New climatic zones in Iran: a comparative study of different empirical methods and clustering technique

Recently in agricultural and industrial sectors, researchers have started to classify the climate of a region using empirical methods and clustering. This study aims to compare four empirical approaches to climate classification (Thornthwaite and Mather, De Martonne, the Extended De Martonne, and the IRIMO (I.R. of Iran Meteorological Organization)) with Ward’s hierarchical agglomerative clustering applied to the climate of Iran. The dataset used in this study comprises maximum and minimum temperatures and precipitation data of 356 weather stations extracted from IRIMO’s databases. Thirty-five synoptic weather stations are selected among 356 stations. These stations are selected regarding the best uniform distribution, elevation, windward and leeward sides of the mountain ranges, and availability of a continuous 50-year data (1966–2015). Compared with the other three empirical reference methods of climate classification, the Thornthwaite and Mather method emphasizes the role of water bodies and air masses in determining the climate type of a region. Highlighting these two factors is identified as the main advantage of this method over the other three. This advantage is the most noticeable for the highlands/mountainous regions, in the vicinity of the Zagros Mountains, and in the western regions of Iran. As a case in point, while in the De Martonne and the Extended De Martonne methods, the Zagros storm cell is climatically classified similar to patchy areas in Caspian Sea coastal zone, this cell is correctly identified as a separate zone in the Thornthwaite and Mather method. The results also reveal that the clusters obtained from Ward’s algorithm are comparable to those of empirical climate classifications, particularly Thornthwaite and Mather method.


Introduction
A comprehensive understanding of the climate characteristics in different regions is essential for environmental planning to achieve sustainable development. Climate classification is an effective way for understanding climate attributes (Aparecido et al. 2016;Piri et al. 2017). Climate classification is an attempt to identify the climatic differences and similarities of various geographic regions and to discover the relationships among different components of the climatic system. Recognizing climatic zones has been a concern for scientists from the time of Ancient Greece (Feddema 2005), when Parmenides made the first global climatic classification based on the subtropical and polar circles in 1 3 500 BC. Hipparchus also made another classification based on the solar inclination angle in the summer solstice. Later, Ptolemy classified the world into seven climates based on latitude.
Development of different types of climate classifications continued with empirical relationships proposed by Köppen (1900), De Martonne (1941, Thornthwaite (1948), and Strahler and Strahler (1978). Some of these methods, such as those of Thornthwaite and Mather (1955), Köppen (1931), and Extended De Martonne (Khalili 1973), were modified by adding new climatic components. Thanks to recent advances in statistical methods, researchers have added new variables to identify the climates of the study areas with a higher accuracy. These researchers believe that empirical (traditional) methods cannot express the climatic facts of different geographic zones due to complexities of the climate system, since they cannot be classified by examining a limited number of climate variables. Further, these types of classifications show little compatibility with the countries' biodiversity (Rahimi et al. 2013). However, comparing the climate zones of different regions using empirical methods is still the main concern of researchers studying climatic classification (Feddema 2005;Baltas 2007; Leech 2014; Mavrakis and Papavasileiou 2013;Aparecido et al. 2016;Piri, et al. 2017;Rubel et al. 2017;Vieira et al. 2017).
Many meteorological studies in Iran have focused on climate classification. The first studies based on empirical methods carried out by Ganji (1954), Adl (1960), Javadi (1966), and Khalili (1973) used the Köppen method. Despite criticism about empirical methods, these methods have garnered interest in recent years. Scientists have recently adopted a new approach for studying climate zones of Iran using empirical methods. This approach mainly emphasizes the role of climate change in the margin shifts of climate zones (Poormohammadi and Malekinezhad 2013;Rahimi et al. 2013;Siabi and Sanaeinejad 2013;Zarei 2014;Bazrafshan Daryasari et al. 2016;Ghorbani et al. 2016;Mirmousavi and Kiani 2017;Raziei 2017). As noted earlier, researchers in recent years have paid more attention to using quantitative methods, particularly factor analysis and clustering, to identify variables affecting the climate of the region under study (e.g., Heidari andAlijani 2000, Masoudian 2003).
Modern statistical methods in climate classification have tried to overcome the shortcomings of traditional methods: studies on these methods show that climatic factors affecting empirical methods such as temperature and precipitation need to be addressed as effective factors in determining climate boundaries in a modern method (Golkar Hamzee Yazd et al. 2016;Netzel and Stepinski 2016). Some studies have emphasized the effect of characteristics of the climatic variables, such as temperature or precipitation (Ghaffari et al. 2015). Dinpajouh et al. (2003) classified the climates of Iran based on precipitation using multivariate methods. They selected 12 out of 57 climatic variables and divided the whole country into 6 homogeneous zones and one heterogeneous zone. Among the former accounts, the Köppen classification system is most used classification methods. Later, the Thornthwaite and Mather classification (1955) presented an improved method in comparison with Köppen. In this method, in addition to temperature and precipitation, the evapotranspiration is considered as well. It is possible to identify the impact of moisture index on climate types in a region using Thornthwaite and Mather method. Because of these, researchers believe this method is more suitable for agricultural planning than Köppen's method (Abounoori 2010;Müller et al. 2018;Umar and Yusuf 2019). Feddema (2005) states that this classification method reveals properly the relationship between the climate variations and moisture index, temperature, and dynamic characteristics of different regions. Despite its higher precision, computational complexities have made this method less popular than Köppen's method in the literature. However, Köppen uses mean monthly precipitation to indicate relative moistness without consideration of water demand values and the most powerful factor of precipitation can only be estimated and comparisons between localities cannot be made (Abounoori 2010). In addition, Köppen's scheme ignores other factors, such as cloudiness and wind.
The climate zones in Iran are mostly identified using De Martonne and Köppen methods, and the Thornthwaite and Mather method (1955) (referred to as T&M going forward) has not been used for this purpose to date. This study adopts T&M to classify Iran's climate using empirical methods augmented with T&M (1955) while preserving homogeneous climatic classes for comparison with other climatic regions of the world. Further, De Martonne (1941), Extended De Martonne (Khalili 1973), and IRIMO climate classification methods along with a hierarchical clustering technique, Ward's algorithm, are used to provide a frame of reference to assess the performance of T&M and to determine which classification method is more aligned with the climate facts of Iran. Climatic zoning is further performed using geostatistical interpolation techniques considering the secondary variables, altitude and latitude, affecting temperature and precipitation.

Climatic features of Iran
Iran is located in the Middle East, a geographical region in southwest Asia. Iran is situated between latitudes 25° 03′-39° 47′ N and longitudes 44° 05′-63° 18′ E, covering an area of about 1.65 million km 2 (Ghaemi et al. 2012). The country is bordered by the Caspian Sea, Persian Gulf, and Oman Sea and by the countries of Iraq, Turkey, Azerbaijan, Turkmenistan, Afghanistan, and Pakistan (Fig. 1). Iran is a rugged country of plateaus and mountains dominated by the Alborz Mountains in the north and the Zagros Mountains along its western border (Zarei et al. 2013).
Iran's climate is mostly arid and semi-arid (Alavipanah et al. 2007) and is controlled by various factors. The different climates of Iran are mainly generated due to differences in latitude, topography, proximity to large water bodies, and frequency and intensity of atmospheric systems. Topography largely impacts the arrangement of different climatic regions in Iran; in particular, Zagros and Alborz Mountains determine the spatial variations of temperature and precipitation in the country. Accordingly, climatic borders of Iran largely follow the topographies.
The external factors affecting the climate of Iran are mainly the components of the general circulation of the atmosphere, including the subtropical high (STH), the upper westerly winds, the Mediterranean Sea and cyclogenesis in the central and eastern part of this sea, the polar front jet, the Sudan low pressure, and the Red Sea inverted trough, the Arabian Sea, and Indian ocean. The Rossby Waves passing north of Alborz Mountain have a major role in precipitation in South Caspian Sea coastal area. The humidity flux from the Caspian Sea heavily affects the heavy rainfall in the coastal provinces. The humidity flux from Arabian Sea, the northern region of the Indian Ocean, and the Red Sea contribute in feeding the precipitation dynamic low pressures of the Rossby waves which pass over the Mediterranean and Iraq before crossing the Zagros Mountains. The low-level stationary anticyclone over the Arabian Peninsula intensifies the feeding of warm and humid air into the warm sector of the developing or strengthening transient cyclones over Mediterranean when they move towards Iran. Heavy and widespread precipitations in Iran are usually caused by this mechanism. Orographic forcing increases the intensity and duration of precipitation in the windward side of Zagros Mountains, in western Iran. In warm season, the hilly and mountainous areas in the north of Persian Gulf and Oman Sea are affected by monsoonal south easterly winds passing over Oman Sea and Persian Gulf. Overall, Iran, with a broad geographic location, has always been under the influence of various atmospheric systems due to the general circulation of the atmosphere (Moradian 2016).
The average annual precipitation in Iran is 245 mm. The annual rainfall in the desert areas is less than 50 mm (IRIMO 2016) and over 800 mm in the south coastal areas of the Caspian Sea (north of Iran). Precipitation in 61% of the country is less than 250 mm, and in only 4% is more than 800 mm. Despite proximity to the water resources of the Persian Gulf and the Oman Sea, the southern coasts of Iran have little rainfall due to missing the required mechanisms such as polar front jet stream, to drive the formation of cloud and precipitation (Masoudian 2011). Air temperature in Iran is highly dependent on altitude, latitude, and atmospheric moisture. The effect of altitude on the air temperature is considerably greater than the latitude. The average annual temperature is 18 °C and decreases from east to west and south Fig. 1 The study area to north. Latitude affects the temperature difference between regions more than longitude does. The spatial gradient of temperature rises 2.6 and 7.8 °C for every 10,000 km from west to east and north to south of the country, respectively. The increase in temperature from west to east is due to the concentration of massive mountains in the west, and from north to south is due to increasing proximity to the equator and increasing solar radiation angles. The climate of Iran can be described by 6 temperature regions, namely, cold, semicold, moderate, semi-warm, warm (Fig. 2), and very warm, with average temperatures of 11, 13.5, 16.1, 19.5, 23.7 and 26 °C, respectively (Masoudian 2011).

Data
The meteorological data used in this study, including the maximum and minimum temperatures, and precipitation, are collected from I.R. of Iran Meteorological Organization for 356 weather stations. Based on the optimal spatial distribution and respecting to have at least one representative weather station for each non-homogeneous area and availability of a continuous 50-year data for each station (from 1966 to 2015), 35 weather stations were selected among these 356 stations for the analysis ( Fig. 2 and Table 1).
The missing data were filled in with average daily data for each station. For example, if the station's minimum temperature value was missing for a day, the long-term average (50-year period) of the same day for the station was used to fill in the gap. Data retrieval was done based on the report of the World Meteorological Organization (Aguilar et al. 2003) which recommends the missing data should not exceed 10% of the total data. Among the 35 stations, only the minimum temperature of Bandar Anzali station (located in north of Iran in Fig. 2) exceeded the 10% limit for missing data with 11.1% of missing data. Missing data in maximum and minimum temperatures as well as precipitation were mostly within the range of 0 to 1% (Table 2) for the remaining stations.

Data homogeneity test
For environmental studies, the quality and validity of the data should be considered. This is one of the most important requirements for statistical analysis in atmospheric sciences, hydrology, and related fields. The Pettit homogeneity test (Pettit 1979), standard normal homogeneity test (SNHT) (Alexanderson and Moberg 1997), cumulative deviations (Buishand 1982), and likelihood ratio test (Worsley 1979) were applied to temperature and precipitation data. Data homogeneity tests were done with different null hypotheses (Table 3). For Pettit and standard normal tests, if the p value were higher than Alpha (α = 0.05, the probability of type I error), the null hypothesis was accepted and hence data was considered homogeneous. For the other two tests, i.e., cumulative deviations and likelihood ratio, the null hypothesis was considered true if "Q" and "W" were less than the critical values 1.36 and 3.16, respectively (Table 3). . Zone 1, cold to very cold with moderate rainfall; zone 2, temperate with lots of rainfall; zone 3, slightly warm with moderate rainfall; zone 4, very warm with low rainfall; zone 5, hot with low rainfall; zone 6, very hot with very low rainfall

Performance evaluation of the models
For climatic zoning, climatic data recorded at the station scale must be interpolated. To select the best interpolation techniques, performance evaluation methods are applied on interpolated temperature, precipitation, and evapotranspiration using RMSE (root mean square error) and IOA (index of agreement). Cross-validation was used to evaluate the accuracy of the applied interpolation methods. Based on the observed and estimated values, the RMSE and IOA were calculated for each method using Eq. 1:  Table 2 The ranges of missing percentages in minimum (T min. ) and maximum (T max. ) temperatures and precipitation (P) relative to the whole data for these three datasets  Range of missing percentages (%) where N is the number of observations, E i is estimated value of the i th data point, O i is the observed value of the i th data point, and O, is the mean of the observed data.
In Eq. 2, IOA is the ratio of the total RMSE to the sum of the difference between each estimated and observed mean and the difference between each observation and the observed mean (Willmott 1982). The index of agreement ranges between 0 and 1, where 1 indicates perfect agreement.

De Martonne method
The aridity index of De Martonne (1941) was calculated based on the annual average temperature and precipitation (for the period 1966-2015) using Eq. 3: where A I is the De Martonne aridity index, P the annual average precipitation (mm), and T is the annual average temperature (°C). Table 4 shows the climate classes of De Martonne method. Khalili (1973) attempted to show more details of Iran's climate by changing the classification defined by De Martonne (1941), with respect to comparative values on a global scale. This method was developed based on the average monthly temperature in the coldest month of the year for each climate (Khalili 1973). Khalili climate classification divides each climate class into four temperature subgroups. Table 5 shows the temperature subclasses corresponding to "m."

Thornthwaite and Mather method
The Thornthwaite (1948) method is one of the best for estimating the moisture balance of a climate (Feddema 2005). Thornthwaite emphasized the humidity and potential evapotranspiration (PE) in each region for classifying climate conditions. PE is calculated using Eqs. 4 to 7.
(4) PE = 16 10T I α   where T is the monthly mean air temperature (°C), and I is the annual thermal index. α is a constant coefficient expressed by the following relationships: where i is the monthly thermal index.
To account for the actual length of the day (h) and the number of days per month (N), PE from Eq. 4 can be modified using Eq. 8: A more recent study by Nouri Mohammadiyeh et al.
(2009) has shown that the calculated potential evapotranspiration was underestimated by the Thornthwaite method (Thornthwaite 1948) for different regions of Iran. Therefore, in this study, an optimum regional coefficient was used to increase the accuracy of the calculated PE by Thornthwaite method for different regions of the country. This coefficient ranged from 0.31 to 1.43 from summer months to winter months (Nouri Mohammadiyeh et al. 2009). T&M (1955) modified the 1948 Thornthwaite method and proposed the following equation to calculate the changes in the soil moisture for classifying the climate conditions of the regions (Eq. 9).

IRIMO method
The IRIMO climate classification method (IRIMO 2016) is a widely used method which is used for sustainable development purposes in Iran. It is based on monthly average minimum and maximum temperature and precipitation data. The altitude is also considered as an effective variable in climate characteristics in the IRIMO method. In this method, six climate regions are defined, namely, cold to very cold with moderate rainfall, temperate with abundant rainfall, slightly warm with moderate rainfall, very warm with low rainfall, hot with low rainfall, and very hot with very low rainfall (Fig. 2).

Clustering
Cluster analysis is a form of data reduction that splits a set of objects into classes that are homogeneous within each cluster but heterogeneous across different clusters (Jain et al. 1999;Norusis 2010). To compare empirical climate classification methods with clustering techniques, a hierarchical clustering (Ward's method) was used in this paper. The squared Euclidean distance was employed for measuring distance, and z-score was used for standardizing the data with different scales (Kent et al. 2014). The data from the 35 weather stations were clustered based on seven features, namely, latitude, longitude, elevation, minimum and maximum temperatures, precipitation, and evapotranspiration. Each resulting cluster would correspond to a climatic group.

Trends of changes in minimum and maximum temperatures and precipitation
Homogeneity tests were applied to minimum and maximum temperature and precipitation datasets. After four homogeneity tests, data were classified into acceptable, doubtful, and suspect categories (Wijngaard et al. 2003).
The spatial distribution of changes for the average minimum temperature trend in climatic classes of Iran ranged from − 2.8 to + 2.8 °C. The highest changes along with significant upward trends are observed in northeast of Iran. The

Climate classifications
To enable classifying the climatic regions, the maps of the average temperature and precipitation for De Martonne method (De Martonne 1941), average monthly temperature in the coldest month of the year for Extended De Martonne method (Khalili 1973), and potential evapotranspiration for T&M method (T&M 1955) were acquired as described in the following subsections.

Statistical regression models
To prepare the isoline maps using Arc GIS version 10.1 for temperature, precipitation, and evapotranspiration, the relationships between temperature, precipitation, and evapotranspiration as dependent variables and altitude, longitude, and latitude as independent variables were inspected. Since the gradient of precipitation and potential evapotranspiration were not the same in different geographic regions of the country (Ghorbani and AghaShariatmadary 2014), the statistical equations were derived separately for each climatic zone. No significant correlations were found in climate zones 6 and 4 for precipitation and potential evapotranspiration, respectively; hence, these two zones were excluded from the regression analysis.
Most of the R values of the regression models are significant, but the relationship between independent variables and precipitation in zone 6 and potential evapotranspiration in zone 4 were not significant. In addition, there was a significant relation (p < 0.05) for zones 4 and 5 between precipitation and latitude and between potential evapotranspiration and latitude and height (p < 0.01) in zones 5 and 6. The results revealed that only the latitude was affecting the precipitation gradient in zone 4. Moreover, the water bodies in region could have an important influence on precipitation gradient in this zone (Masoudian 2011).

Selecting interpolation method
Different interpolation techniques were assessed using cross validation to improve the accuracy of climate zoning. Inverse distance weighting with powers 2 and 5 were selected to interpolate the potential evapotranspiration and precipitation, respectively, using RMSE and IOA (Table 7). The kriging method was chosen to interpolate the average and the minimum temperature data. Accordingly, climate variables were mapped and used as base maps to create the climate classes by different climate classification methods. However, the climate variables maps are not displayed here for brevity. Table 8 shows four climatic clusters that are created for the selected weather stations along with the number of stations included in each cluster (N). The clustering results are comparable to the actual climate zones of the country. The dendrogram of cluster analysis is shown in Fig. 3.  Figure 4 shows the climatic zones of Iran determined using De Martonne method. The arid zone covers 69% of the country, and 22% of Iran's climate is semi-arid. The smallest climate zone belongs to per-humid type B (covering 0.02 percent of Iran), with small patches in the northern part and in Zagros Mountains. Other climate types range from 1 to 3% of the total area (Fig. 4). Figure 5 depicts the climatic regions determined using the T&M method (T&M 1955). Both De Martonne and T&M methods confirm the real climate types (IRIMO 2016) of Iran. However, the T&M method provides a more detailed description of the precipitation conditions throughout the country. The patchy areas with highest rainfall in northern foothills of Alborz and western foothills of Zagros Mountains are prominent (Fig. 5) which emerge due to numerous numbers of humid subclasses in these methods. Comparing the IRIMO and T&M maps (Figs. 2 and 5) revealed that the climatic conditions of the climate classes determined using the two methods were almost similar. The prominent difference was that the dry climate class was much wider in the T&M method, but in the IRIMO climate map, this area was divided into microclimates. Moreover, the Caspian zone (north of the country) was divided into three classes in the T&M map, but in IRIMO climate classification was presented just in one class. In the T&M method, the changes in the humid zones of the Caspian Sea region were along the southnorth direction and the semi-humid to very humid classes were seen in coastal area of the Caspian Sea. Regardless of both elevation and latitude, the T&M method clearly showed the role of air masses and water bodies in diversity of humid climate zones in the country (Fig. 5). Feddema highlights this by pointing out that the T&M method not only shows climatic moisture gradients but also defines a single seasonality index responsive to mean seasonal variation in both thermal and moist conditions (Feddema 2005). Some studies have also attempted to consider the atmospheric pressure patterns using a statistical approach (Heidari and Alijani 2000;Shi and Yang 2020). They concluded that ignoring the pressure patterns in empirical methods such as T&M is a limitation of the method. Nevertheless, the effectiveness of Caspian Sea moisture and altitude was well evident in T&M classifications in Iran (Fig. 5). In addition, in this method, the effect of moisture source distribution and variations in altitudes can be seen in western foothills of the Zagros Mountain. For example,  (Khalili 1973), the country was divided into more microclimates as compared to T&M method (Fig. 6). The dry zone in central Iran was divided into three categories, including warm arid, temperate arid, and cold arid using Extended De Martonne method. However, this region is categorized into one arid climate zone by T&M method (Fig. 6).

Discussion
The results of the climate classification methods are shown for the selected 35 weather stations in Table 9. Considering the minimum temperature as the third most effective variable in Khalili method, the Extended De Martonne method (Khalili 1973) delineated more microclimates than the T&M method (T&M 1955). However, marked differences in climate types of weather stations were found between Extended De Martonne and IRIMO methods, such as in Isfahan, Kerman, Khorramabad, Zabol, and Zahedan weather stations (Table 9). Compared with other methods of climate classification, the T&M method clearly highlights the role of water bodies and air masses for determining climates of different regions. This is the main advantage of this method compared to the others studied here (Feddema 2005). This advantage is especially noticeable in highlands/ mountainous regions, in the vicinity of the Zagros Mountains (Fig. 5). Figure 7 shows the geographical distribution of weather stations colored according to the four clusters identified by the Ward's algorithm. Clusters 1 and 2 contain, respectively, the maximum and minimum number of weather stations. Comparing the results of Ward's algorithm with the other empirical methods studied in this paper shows considerable similarities between climate groups identified using clustering.

Cluster 1
Of the total 35 weather stations, this cluster had 5 members and included 14.3% of total stations. These stations are located at the Caspian Sea shore (in north) with a very humid climate (Fig. 7). This cluster has a mean annual maximum and minimum temperatures, precipitation, and evapotranspiration of 21.6 °C, 13.7 °C, 1162 mm, and 815 mm, respectively (Table 8), which is similar to the climate class obtained using the T&M climate classification (Fig. 5).

Cluster 2
This cluster had the largest number of members (15 members) and consisted of 40% of total stations. These stations are distributed in the center, southeast, and northeast of the  (Fig. 7). The climate of the cluster 2 is arid and similar to that of the T&M and the De Martonne methods. This region of the country has low precipitation, less than 165 mm, with an annual average of maximum temperature of 25 °C (Table 8).

Cluster 3
Cluster 3 was a cluster with 9 members which included the stations located in the direction of northwest to southwest of the Zagros Mountains (Fig. 7). The most prominent feature of this cluster is a mean annual minimum temperature of 6.2 °C which is the lowest temperature between clusters (Table 8). The climate of this region is semi-arid and comparable with that of T&M and De Martonne methods.

Cluster 4
This cluster consists of 6 weather stations. These six stations are located at the coast of the Persian Gulf (in south). The climate of the cluster 4 is semi-arid warm which is similar to the T&M climate classification. The main characteristic of this cluster is the mean annual temperature and potential evapotranspiration rate, 29.1 °C and 2214 mm (Table 8), respectively. The results revealed that the clusters obtained from Ward's algorithm are comparable to those of empirical climate classifications, particularly the Thornthwaite and Mather methods.

Conclusions
In this research, the Thornthwaite and Mather climatic classification system was used to classify the climate of Iran. The results were compared to those of De Martonne, Extended De Martonne, and IRIMO methods. Thornthwaite and Mather methods identified the climate classes in Iran more accurately than the other methods, especially in Zagros mountainous regions in the west of Iran. In De Martonne and the Extended De Martonne methods, the area with the most rainfall in the west Zagros Mountain was climatically classified similar to patchy areas in Caspian Sea coastal zone. This area was classified as a new zone in modified Thornthwaite and Mather methods. To address this issue, the humid B zone of the Zagros storm cell was changed to the highland/mountainous humid climate which encompasses Koohrang, Yasuj, Sardasht, and Marivan regions. In the Thornthwaite and Mather methods, the humid B class was changed to highland/mountainous humid climate to be similar to the IRIMO method as the reference method. From a synoptic point of view, one of the main sources of moisture transfer into the lower layers of troposphere is from the Arabian Sea which directly affects precipitation variability over Iran. The Thornthwaite and Mather methods are reliable for identifying the role of the tropical sea and ocean. These two elements determine the precipitation patterns over mountainous regions in the west of Iran. The comparison of climate classification methods has also showed that in addition to temperature and precipitation, evapotranspiration plays a critical role in determining real climate classes. The clusters obtained using Ward's algorithm is also comparable to those of empirical climate classifications, particularly to the Thornthwaite and Mather's method. In conclusion, since the climate classification approach used by Thornthwaite and Mather was more rational and practical for Iran, this method is recommended as a framework to lead the environmental projects for improving the food security and sustainable development goals in the region.