Estimation of Vegetation Changes Concerning Annual Rainfall and Temperature in Semnan Province

Vegetation is one of the most important components of terrestrial ecosystems, which plays a vital role in carbon regulation and balancing, energy exchange and climate stability. Also, temperature and rainfall are the most important influential factors in changing the vegetation index, NDVI. Understanding the relationship between rainfall, NDVI and temperature is essential in forestry planning over each region. Accordingly, the main objectives of this research are 1) to monitor the annual changes of NDVI from 2001 to 2020 using linear regression’s slope and Sen’s slope estimator methods, and 2) to investigate and determine the relationship between NDVI and the climatic components, specifically rainfall and temperature, in Semnan province.

Semnan province was selected as the study area to evaluate the relationship between climatic components and vegetation cover by using remote sensing data. The NDVI data was extracted from the Terra MODIS product in a spatial resolution of 500 meters and was processed to evaluate vegetation changes in Semnan province during the years between 2001 and 2020 on monthly scale. After that, the monthly rainfall and temperature data were obtained from both synoptic and climatology stations; then they were converted into annual scale. Furthermore, the rainfall and temperature reanalysis grid-base data, ERA5-Land, was downloaded in about 9 km spatial resolution from 2001 to 2020. Reanalysis data usually contains systematic error compared to observational data, which can affect the output results and requires to be corrected. Consequently, we utilized one of the recent bias-correction approaches, the Quantile Mapping (QM) bias-correction method, to correct biases over the entire distribution of the rainfall and temperature reanalysis data. At this point, each set of the rainfall and temperature grid-based data was resampled to 500 meters based on the spatial resolution of NDVI pixels. Next, each series of rainfall and temperature data were corrected based on QM method from the years 2001 to 2020 according to the availability of the NDVI time series data. The relationship between annual rainfall and temperature with the NDVI was calculated in each month of the year (2001-2020). In this study, linear regression and the non-parametric method of the Sen’s slope estimator were used to investigate the changes in NDVI trend for each pixel from 2001 to 2020. Finally, to check the accuracy of the relationship between vegetation, temperature and rainfall, the coefficient of determination was used.

The linear regression’s slope indicated that 25% of ​​Semnan’s area had vegetation variations close to zero in each month during the years 2001 to 2020. It means that NDVI values did not change significantly and it was almost unchanged. Moreover, based on the Sen's slope estimator, the results showed that there was no noticeable change in decreasing or increasing the amount of vegetation in about 75% of ​​Semnan’s area. The analyses also showed that the coefficient of determination between NDVI and rainfall varied from 18% to 44% in different months, and the highest relationship values were observed in September and December. Moreover, in more than 50% of the Semnan’s area, the relationship between NDVI and rainfall has varied from zero to more than 42%. The results indicated that vegetation cover has no significant relationship with annual rainfall in both winter and spring. Furthermore, in 50% of the study area, the estimated NDVI variation indicated zero or negative values in each month, which confirms that the vegetation cover has not changed significantly or it has decreased slightly in response to the temperature.

In recent years, the evaluation of changes in NDVI time series has been developed by using satellite images and remote sensing techniques. In addition, the climate components like rainfall and temperature are among the factors affecting the growth of vegetation cover, which has attracted the attention of many researchers. By considering the linear regression’s slope and Sen’s slope estimator, the results showed that NDVI did not change significantly during the years 2001 to 2020 in Semnan province, but it decreased in some areas. Moreover, the effect of rainfall and temperature demonstrated that vegetation has either direct or indirect relationships with rainfall or temperature in some months (positive or negative values, respectively). The linear regression’ slope between annual temperature and NDVI showed that in 50% of ​​Semnan’s area, NDVI variation was estimated to be nearly zero or negative in each month. Generally, in arid regions like Semnan province, the growth of plants is controlled by two climatic factors, rainfall and temperature. In arid and semi-arid regions (where the amount of rainfall is low, such as Semnan province), or in regions where the percentage of humidity is high, the maximum relationship between NDVI and the rainfall was not observed.

In arid and semi-arid regions, due to the fragility of the ecosystem, the reduction of vegetation cover can have irreparable consequences such as increasing the movement of soil particles. This action will eventually lead to wind erosion and will increase dust in the region. Therefore, the present study was conducted to evaluate the monthly changes of NDVI from 2001 to 2020 by the linear regression’s slope and Sen’s slope methods. In addition, the relationship between NDVI and climatic components of rainfall and temperature were discussed. The findings showed that results of the linear regression’s slope and Sen's slope estimator are both almost the same. The relationship between rainfall and NDVI indicated that in 75% of Semnan’s area, the estimated value of coefficient of determination had the highest value in September and December. It means that with the increase of rainfall, vegetation cover also increases. Based on the linear regression’s slope between the annual temperature and NDVI, it was observed that in almost 50% of the study area, NDVI variations were approximately estimated to be zero or negative in each month, which means the vegetation cover has not significantly changed or even decreased by changing temperature. In future studies, it is suggested to use other remote sensing NDVI products such as Landsat satellite and even other climate reanalysis data to have a more accurate view of vegetation cover changes in the study area.