Using OCO-2 Satellite Data for Investigating the Variability of Atmospheric CO 2 Concentration in Relationship with Precipitation, Relative Humidity, and Vegetation over Oman

: Recognition of the carbon dioxide (CO 2 ) concentration variations over time is critical for tracing the future changes in climate both globally and regionally. In this study, a time series analysis of atmospheric CO 2 concentration and its relationship with precipitation, relative humidity (RH), and vegetation is investigated over Oman. The daily XCO 2 data from OCO-2 satellite was obtained from September 2014 to March 2019. The daily RH and precipitation data were also collected from the ground weather stations, and the Normalized Di ﬀ erence Vegetation Index was obtained from MODIS. Oman was studied in four distinct regions where the main emphasis was on the Monsoon Region in the far south. The CO 2 concentration time series indicated a signiﬁcant upward trend over di ﬀ erent regions for the study period, with annual cycles being the same for all regions except the Monsoon Region. This is indicative of RH, precipitation, and consequently vegetation cover impact on atmospheric CO 2 concentration, resulting in an overall lower annual growth in the Monsoon Region. Simple and multiple correlation analyses of CO 2 concentration with mentioned parameters were performed in zero to three-month lags over Oman. They showed high correlations mainly during the rainfall period in the Monsoon Region.


Introduction
Increases in greenhouse gases, particularly carbon dioxide (CO 2 ), have accelerated the rise in global air temperatures in recent years [1]. Forster et al. showed that the global mean concentration of CO 2 and the atmospheric radiative forcing associated with it increased to a substantial degree within the last two centuries, causing further disruption in the earth's energy budget balance [2]. The atmospheric radiative forcing changes cause alterations in the atmospheric heating distribution by trapping additional heat, resulting in changes in the different elements of the hydrological cycle and atmospheric moisture content [3], which are subsequently expected to affect precipitation [4][5][6]. At the same time, precipitation, water vapor content [7], and consequently, vegetation cover have a direct effect [8] on the monthly CO 2 concentration fluctuations. example, Falahatkar et al. investigated the interaction of land surface temperature and Normalized Difference Vegetation Index with GOSAT XCO 2 data for one year over Iran [27], while Fu et al. tracked the fossil fuel consumption in urban areas and revealed the impacts of urban vegetation on seasonal variation of XCO 2 using OCO-2 observations [28]. In addition, Siabi et al. assessed the CO 2 distribution over Iran using XCO 2 data from OCO2 during the growing season [29]. In their study, land cover and wind direction showed the highest impact in CO 2 concentration over Iran. However, no previous studies on CO 2 concentration have been conducted over Oman using ground-based measurements as they lack or space-based observations. The Sultanate of Oman is a part of Western Asia occupying the southeast of the Arabian Peninsula on the Indian Ocean. The average annual rainfall in the Sultanate of Oman is about 110 mm but can range from <100 to 500 mm [30]. Although July-August is known as the "wet" or "monsoon" period in the south of Oman, this country is one of the most water-stressed regions in the world, and it is relatively sensitive to global climate change [31]. Despite the importance of investigating the atmospheric CO 2 trend in Oman and its monthly variability, no previous studies were carried out, which can be critical in predicting and adapting to the future climate.
Thus, the objectives of the current study can be summarized as analyzing the time series of atmospheric CO 2 concentration over Oman using Earth Observation Satellites as well as investigating the relationship of monthly variability of CO 2 concentration with precipitation along with relative humidity and vegetation cover changes.

Study Area
A diverse climate and geography characterize Oman. The terrain varies from the desert plains in the central region to the rugged mountains in the north and south. As for the climate, it is both hot and humid in the northern coastal plains and hot and dry in the central part. In the far south, there is an active southwest summer monsoon in late June to August [30]. Furthermore, because of complex topography, Oman has several local climates across its terrain, according to the Köppen and Geiger classification system varying from dry-arid to semi-arid conditions [32]. As a result, vegetation biomass and distribution are mainly controlled by the low and inconsistent rainfall and the fluctuating temperatures [31]. Therefore, in this study, considering the climate features and the rainfall patterns, Oman is assessed in four distinct regions, as displayed in Figure 1. "Monsoon Region" is referred to the Dhofar Mountains and coastal areas in the southwest with monsoon rainfall pattern and thus exhibit different vegetation cover compared to the rest of Oman. In this region, three distinct periods are considered during the year. Pre-monsoon, covering the period from 1 November to 19 June; monsoon, from 21 June to 30 August; and post-monsoon, from 1 September to 30 October. The second region considered is the "South Region," which despite the slight impact of the monsoon rainfall on this area, still has a different climate and vegetation pattern. The third region, called "Central Region" is located in the center of Oman and is mostly desert with limited rainfall and vegetation. "North Region" is the fourth area that encompasses the northern part of the country, which has varied precipitation and vegetation changing with elevation.

Dataset
In this study, daily precipitation and RH data are collected from 36 ground weather stations over Oman (Figure 1) for the study period from September 2014 to May 2019. As shown in Figure 1, there is a sufficient number of stations for each of the four study regions and the appropriate distribution of stations over Oman.

Dataset
In this study, daily precipitation and RH data are collected from 36 ground weather stations over Oman ( Figure 1) for the study period from September 2014 to May 2019. As shown in Figure 1, there is a sufficient number of stations for each of the four study regions and the appropriate distribution of stations over Oman.
Because of ground CO2 observations absence in Oman, the column-averaged CO2 dry air mole fraction, XCO2 (hereafter, CO2 concentration), from the Orbiting Carbon Observatory 2 satellite is obtained in the current study for the period from September 2014 to May 2019 (Daily bias-corrected OCO2_L2_Lite_FP.9r products were downloaded from https://disc.gsfc.nasa.gov). OCO-2 was launched in July 2014 and started providing data in September 2014, intending to estimate CO2 with high precision and resolution to characterize sources and sinks of this critical greenhouse gas [1,[22][23][24]26]. The satellite flies in a sun-synchronous orbit at an average altitude of 705 km above the earth's surface, with a descending node at around 13:30 local time, at a spatial resolution of roughly 3 km 2 and a temporal resolution of 16 days [26]. Validation of the OCO-2 data set versus measurements from the TCCON, as reported by Wunch et al. [22], shows regional biases of around 0.5 ppm and standard deviations of 1.5 ppm (similar results obtained from other studies in different regions [1,23,26]). These errors, as mentioned by Kaluwik et al. [24], are not entirely due to OCO-2, but TCCON, as well as colocation errors, are also contributing. For this study, the XCO2 data from Greenhouse gases Observing Satellite (GOSAT) with a more extended period from May 2009 was also considered (Daily Level 2 GOSAT bias-corrected products from https://disc.gsfc.nasa.gov). However, it showed very few numbers of overpasses (23 points) over Oman from 2009 to 2016, while the number of OCO-2 total measurements over Oman during the mentioned period is 473,618 points. Figure 2 shows the OCO-2 measurement points over Oman for the year 2015 as a sample of the satellite overpasses.
With a spatial resolution of 500 m, the Normalized Difference Vegetation Index (NDVI) products from MODIS (Moderate Resolution Imaging Spectroradiometer) images were used for monitoring the overall changes of vegetation over Oman during the study period. The MODIS MCD43A4 Version 6 data used here was extracted and processed over the four study regions using the JavaScript code editor in the GEE (google earth engine) platform. This MODIS dataset is produced daily using 16 days of Terra and Aqua MODIS data temporally weighted to the ninth day (https://lpdaac.usgs.gov/products/mcd43a4v006/). Generally, the NDVI ranges from −1 to +1. The negative values of vegetation index correspond to water, snow, and ice cover. In contrast, bare land Because of ground CO 2 observations absence in Oman, the column-averaged CO 2 dry air mole fraction, XCO 2 (hereafter, CO 2 concentration), from the Orbiting Carbon Observatory 2 satellite is obtained in the current study for the period from September 2014 to May 2019 (Daily bias-corrected OCO2_L2_Lite_FP.9r products were downloaded from https://disc.gsfc.nasa.gov). OCO-2 was launched in July 2014 and started providing data in September 2014, intending to estimate CO 2 with high precision and resolution to characterize sources and sinks of this critical greenhouse gas [1,[22][23][24]26]. The satellite flies in a sun-synchronous orbit at an average altitude of 705 km above the earth's surface, with a descending node at around 13:30 local time, at a spatial resolution of roughly 3 km 2 and a temporal resolution of 16 days [26]. Validation of the OCO-2 data set versus measurements from the TCCON, as reported by Wunch et al. [22], shows regional biases of around 0.5 ppm and standard deviations of 1.5 ppm (similar results obtained from other studies in different regions [1,23,26]). These errors, as mentioned by Kaluwik et al. [24], are not entirely due to OCO-2, but TCCON, as well as colocation errors, are also contributing. For this study, the XCO 2 data from Greenhouse gases Observing Satellite (GOSAT) with a more extended period from May 2009 was also considered (Daily Level 2 GOSAT bias-corrected products from https://disc.gsfc.nasa.gov). However, it showed very few numbers of overpasses (23 points) over Oman from 2009 to 2016, while the number of OCO-2 total measurements over Oman during the mentioned period is 473,618 points. Figure 2 shows the OCO-2 measurement points over Oman for the year 2015 as a sample of the satellite overpasses.
With a spatial resolution of 500 m, the Normalized Difference Vegetation Index (NDVI) products from MODIS (Moderate Resolution Imaging Spectroradiometer) images were used for monitoring the overall changes of vegetation over Oman during the study period. The MODIS MCD43A4 Version 6 data used here was extracted and processed over the four study regions using the JavaScript code editor in the GEE (google earth engine) platform. This MODIS dataset is produced daily using 16 days of Terra and Aqua MODIS data temporally weighted to the ninth day (https://lpdaac.usgs.gov/products/ mcd43a4v006/). Generally, the NDVI ranges from −1 to +1. The negative values of vegetation index correspond to water, snow, and ice cover. In contrast, bare land value is almost zero, and the typical range of 0.1 and higher is linked with the greenness of plant canopies [33].
Water 2020, 12, x FOR PEER REVIEW 5 of 15 value is almost zero, and the typical range of 0.1 and higher is linked with the greenness of plant canopies [33].

Methods
During the study period, the spatial daily and monthly means of CO2 concentration were calculated based on the satellite passes over each region. The spatial mean of NDVI for the available days was also calculated using GEE for all four study regions. The spatial daily and monthly mean of RH, as well as daily mean and monthly sum of precipitation, were also calculated based on the stations located in each region over the study period.
Time-series analysis is a suitable way to investigate the variation of parameters over time. In this study, the trend analysis of CO2 concentration as well as precipitation, RH, and NDVI was considered for the four regions over Oman. Simple linear regression (SLR) was applied to test the linear trend, with CO2 concentration as well as precipitation, RH, and NDVI considered as the dependent variables in the SLR. Also, the non-parametric Mann Kendal test (M-K test) for trend analysis was applied here to identify the significant upward or downward trend of the mentioned parameters. The M-K test, developed by Mann [34] and Kendall [35], statistically detects monotonic increasing or decreasing trends of a variable of interest in a series of data. In this test, the null (H0) and alternative hypotheses (H1) identify the absence or presence of a trend in the time series of the investigated data, respectively. The following equations give the M-K test statistic S and the standardized test statistic ZMK used for testing the hypotheses [36]:

Methods
During the study period, the spatial daily and monthly means of CO 2 concentration were calculated based on the satellite passes over each region. The spatial mean of NDVI for the available days was also calculated using GEE for all four study regions. The spatial daily and monthly mean of RH, as well as daily mean and monthly sum of precipitation, were also calculated based on the stations located in each region over the study period.
Time-series analysis is a suitable way to investigate the variation of parameters over time. In this study, the trend analysis of CO 2 concentration as well as precipitation, RH, and NDVI was considered for the four regions over Oman. Simple linear regression (SLR) was applied to test the linear trend, with CO 2 concentration as well as precipitation, RH, and NDVI considered as the dependent variables in the SLR. Also, the non-parametric Mann Kendal test (M-K test) for trend analysis was applied here to identify the significant upward or downward trend of the mentioned parameters. The M-K test, developed by Mann [34] and Kendall [35], statistically detects monotonic increasing or decreasing trends of a variable of interest in a series of data. In this test, the null (H 0 ) and alternative hypotheses (H 1 ) identify the absence or presence of a trend in the time series of the investigated data, respectively. The following equations give the M-K test statistic S and the standardized test statistic Z MK used for testing the hypotheses [36]: Water 2020, 12, 101 6 of 16 where X i and X j denote the successive data values of the time series collected over the years at times i and j, n is the length of the time series, t p is the number of observations for the pth value, and g is the number of tied values. Positive values of Z MK indicate that the observed data increase with time, while negative Z MK values indicate data decreasing in the time series. When |Z MK | > Z 1−α/2 , the null hypothesis is rejected and the alternative is accepted indicating a significant existence of a trend in the time series. The statistically significant trend was evaluated in a two-tailed M-K test at a probability level of 0.95 (α = 0.05).
Multiple correlation analysis (MCA) was applied in this study to analyze the interaction and the strength of the relationship between CO 2 concentration with precipitation and relative humidity. MCA is commonly used to explain the association between one continuous variable and two or more response variables [37] using a linear function of a set of variables which generalizes the standard coefficient of correlation. It should be mentioned, and as discussed by Huberty [37], there is a distinction between the multiple regression analysis (MRA) and MCA. Such that "in predictive research (MRA), the main emphasis is on practical applications, whereas in explanatory research (MCA), the main emphasis is on understanding phenomena [37]." However, as explained by [37], an initial choice of the response variables for an MCA study should be conducted based on some relevant substantive theory. Here, the response variables were chosen based on the initial simple correlation analysis. Then, the correlation coefficient Equation (5) and p-value of 0.95 probability level are used to test the statistical significance of MCA. However, some response variables have been deleted as a result of low/no contribution to explain the variability.
where Y i and hatY i are response variable and continuous variable values, respectively.

Time Series Analysis and Monthly CO 2 Concentration Variability along with Other Parameters
The monthly CO 2 concentration time series for different regions in Oman, as well as the fitted linear regressions, are plotted in Figure 3 for the whole study period from September 2014 to May 2019. The CO 2 concentration is shown to be increasing for all the regions under study with an estimated slope of 0.19 to 0.23 ppm/month (Figure 3). The M-K test for trend analysis of CO 2 concentration showed a significant upward trend for all regions but with a sharper increase in the Central and North Regions. It should be mentioned here that the trend analysis was performed for RH, precipitation, and NDVI over the different regions in Oman and showed a significant positive trend of RH for the Monsoon and South Regions, while no significant trend was found for precipitation and NDVI for all regions in Oman. Figure 3 shows that the Monsoon Region is marked by more intense fluctuations during the study period, with sharp decreases in CO 2 concentration levels in some months of this region, while other regions show smoother fluctuations. These sharp decreases in CO 2 concentrations in these months characterizes the lower annual growth in CO 2 concentrations not only in the Monsoon region but the South region as a whole. Thus, to explore these differences and for better comparison, the monthly CO 2 concentration time series is plotted against monthly precipitation as well as RH and NDVI to show the CO 2 fluctuation in association with other parameters (Figure 4). In general, there seems to be an association between CO 2 concentration and RH on a monthly basis, as higher peaks of CO 2 concentration seem to coincide with lower peaks of RH in all regions. The same pattern applies to CO 2 concentration and NDVI in the Monsoon and South Regions, but the influence of time lags is detected here. This association between CO 2 concentration and NDVI is not visible in other regions because of the sparse vegetation cover and the fact that a large part of these regions is mainly deserts. The interaction between monthly mean carbon dioxide concentration and monthly precipitation seems to be more complicated to be discussed based on the graphs in Figure 4. Hence, for more precise comparison, the climatology mean (CM) of monthly CO 2 concentration in association with CM of monthly RH and NDVI and particularly precipitation is investigated.
in March and then increases slightly in April and drops again in August during the monsoon period to 388 ppm. To study and clarify the monthly changes of CO2 concentration, 3Y axes CM of monthly CO2 concentration, precipitation, and RH as well as 3Y axes CM of monthly CO2 concentration, precipitation and NDVI are plotted and shown in Figure 6. Based on Figure 6, the sharp decrease in CO2 concentration in the Monsoon Region in August occurred simultaneously with the highest precipitation, RH, and NDVI during the monsoon period, though there are time lags between minimum CO2 concentration and maximum NDVI. In other words, during the monsoon season, there is an increase in relative humidity and precipitation and subsequently vegetation, which eventually causes a decrease in CO2 concentration in this region. There is a decrease in CO2 concentration in March, and although it does not seem to be linked with precipitation, RH, or NDVI within the boundary of monsoon region, it seems that there might be an impact of precipitation in the vicinity of this region on CO2 concentrations as ground stations close by in the south region have recorded especially high precipitation in the Month of March ( Figure 6). This decrease in CO2 concentration could be associated with other parameters related to regional land characteristics as well, but this needs to be further investigated. The CM for monthly CO 2 concentration during the study period for the four regions is plotted in Figure 5. The monthly variation for all regions excluding the Monsoon Region ranges from 399 to 408 ppm, while for the Monsoon Region it ranges from 388 to 407 ppm. The CO 2 concentration for all regions except the Monsoon Region is lowest in October and highest in April. In the Monsoon Region, however, the CO 2 concentration starts decreasing in January, reaching the first minimum (402 ppm) in March and then increases slightly in April and drops again in August during the monsoon period to 388 ppm. To study and clarify the monthly changes of CO 2 concentration, 3Y axes CM of monthly CO 2 concentration, precipitation, and RH as well as 3Y axes CM of monthly CO 2 concentration, precipitation and NDVI are plotted and shown in Figure 6. Based on Figure 6, the sharp decrease in CO 2 concentration in the Monsoon Region in August occurred simultaneously with the highest precipitation, RH, and NDVI during the monsoon period, though there are time lags between minimum CO 2 concentration and maximum NDVI. In other words, during the monsoon season, there is an increase in relative humidity and precipitation and subsequently vegetation, which eventually causes a decrease in CO 2 concentration in this region. There is a decrease in CO 2 concentration in March, and although it does not seem to be linked with precipitation, RH, or NDVI within the boundary of monsoon region, it seems that there might be an impact of precipitation in the vicinity of this region on CO 2 concentrations as ground stations close by in the south region have recorded especially high precipitation in the Month of March ( Figure 6). This decrease in CO 2 concentration could be associated with other parameters related to regional land characteristics as well, but this needs to be further investigated.

CO2 Concentration Relationship with Other Parameters
In this section, the correlation between CO2 concentration, relative humidity, and precipitation, using daily records on a monthly basis is investigated. Because of the limited number of NDVI data each month, this parameter is not taken into consideration here. It should also be mentioned that for

CO 2 Concentration Relationship with Other Parameters
In this section, the correlation between CO 2 concentration, relative humidity, and precipitation, using daily records on a monthly basis is investigated. Because of the limited number of NDVI data each month, this parameter is not taken into consideration here. It should also be mentioned that for precipitation data, only days with recorded rainfall are included in the correlation with coinciding days of other parameters. Because of the different characteristics of the Monsoon Region, a detailed correlation investigation was performed, and the same procedure was then applied to other regions.
Linear correlation analysis of CO 2 concentration with precipitation in the Monsoon Region showed a negative correlation coefficient for all months, which indicates atmospheric CO 2 decline because of precipitation [7]. The correlation coefficients for this region are presented in Figure 7. In this study, only the months of January to April, as well as June, have shown a non-significant correlation based on p-values of α = 0.05. Here, the lack of continuous days with significant precipitation can be the reason for the insignificant correlation.
between CO2 and another variable, for example, in August for the Monsoon Region (Figure 8), is not necessarily confirmed in a multivariate correlation (Figure 9a,b). Also, the existence of multiple significant correlations between CO2 and other variables, for example, in the Monsoon Region in October (Figure 8), does not necessarily lead to the definition of a significant MCA in the same month (Figure 9b). In other words, some response variables which had a significant impact on simple correlation were insignificant in MCA, which could be due to the interaction between the response variables. Figure 9 also shows significantly correlated variables for months with acceptable MCAs. As it is shown in Figure 9a, if only RH is considered in MCA, it will be possible to establish multivariate correlations for a longer period from May to November, which reveals more significant correlations in the monsoon period. Besides, RH with zero and one-month lags are recognized as more effective variables in local carbon dioxide fluctuation, whereas two-month and three-month lags have been eliminated in most of the months. At the same time, when both RH and precipitation are involved in MCA (Figure 9b), the period with significant MCA is limited to the months with more frequent rainy days. As a result, in MCAs including days with both RH and precipitation, RH displays less impact on carbon dioxide fluctuations compared to when it is considered individually. Furthermore, fewer correlation coefficients are obtained in months when both RH and precipitation are involved in MCA.
In the North Region, multiple correlations in CO2-RH analysis (Figure 9c) were significant in May and June for zero and one-month lags, whereas only in August, a significant MCA with RH and three-month lag precipitation is established (Figure 9d).  The daily CO 2 concentration correlation with precipitation was investigated with one, two, and three-month lags. This was done to evaluate the possible effect of monthly time lags on the consequent month's CO 2 concentration. This analysis showed a significant CO 2 -precipitation relationship with one to three-month lags for some periods in the Monsoon region. Significant correlations in different time lags based on a p-value of α = 0.05 are shown in Figure 8. These results show that an increase in precipitation in June, July, and August results in a decrease in CO 2 concentration of the following months. The same applies for September and October precipitation. A significant two-month lag correlation in CO 2 concentration was also seen with precipitation in August and September, and a significant three-month lag correlation with September rainfall.
In a research done by [7] for investigating the monthly variability in atmospheric CO 2 growth rate in Mauna Loa, air temperature and precipitation correlation with carbon dioxide concentration changes with 1 to 12-month lags were considered. As a result, significant correlations in 4-month lag in precipitation and 1-month lag for temperature with CO 2 growth rate were reported in their study. To further explain our recent findings, it can be added that along with rainy days' impact on the CO 2 concentration decrease of the same month, the effect of the previous month's rainfall on vegetation and, consequently, carbon dioxide absorption can be considered as another factor. Water 2020, 12, x FOR PEER REVIEW 13 of 15

South Region Monsoon Region
North Region Central Region

Conclusions
In this study, because of the lack of ground-based measurements, the CO2 concentration data captured by the OCO-2 satellite was used. The trend analysis of CO2 concentration during the study period shows a significant upward trend all over Oman. However, the Monsoon Region shows a lower increasing rate but more interannual variations. Along with being aware of the global upward trend and seasonal cycle of CO2, the effects of regional characteristics such as precipitation, relative humidity, and vegetation on monthly fluctuations of CO2 were considered over Oman. This research Furthermore, monthly carbon dioxide concentration correlation with relative humidity as well as one-month to three-month lags is investigated in this study. The CO 2 -RH correlation with lags is performed to test the possible impact of the preceding month's relative humidity on the carbon dioxide concentration of each month. Significant negative CO 2 -RH correlations are detected during the year except for January, February, April, and May ( Figure 8). Meanwhile, significant correlations in one-month lag are recognized throughout the year except for February, March, and October. Significant correlations for two and three-month lags are also shown in Figure 8. However, the CO 2 -RH association is not similar across the two and three-month lags varying between positive and negative correlations, where the negative correlation coefficients were more frequent. This could be related to complex atmosphere thermodynamic components of relative humidity and its interaction with precipitation [10].
As it is presented in Figure 8, the same approach was applied for the three remaining regions of Oman. More stable atmospheric conditions with rare rainy days during the year, less relative humidity changes, and less fluctuation in CO 2 concentration can be the reason for fewer months with significant correlation in South and Central regions, particularly in correlations involving precipitation. However, in the North Region, there are more months with significant correlations that reflect the impact of regional interaction of precipitation and RH on CO 2 fluctuation due to different topography features and rainfall patterns. In general, along with the global increase in carbon dioxide as well as its increasing trend over Oman, the points discussed above highlight the impact of regional atmospheric conditions on CO 2 concentration changes, particularly in the Monsoon Region.
Therefore, defining multiple correlation analysis between CO 2 and precipitation as well as RH in the months with more than one significant correlation is considered. This attempt aims to obtain a multivariate relationship to justify the interaction of different parameters on local carbon dioxide fluctuations. It should be mentioned that in MCA, the main emphasis is on understanding interactions, which explain the relationship between a single variable and a collection of response variables [37]. Predicting regional CO 2 concentration fluctuations is not the purpose of this study.
Here, the CO 2 concentration interaction with multiple variables, including precipitation and RH in different time lags for the Monsoon Region as the region with a higher number of months with significant correlations, was investigated. It is worth to mention that two sets of MCA are considered. The first is the MCA of CO 2 -RH, which is carried out for all days with data (Figure 9a), and the second is the MCA of CO 2 -RH-precipitation, where only days with precipitation are considered (Figure 9b).

Conclusions
In this study, because of the lack of ground-based measurements, the CO2 concentration data captured by the OCO-2 satellite was used. The trend analysis of CO2 concentration during the study period shows a significant upward trend all over Oman. However, the Monsoon Region shows a lower increasing rate but more interannual variations. Along with being aware of the global upward trend and seasonal cycle of CO2, the effects of regional characteristics such as precipitation, relative humidity, and vegetation on monthly fluctuations of CO2 were considered over Oman. This research The multiple correlation analysis shows that the existence of a significant simple correlation between CO 2 and another variable, for example, in August for the Monsoon Region (Figure 8), is not necessarily confirmed in a multivariate correlation (Figure 9a,b). Also, the existence of multiple significant correlations between CO 2 and other variables, for example, in the Monsoon Region in October (Figure 8), does not necessarily lead to the definition of a significant MCA in the same month ( Figure 9b). In other words, some response variables which had a significant impact on simple correlation were insignificant in MCA, which could be due to the interaction between the response variables. Figure 9 also shows significantly correlated variables for months with acceptable MCAs. As it is shown in Figure 9a, if only RH is considered in MCA, it will be possible to establish multivariate correlations for a longer period from May to November, which reveals more significant correlations in the monsoon period. Besides, RH with zero and one-month lags are recognized as more effective variables in local carbon dioxide fluctuation, whereas two-month and three-month lags have been eliminated in most of the months. At the same time, when both RH and precipitation are involved in MCA (Figure 9b), the period with significant MCA is limited to the months with more frequent rainy days. As a result, in MCAs including days with both RH and precipitation, RH displays less impact on carbon dioxide fluctuations compared to when it is considered individually. Furthermore, fewer correlation coefficients are obtained in months when both RH and precipitation are involved in MCA.
In the North Region, multiple correlations in CO 2 -RH analysis (Figure 9c) were significant in May and June for zero and one-month lags, whereas only in August, a significant MCA with RH and three-month lag precipitation is established (Figure 9d).

Conclusions
In this study, because of the lack of ground-based measurements, the CO 2 concentration data captured by the OCO-2 satellite was used. The trend analysis of CO 2 concentration during the study period shows a significant upward trend all over Oman. However, the Monsoon Region shows a lower increasing rate but more interannual variations. Along with being aware of the global upward trend and seasonal cycle of CO 2 , the effects of regional characteristics such as precipitation, relative humidity, and vegetation on monthly fluctuations of CO 2 were considered over Oman. This research showed that in the Monsoon Region with higher precipitation as well as higher relative humidity and vegetation cover, the monthly atmospheric CO 2 concentrations were significantly affected by the regional parameters compared to the rest of Oman. Precipitation was shown to have a higher impact in defining the changes in CO 2 concentrations. Ultimately, despite the short study period, the findings of current research are sound and reliable. However, other parameters such as soil temperature and soil moisture relationship with atmospheric CO 2 concentration variations are suggested in further studies.