Drought Analysis for the Seyhan Basin with NDVI and VCI Vegetation Indices

Various drought indices have been developed to monitor the drought, which is one of the results of climate change and mitigates its adverse effects on water resources, especially agriculture. Vegetation indices determined by remote sensing have been the subject of many studies in recent years and shed light on drought risk management. This study is examined in the Seyhan River Basin, a basin with Turkey’s considerable population density counts and is situated south of the country. Normalized Difference Vegetation Index (NDVI) and Vegetation Condition Index (VCI) are the most widely used vegetation indices and are very useful because they give results only based on satellite images. This study examined the Seyhan Basin by using satellite data in which the vegetation transformation occurring due to the decline of agricultural and forest areas was also seen. An increase in drought frequency was detected in the Seyhan Basin using NDVI and VCI indices. It was determined that climate change and drought increased with a linear uptrend. It is recommended that decision-makers should take the necessary measures by considering the drought risk maps and that long-term drought management plans should be made and implemented.


Introduction
The status of water resources adversely affected by climate change, hence drought, is a major concern in agriculture, water resources management, and human use. Drought is de ned as a natural event that negatively affects land and water resources, and hydrological equilibrium is disrupted due to precipitations falling below normal levels (Benson et al., 1997). It is possible to classify drought as meteorological, agricultural, and hydrological drought. Meteorological drought is the decrease, which occurs according to the long-year averages, in precipitation. On the other hand, agricultural drought is based on the amount of water available in the root zone of the plant. In terms of agriculture, the periods in which the amount of water to meet the water needs of plants in the soil is not present are de ned as arid. Precipitation, plant water consumption, and soil characteristics can be shown as the main factors for agricultural drought. Hydrological drought refers to the decrease in surface and groundwater resources due to the lack of long-term precipitation. Even long after the end of the meteorological drought, hydrological drought may exist (Diṅç et al., 2016). Drought risk management constitutes a vital part of water-resource management policies and strategies. National drought policies have an essential role in managing drought risk (Meza et al., 2020, Vogt et al., 2018, Wilhite et al., 2014. It is necessary to prepare drought management plans to mitigate the effects of drought depending on the legislation of the country and by considering the speci c drought characteristics and effects of the basin (EuropeanCommission, 2007). Elements of the drought management plan include knowing the characteristics of the river basin, investigation of historical drought events in the basin, evaluation of the possible risk(s), determination of indicators and threshold values for drought analysis, creation of a program of measures to reduce the effects of drought, and establishment of the early warning system and organizational structure (GWP, 2015, Pérez-Blanco & Gómez, 2014). Drought risk management includes the following stages: hazard, impact assessment, and affectability, early warning system, including drought monitoring and forecasting, and preparedness and harm reduction (Wilhite, 2000).
Drought early warning systems typically aim to monitor, evaluate, and present information on climate, hydrological features, water supply conditions, and trends (Funk & Shukla, 2020). The goal here is to provide early information before or during the onset of drought within a drought risk management plan to mitigate the potential impacts. Since drought is a slow-starting and progressive hydrological event, monitoring and analyzing drought is of great importance. Monitoring and analysis of drought are done employing various indicators and indices. These indicators and indices help characterize drought by providing information on the severity, location, duration, and timing of drought to determine, classify, and monitor drought conditions. Some indicators and indices can also be used to validate indicators of drought data that have been modelled, remotely detected, or assimilated into the model. The geographic information systems have made it possible to overlap, map, and compare different indicators and indices thanks to the power of evolving computational and imaging techniques (Svoboda & Fuchs, 2017).
Indicators are variables or parameters used to describe drought conditions. In general, drought indicators include the variables summarized in Table 1.
The index values initiate or terminate the implementation stages of a drought management plan.
Therefore, drought management plans should be formed based on index values (Svoboda & Fuchs, 2017  Several studies mentioned that NDVI and VCI indices are useful methods to detect drought and make a prediction. The main aim of the study was to answer the following questions based on remote sensing technologies. Advances in space technologies and computer systems have brought along a broader and more e cient use of remote sensing technologies and geographic information systems (GIS). The ability to easily transfer various geospatial data to the GIS environment with images taken from space via satellites has increased the possibilities for analysis of issues such as natural resource management, land use and land cover, environmental and ecological analysis, disaster risk assessment, and meteorological, hydrological and agricultural applications. Remote sensing technologies, especially satellite products, are used effectively and extensively in various hydrological applications for various High-resolution vegetation change information provided both temporal and spatial by vegetation indices (e.g., NDVI), can contribute to drought-related research without requiring additional information on drought. Vegetation indices are preferred because they are easy to use, and they do not require any assumptions and/or additional information other than themselves (Bulut & Yilmaz, 2016). Vegetation indices can be determined by remote sensing methods and have a wide range of applications because the green vegetation gives high re ectivity values in the near-infrared region of the electromagnetic spectrum (Gökdemir, 2002). Most of the satellite sensors measure red and near-infrared light waves re ected from the land surface. Using mathematical formulas, raw satellite data related to these light waves are converted into vegetation indices. Vegetation indices describe the greenness (relative density The authors were motivated to get answers to the following questions during the work.
-Is it possible to apply NDVI and VCI indices to the Seyhan Basin?
-Are the results that NDVI and VCI indices will produce in the Seyhan Basin reasonable?
-Is there a drought problem in the Seyhan Basin, and how is it progressing?
-What is the frequency of the drought in the basin?
In this study, the remote sensing method was used, and drought analysis was performed for the Seyhan Basin with NDVI and VCI vegetation indices.

Remote sensing and data sources
Within the scope of the drought analysis studies of the Seyhan Basin, vegetation indices, which provided information about the change in plant greenness, were preferred. Accordingly, the "Normalized Difference Vegetation Index (NDVI)" and the "Vegetable Condition Index (VCI)" were analyzed both temporal and spatial within the boundaries of the Seyhan Basin.
The study in which the vegetation index was associated with precipitation (Şahin et al., 2009) showed a correlation between the data obtained from the precipitation stations in different regions of Turkey and the NDVI data. Similar studies performed with data for rainfall monitoring stations in Turkey and compliance with the drought indices data were discussed (Dikici, 2013, Dikici & Aksel, 2021. NDVI is one of the vegetation indices that is quite widely used in forest classi cation and agricultural studies as well as in the detection of the change in land cover. On the other hand, high NDVI values indicate areas in which there is healthy plant development (Yıldız et al., 2012).
NDVI data obtained from National Oceanic and Atmospheric Administration (NOAA), Advanced Very High-Resolution Radiometer (AVHRR), and Moderate Resolution Imaging Spectroradiometer (MODIS) satellites are satellite images commonly used to monitor vegetation changes in large sites. AVHRR and MODIS satellites provide NDVI data as ready to use. Therefore, the atmospheric correction is not needed in these satellite data, thus, no additional data are required for the atmospheric correction. Data at the NIR and RED wavelengths obtained from the LANDSAT satellite need atmospheric correction before NDVI is calculated. Although the normalizing phase reduces these atmospheric components' effect on NDVI, NDVI data obtained from AVHRR and MODIS satellites that do not require atmospheric correction were used in this study. The time interval, resolution, and recurrence time of the NDVI values obtained from these two satellites are given in Table 2.
Several studies in the literature show that NDVI values calculated using AVHRR satellite data (Cracknel, 1997)  . As a result of the study conducted to partial correction of the AVHRR NDVI time series, the AVHRR NDVI3g product has been obtained. In the context of this conducted study and the AVHRR NDVI product, the time series of the AVHRR NDVI3g products was also used.
Modis's difference from the other sensors is that it has a high temporal and positional resolution and can collect data from 0.4 µm to 14 µm in 36 separate spectral bands (Hall & Riggs, 2007). MODIS sensor has 250 m spatial resolution between bands 1-2, 500 m spatial resolution between bands 3-7, and 1 km spatial resolution between bands 8-36 (Lillesand et al., 2015). Although MODIS images are shot twice a day, NDVI products are broadcasted as 8-day composites. MODIS NDVI images, consisting of 4800 rows and 4800 columns, provide the opportunity to analyze the change in vegetation activity in an extensive area (Çelik & Karabulut, 2013 (Figure 2). When calculating the data series, the period was selected as 2001-2015.
For this study, the VCI was calculated using the NDVI values obtained from MODIS and AVHRR satellite data to compare the drought determined by drought analyses conducted for the Seyhan basin. Accordingly, the VCI time series obtained using satellite data covering the years 2001-2016 for each 250 m satellite cell within the basin area boundaries are presented in Figure 3.
In the time series, the VCI values shown by red indicate drought in the plant state, while blue values can be interpreted so that the plant state is at the seasonal and climatic normal conditions. VCI can provide information about the onset, duration, and severity of drought by considering the impact of drought on vegetation. VCI compares the NDVI data of a given period with the highest (max) and lowest (min) data values of the NDVI values belonging to the analyzed entire period (Quiring & Ganesh, 2010). VCI is expressed as a percentage (%) and provides information on when the observed value's highest and lowest values occurred in past years. Whereas low VCI values indicate poor vegetation status, high VCI values indicate that vegetation is good [59].
VCI can be considered as a normalized version of the NDVI. In addition to NDVI, VCI was also evaluated in this study since VCI is a more appropriate index in assessing the deviation of vegetation from the normal state. Therefore, VCI allows the comparison of simultaneously measured NDVI values for different ecosystems, i.e., for different vegetation in different geographies. VCI is a better indicator of soil moisture vulnerability than NDVI because it can distinguish short-term climate signals from long-term ecological signals. The importance of VCI relates to the vegetation index's viability studied by the vegetation index (Jain et al., 2010). VCI data, like NDVI, have high resolution and reasonable areal extent. In the literature, several studies related to the use of VCI for drought analysis purposes , Quiring & Ganesh, 2010).

Results
The scope of drought index studies aimed to analyze climatic change responses of irregularly irrigated or non-irrigated agricultural areas and forest-vegetation areas within the Seyhan Basin. Coordination of Information on the Environment (CORINE) layers used in these analyzes was selected, and temporal NDVI changes in these layers were calculated. The distribution of these CORINE layers over the basin is given in Figure 4. In contrast, the layer list is given in Table 3 (In the CORINE classi cation, while the layers beginning with the number two represent agricultural areas, the layers beginning with the number 3 represent the classes of forest and semi-natural areas).
The fact that the AVHRR-3G and MODIS NDVI data used in the analysis were at different resolutions led to the differentiation of the classi ed areas. The eight km-resolution pixels in AVHRR-3G data corresponds to 1024 pixels at 250 m resolution in MODIS data. Therefore, some pixels sometimes consist of a mosaic of classes with very different properties alongside the class designated as the dominant class. Another issue to be considered when making the assessment is that land usage may change over time.
The NDVI temporal change series for agricultural areas starting with Code 2 in the CORINE 2012 land use data is given annually in Figure 5. While the dashed lines represent the AVHRR-3G data, the continuous lines represent the MODIS data. December-January-February (DJF), March-April-May (MAM), June-July-August (JJA), and September-October-November (SON) refer to the winter, spring, summer, and autumn months, respectively.
When the time series of the examined layers were evaluated in the general framework, it was observed that NDVI values decreased between 2002 and 2004 and rose signi cantly between 2007 and 2013. Except for short-term uctuations, it has been calculated that NDVI values are generally higher in spring and decrease in autumn. Since olive grove, pasture, and non-irrigated mixed agriculture classes covered areas too small to be represented in AVHRR-3G resolution, they were not included in the related charts. In olive groves, which are resistant to cold and known as the evergreen undead tree, values above the average have been observed in winter, unlike other classes. The natural vegetation class is a vital suppressor in the results as it covers large areas in the Seyhan Basin. For this reason, the value calculated from different satellite data has always been close. This layer in which the human in uence is limited is one of the classes where the effects of drought on plants can be better observed, and it has shown signi cant declines in 1982, 1989, 2004, 2007, 2012, 2014. On the other hand, 2011 and 2015 are the years in which the highest values were observed. The NDVI annual temporal change series for forest and semi-natural areas starting with the Code 3 of CORINE classi cation is given in Figure 6. In the studied forest layers, annual changes seen in agricultural areas were observed similarly. In their NDVI calculations, the lowest values were determined during winter, and the highest values were determined in summer and spring.
Coniferous trees, which are cold-resistant and evergreen species, received the highest MODIS data values during the winter months. They received the lowest values in the AVHRR-3G data. In this case, it was evaluated that the coniferous forests class was confused with other classes as a result of the failure of a good class differentiation at 8 km resolution and that because deciduous trees are located within the same pixel as conifers, they received much lower values in winter than expected.

Conclusions
The years in which two indices jointly indicated drought for the Seyhan Basin were determined as 1973-1974, 1989, 2001, 2007-2008, 2014, and 2016. The drought return period for the Seyhan Basin is decreasing over the years. On the other hand, overall NDVI mean values have been increasing since the 2000s for all seasons. This increase may result from the fact that the snow cover, which decreases the NDVI values due to climate change, has reduced in terms of process and area. It is thought that contribution of the increasing temperature values positive effects on the photosynthesis process.
Changes in land use may also cause differentiation in NDVI values. Especially considering the longperiod (e.g., 1982 -2016), this change is inevitable. It should be taken into account that with the population growth, forestlands can transform into agricultural areas and agricultural areas can transform into arti cial pavement areas, or sometimes the opposite situations can occur due to reasons such as migration from village to city, and this can cause arti cial tendencies in the results. Based on the plant indices, it is understood that there is a drought trend in the press. It is clear that with the decline of forestland and agricultural areas, the drought seen in the past years will increase with each passing time for the Seyhan Basin. In the case of drought estimation at intervals covering long periods, the changes in the land-use patterns and demography of the region should also be considered.
It is possible to make plans covering different purposes with drought indices, which have a wide application area, include practical application methodology and can make a higher resolution and precise solutions thanks to remote sensing technologies. However, these indices should be used to associate eld data and other GIS layers such as land-use, population growth, etc.   Temporal change of calculated NDVI value for plant change areas coded CORINE 324 in the Seyhan Basin