstyan et al. ). Secondly, during the last decades, flood events became more frequently observed in the RA (Hovhannisyan ). Thirdly, the damage caused by floods calls for immediate attention due to its increasing trend as well. Thus, the study of spatial and temporal  Armenia is a country of climate contrasts. Due to complicated topography and microclimatic particularities, significant climate diversity is observed through neighboring small watersheds. The country includes almost all climatic types starting from arid subtropical to cold high mountainous (Nersisyan ).
According to Armenia's Third National Communication on Climate Change (), the average annual air temperature for the given 1920-2014 period is 5.5 C (it is below zero at altitudes above 2,500 m a.s.l.). The highest annual average temperature is 12-14 C. The summer is hot in valleys and temperate in mountains: the average temperature at the end of July is 16.7 C, while in the Ararat valley it ranges between 24 and 26 C. The recorded absolute maximum temperature is 43.7 C. Winters are cold. January is the coldest month, with an average temperature of 6.7 C.
The recorded absolute lowest temperature is À42 C. Winters In the RA, the rivers belong to the Araks (76.4% of the territory) and the Kur (23.6%) river basins ( Figure 1).
There are 380 rivers with more than 10 km length in the country. The main sources of water in the rivers are snow and rainwater, which constitute an average of 54%, and for individual rivers (Aghstev, Pambak), they reach 65-70%. For the rivers located in volcanic areas, underground supply plays a significant role (Baghdasaryan ). The water resources of Armenia are formed very unevenly in terms of spatial and seasonal distribution. They are scarce, particularly in the densely populated Hrazdan River watershed in the central part of RA. Around 50% of the total volume of the rivers' runoff is subject to significant annual variations: the flow in water-scarce years amounts to less than 65% of the average. In addition to annual variations, there are also significant seasonal variations in the river flow. Around 55% of the total river runoff on average comes from spring snow melts and rainfall. The ratio for maximum to minimum flow can reach 10:1. In most of the rivers in Armenia, the maximum flow has shown a 3-5% declining trend, except for Aghstev, Hrazdan, Marmarik, and Dzknaget Rivers, where a very small increase in the maximum flow is observed. In recent decades , the intensity and frequency of hazardous hydro-meteorological phenomena have increased. In the last 30 years, the total number of hazardous hydro-meteorological phenomena (as well as floods) increased by 1.2, and in the last 20 years it increased by 1.8 per year (Second National Communication on Climate Change (SNCCC) ).
In the basin of the Kura river, 15 mm and more precipitation are sufficient for flood formation, while in the basin of the Araks river it is 20 mm and more (Margaryan ). The reason lies in differences of geologic formationin the Kura river basin, sedimentary and metamorphic rocks are prevailing, while the Araks river basin is mostly covered by volcanic rocks (Baghdasaryan ).
In the RA, the maximum precipitation is observed in spring months and early summer, due to which river flooding mostly coincides with this period. At the beginning of April, in most of the cases, air temperature increases, heavy rains in the high-altitude zone cause snow melting and the rivers' levels rise sharply, and the flow of the rivers grows accordingly.
The flood period of the rivers begins in the second half of March and continues until June, and sometimes until July. The average duration of the flood period is 80-120 days (maximum 150 days and minimum 60 days). On the territory of the RA, on average, about 58% of 6.9 million m 3 river runoff forms in the spring months. For this period, due to a sharp increase in temperature and seasonal heavy rains, the water levels in rivers may rise by 100-150 cm during a day. The largest daily fluctuations reach up to 3-5 m for large rivers -Araks, Akhuryan, and Debed and 1.5-2 m for smaller rivers (Hovhannisyan ).
The hydrometric and hydrological features of the relatively large rivers of the RA are shown in Table 1. As could be observed, the average altitude of the river catchments is more than 2,000 m a.s.l., which means that the stability of the rivers' level mostly depends on the melting snow, which, on the other hand, depends on temperature variations during a year.
As typical mountainous rivers, they have different water discharge supply during the year (for example, snow melting from low elevations to the top of the mountains continues for months and thus provides yearlong runoff even if the rain is not sufficient). Thus, instability of the river's level, in this case, depends on precipitation, which varies from time to time and from place to place.

Materials
The official meteorological data of the Armenian Hydrometeorological Center (air temperature and precipitation data) were used for this study. The observed flood data were pro- The datasets used for analyses were checked for data quality, screened for data errors, and the correlation had been calculated and performed by the Armenian Hydrometeorological service before giving them to us. The human regulated watersheds, e.g., one of the largest rivers in the central part of Armeniathe Hrazdan river basin with an area of 2,310 km 2were excluded from this study.

Methods
The methodological approach applied in this study consists of the following logical steps: the temporal statistical analysis, the spatial statistical analysis, and the evaluation of flood-prone areas in Armenia ( Figure 2). For spatial distribution analyses, the ArcGIS toolbox was applied. The ArcGIS toolbox is an integrated application and provides a reference for the toolboxes to facilitate user interface in ArcGIS for assessing and organizing a collection of geoprocessing tools, models, and scripts (Chapi et al. ).
The MATLAB was used for temporal statistical analysis (standard deviation, correlation coefficient, the fitting formula, slope of trend, etc.) and graph representation. Thus, for analyzing the trend detection and statistical significance of the studied parameters, the above-mentioned methodology gave accurate and reliable results. The flowchart of the methodology is illustrated in Figure 2.
The records of past floods since 1994 available in the MES database were used. This information was classified as required and used to create different types of maps.
As a primary step, a temporal statistical analysis was performed to identify trends in air temperature, precipitation, and flood events during the study period. Then, the spatial distribution of flood events was examined in the territory of RA.
Additionally, to identify the seasonal distribution of flood events, these parameters were analyzed for each month. and an input value that was larger than 5, such as 5.01, would be assigned to 200.
Moreover, using the 'Extract by Mask' tool from the same section of the Arc Toolbox made it possible to identify the matching territories according to our classification, which were considered the most vulnerable zones. In this case, the automatic classification was selected as a basis.
Thus, areas having a slope up to 5 , conditional flat areas and southern aspects of mountains were considered to be the risky flood areas.
Thus, according to the principal approach, layers of two risk assessment factors must be compared and the matching areas should be considered as the most dangerous zones for flooding in Armenia.

Results
Although the occurrence of flooding is often associated with atmospheric precipitation, it has been acknowledged that Moreover, their changes were characterized from 1994 to 2019 (Figures 3 and 4).
According to the analyses, the average air temperature The temporal trend analysis of flood events Simultaneously, with air temperature and precipitation, flood events data were examined during the same study period ( Figure 5). In Figure 5, the normalized data were In contrast, the correlation between temperature and flood events is almost impossible to find (with III-XI months, it is 0.34, with an annual temperature of 0.2).
From all the above, it could be concluded that even the correlation is not very significant, but as illustrated in     from April to June․ Moreover, during this period, heavy precipitation and snow melting (as the temperature was gradually increasing) were happening at the same time.
Heavy precipitation is also considered one of the most significant reasons for the mountainous snow melting process. They are often associated with an air-mass process in the warm period (April-October). The contributions of circulation types during the heavy precipitation events are associated with the cyclonic activities. It is worth noting that those synoptic processes are known to be flood generating systems (Gevorgyan ).
Thus, it appeared that increasing air temperatures and precipitation contributed to the growth of river water discharge. Therefore, the risk of possible floods increased. As a result, the snow was not accumulated in many river basins and from time to time, it melted due to increasing air temperature in winter. It gradually melted from early spring, and the possibility of an extreme maximum runoff in early spring increased the risk of the occurrence of floods as well.
According to the results gained from linear trend analysis, we have increasing trends for all studied parameters. For deeper understanding and to obtain appropriate results, we applied the MK nonparametric test to analyze the temporal trend significance of air temperature, precipitation, and flood events.
The temporal trend analysis using MK nonparametric test Trend analysis of the observed air temperature, precipitation, and flood events data was carried out using the MK test and the results are shown in Table 2. If the used data have values P < 0.05, they are statistically significant. As it is given in Table 2, the statistical analysis of the data showed that P is 0.07 and 0.2 for annual air temperature and atmospheric precipitation, respectively. It was only statistically significant for III-IX-month temperature (0.02). The statistical analysis of the flood events showed that P is equal to 0.14 and it is not statistically significant.   (Table 3). A spatial database was created, and ArcGIS 10.7 software was used to process the collected data.  Thus, the analysis revealed that the highest value of flood events was observed in the Hrazdan river basin - The second factor was considered relief condition; slope degrees were more than 20 (Figure 9(b)), which means less flood favorable conditions according to our study (Table 5). As for the south, southwest, and southeast of Aragats mountain, the most significant factor here is a massive volcanic shield and deep river canyons (such as the Kasakh river).
Naturally, there are two types of factors that affect flooding: stable and variable (changing from time to time and from place to place). We have already studied two important factors, air temperature and precipitation, which were considered important for flood formation. On the other hand, the stable flood creator factor is relief, which has a significant role in the formation of runoff and, hence, flooding in rivers.
Thus, the geomorphological setting of an area affects applicable for the relief of the RA. Hence, the slope map was classified into the following five categories: 0-5 , 5-10 , 10-20 , 20-30 , and 30-70 (Figure 9(b)). The aspect map shows the slope direction and identifies the flat areas that have no slope. Thus, the aspect map was categorized into nine classes: flat, north, northeast, east, southeast, south, southwest, west, and northwest (Figure 9(a)). The aspect identifies the downslope direction of the maximum rate of change in value from each cell to its neighbors. It could be thought of as the slope direction. The values of each cell in the output raster indicate the compass direction that the surface faces. It is measured clockwise in degrees from 0 (due north) to 360 (again due north), coming full circle. Flat areas having no downslope direction are given a value of À1. Figure 9 shows the slope and aspect maps and their categories.
The lowest and flattest areas were mainly located in the low stream of drainage basins of the Hrazdan, the Vedi (which flows through the Ararat valley), the Arpa, the Aghstev, the Debed rivers, and in Lake Sevan coastal zone. Generally, the lowest and flattest places of river basins, in this case, also in Armenia, are considered one of the densest populated areas.
Thus, flooding damage is much higher and more diverse in these valleys than anywhere else.
As is already mentioned, according to the analyses, we separated nine different aspects and the flood occurrences in each aspect (Table 4). The 'Select by location' tool is used from the Arc Toolbox to calculate the number of floods on each side. Thus, maximum floods are seen in flat and southern aspects (29.8 and 21.5%, respectively). This means that more than half of the observed floods are located in the flat and south aspects.
The appearance of more floods in the south aspects of the mountains in the RA is mostly conditioned by the southeastern invasion of warm air masses from the Arabian deserts in early spring (Surenyan ), that creates favorable conditions for the intensive snow melting  process. Moreover, the south aspects have more floods because of more sunlight energy. Hence, the snow melting process (mostly at the beginning of spring, simultaneously increasing air temperature and occurring rain) was happening much more intensively than in other aspects of the mountains. It is worth mentioning here that the southern aspects of the mountains are windward side and get more precipitation than the other aspects (Galstyan et al. ). This is especially noticeable in the northern and the northeastern part of the country, where the most flood events were registered.
As a result, it was discovered that 50% of the observed floods occurred in the flat areas and southern aspects during the last three decades (Figure 10(a)). Furthermore, they cover only 13% of the territory in Armenia. Mainly located in the lower parts of the valleys, the largest area was Ararat valley and the valleys located on the southern slopes of the mountain.
On the other hand, studying the slope case, it became clear that there are some regularities here as well. Almost 66% of floods are observed at the slopes up to 5 , which covered 34.6% of the country's area, mostly with a dense population ( Figure 10(b)). The second regularity is class of slope from 5 to 10 , where almost 20% of the flood occurrences are happening (Table 5). Hence, it could be said that less sloped mountains are less risky. Based on these results, the maps were drawn to see the position of flat, southern aspects and slopes up to 5 separately (Figure 10(a) and 10(b)).
However, this was considered insufficient and it was decided to evaluate the joint impacts of the slopes and aspects ( Figure 11). Thus, we combined those two values

DISCUSSION
Though the data performed and analyzed in this study were taken from reliable and official sources some uncertainties could be found. The standard deviations from air temperature and precipitation were low which indicated that the data points tended to be close to the mean, while the standard deviation for flood events was higher than mean.
This indicated that the data points were spread out over a wider range of values. The uncertainties that may affect the findings of this study were connected with flood events data. The possible explanation was the fact that the official data were registered only from the residential area (such as villages or cities), where flood events have been observed.
Thus, the flood events that happened in nonresidential areas were missing in this study. June is the second month with many flood records ( Figure 7).
The study area is subdivided into 14 sub-basins to examine the spatial distribution of flood events (Table 3) Furthermore, other reasons favoring the flood genesis due to human activities are bridges with inappropriate height and the construction by the farmers of 'handy' barriers in the river channel for the storage of irrigation water. The best example of this is the river Hrazdan, which is more or less regulated by people.
In the study area, the endogenic processes, such as active tectonics, do not seem to be affecting the flooding hazard within a short period of time. On the other hand, exogenic processes, such as climate change, may significantly influence the formation of future flood events.
Expected climate changes will cause annual and seasonal temperature increase and will bring precipitation changes in Armenia. According to the 'Medium Impact Scenario', the average annual air temperature will increase by 2. In addition to air temperature and the amount and intensity of precipitation changes, slopes and aspects are considered to be among the most important flood determinants. Although we found out that only 5% of the study area (which is equal to 1,500 km 2 ) was more vulnerable to flooding, the possible temperature and precipitation changes will certainly increase in this area.
Overall, in the RA, the frequency and destructive force There is a likelihood that the risk of flood events will continue in the coming years as an effect of climate change. Despite a number of actions and plans executed in the past few decades, there is still much to be done.
More concerted efforts are required in the direction of climate change management. Plans related to the control of flood effects in the RA need to be intermingled with climate change management efforts by employing better planning policy. There is also a need to learn from the experiences gained in the past. Consequently, considerable territories of the RA will appear at flood risk, which will require significant extra costs. Proper land use for specific areas (i.e., agricultural usage, residential areas, etc.) have to be determined. The study results will be communicated to the Ministry of Emergency Situations of Armenia (Disaster Management Agency) to guide the disaster risk reduction activities of floods.
In the future studies and analytic projects integration of MATLAB, the MK test and GIS should be applied. To foster improvements of this methodology related to flood analyses, it is planned to implement a systematic mapping so as to gain a more detailed classification of the related works and to present a useful guide to aid in the future flood management plans. The only limitation for now is considered flood events data registration methodology, which must be more operative and accurate.

DATA AVAILABILITY STATEMENT
All relevant data are included in the paper or its Supplementary Information.