A novel integrated approach for monitoring drought stress in an aeolian desertification area using Vegetation Drought Status Index

Drought is a universal natural phenomenon in the world, which has a wide range and a long duration. It is one of the most serious natural disasters affecting agricultural production and human life. The traditional drought monitoring method is based on limited data from meteorological observation stations. Although spatial interpolation can be carried out for the data, there are many uncertainties in drought monitoring due to the influence of many factors, such as sample size, spatial distribution characteristics and so on.

Remote sensing (RS) can quickly and accurately collect the related information of the surface features and has the characteristics of high time efficiency and wide monitoring range (Bowers & Hanks 1965; Tian et al. 2019; Huo et al. 2022a), which makes it become the research hotspot and frontier in the field of drought monitoring. In order to monitor drought quickly and widely, it is necessary to use satellite RS image data to monitor the surface of the observation area. The thermal inertia model was established using the diurnal range of surface temperature (Watson 1952). Rosema et al. (1992) improved upon this model for daily evapotranspiration and thermal inertia. Qi et al. (2003) combined plant cover information and land surface temperature to research Chinese drought using the Temperature Vegetable Drought Index.

In any drought monitoring technique, soil moisture is the key factor in agriculture domains, ecology, hydrology, and climate. Soil moisture is important in energy exchange and matter between the atmosphere and land surface. Because of the sparsity of sample points, the limited representative range, and the poor timeliness of data collection, especially in the macro and large-scale monitoring of soil moisture content, most of the conventional soil moisture measurement methods have the problem of replacing areas with points, and it is difficult to give the boundaries between different soil moisture content areas (Rhee et al. 2010; Luo et al. 2018a; Gorlé et al. 2019). In the RS monitoring of drought, because soil moisture is affected by many factors, such as temperature, rainfall, soil texture, and the interference of vegetation coverage cannot be ignored, the RS inversion of soil moisture has always been difficult.

Soil moisture can be rapidly depleted when precipitation is insufficient, so soil moisture can be used as the evaluation basis for areas suffering rain shortages. A variety of feasible and reliable methods to measure soil moisture currently exist from the point scale to the area scale (Huo et al. 2016a, 2020a; Luo et al. 2018b). Soil moisture is an important indicator of agricultural drought monitoring. Vegetation status and surface temperature can reflect water stress status. In the visible and near-infrared bands, the focus of RS drought monitoring is the change of vegetation index. The theoretical basis is the strong absorption and strong reflection characteristics of vegetation (crops, etc.) in these two bands (Huo et al. 2011a, 2020b).

There are many methods to monitor drought using RS technology. At present, thermal inertia model (Huo et al. 2016a), vegetation water supply index (Huo et al. 2011a), vegetation water shortage index (Huo et al. 2009), vegetation state index (Huo et al. 2010), vegetation index (Luo et al. 2019), and surface temperature are mainly used (Bayissa et al. 2019; Nguyen et al. 2019). Although RS has been applied to regional drought monitoring for a long time, the research of drought RS monitoring is mostly applied to the areas with single vegetation or flat terrain in the North of China, but less for the Northwest arid areas with diverse landforms, sparse vegetation, and less rainfall.

The primary focus of this research is a new drought monitoring methodology called the Vegetation Drought Status Index (VDSI), which satellite-based earth observations and integrates historical measurement data with soil, land use, and land cover. In order to monitor drought quickly and widely, it is necessary to monitor the surface of the observation area. Based on the spectral reflection characteristics of water, this study analyzes the common characteristics of water-bearing soil and vegetation and proposes the new VDSI suitable for the mixed region of soil and vegetation. It combines historical survey data with satellite-based earth observation data for drought monitoring in mixed areas, such as bare land and plants. This method combined with the new data mining analysis technology to identify the historical climate vegetation relationship related to the drought phenomenon.

MODIS data have a wide range of bands and high time resolution, which has a great application value for real-time earth observation (Huo et al. 2011b, 2019). Therefore, MODIS data were selected as the main RS data source in this paper. MODIS data come from Warehouse Inventory Search Tool (WIST; http://Wist.echo.nasa.gov/api/). MOD11A2 is an 8 d synthesis data with a spatial resolution of 1 km × 1 km. MOD13A3 is a monthly vegetation index product including NDVI (Normalized Difference Vegetation Index), EVI (Enhanced Vegetation Index), blue light and other visible light, near-infrared, mid-infrared, thermal red light band reflectivity, and other auxiliary information.

Satellite RS images are reflections of surface features characterized by their thermal radiation and electromagnetic waves. Their electroreflectance and thermal radiation are different because of different structures and physicochemical properties of surface features. Almost all incident energy, ranging across mid-infrared and near-infrared bands, is absorbed by water. Contrary, soil and plants absorb very little energy but reflect most of the energy in these ranges. Because of these different characteristics, the relationship between spectral reflectance and soil drought is clearly to be distinguished.

Soil is a complex containing many components. The soil spectrum is influenced by water, soil parent material, organic matter, and other elements; the effect of soil moisture is specifically important. In a word, reflectivity decreases with the increase of soil moisture.

The drought course is relevant to the water cycle as a whole including storage in rivers and lakes, vegetation water consumption, soil moisture, and precipitation. However, soil moisture is usually utilized to represent the status of a drought. It is the key factor among them.

Based on the MODIS 1B data set, the relationship between the retrieval values from the RS image data sets and the fieldwork measure values of soil moisture were analyzed at six spatial scales from 1 km × 1 km to 32 km × 32 km, respectively. Huo et al. (2016b) indicate that 2 km scale is the suitable scale for the retrieval of soil moisture in the study area; 1 and 4 km are the second best scale. Compared with references (Huo et al. 2018), this paper used the newly resampled 2 km spatial resolution MODIS band B2 and band B1 to calculate the vegetation index and which can obtain more accurate calculation results. Therefore, it can be said that this paper introduces the revised vegetation index, develops the calculation formula in the reference (Huo et al. 2018), and makes the calculation results more accurate.

The reason for Xinjiang's arid environment is that it is far away from the sea and surrounded by mountains. The water from the ocean gradually decreases in the long-distance transportation process and is blocked by mountains when it arrives at Xinjiang, which not only reduces the water content but also forms the regional difference in precipitation distribution.

There are two kinds of drought in Xinjiang caused by natural factors and human factors. Natural factors refer to the fact that Xinjiang is located in the Northwest inland of China, far away from the ocean and hard to reach by water vapor. In the North, West and South of Xinjiang, there are tall mountains to block the water from outside. In addition, the climate of Xinjiang belongs to the temperate continental climate, which is characterized by little precipitation and high temperature in summer. Human factors mainly refer to the destruction of forest vegetation here. Since the western development of Xinjiang, the population has increased rapidly, and the expansion of the city has destroyed a large number of vegetation. In addition, due to the increase of population and the increase of domestic and industrial water, the surface water and groundwater are overdrawn, which directly leads to the drought in Xinjiang.

The data source selected one MODIS image data from 2007. Spatial resolution is 1 km. The distribution of drought was analyzed by using the homochronous soil moisture data collected in the field in the study area; fieldwork is a significant part of model verification.

MODIS is the vital spectroradiometer on Terra. Its data are shared directly to the world. MODIS has three scales of resolution: 1,000, 500, and 250 m. The scanning width is 2,330 km. The satellite's transit times are 00:30 and 12:30 (Beijing time) in the study area. MODIS images at 12:30 were used to minimize the influence of cloud cover. The MODIS 1B data set was downloaded from WIST with 1 km resolution as a 16-bit attribute dataset (Luo et al. 2018a, 2018b; Sundell et al. 2019).

The surface observation time corresponds to the time when MODIS data were taken over. The observation site is in the area with desertified land distributed on both sides of Tarim River and the edge of Gurbantonggut desert in Xinjiang. The location of 81 measured data detected in July 2007 is shown in Figure 1. The data collection method was as follows: Firstly, we selected the area with a uniform surface cover distribution, with an area of about 1 km. Then we measured the longitude and latitude coordinates of the center points of each pixel near the area from the MODIS image and used GPS navigation to find the center point nearest to the center point of the actual measurement area, so that the measured area was completely arranged in a single pixel as far as possible. Lastly, from the middle point of the survey area, 100 m sample lines were arranged in four directions (South, East, North, and West), respectively, and the soil moisture content was measured at an interval of 20 m in each direction. The mean value of 21 sample points (including the center point, corresponding to the pixel location) was used to represent the actual soil moisture content of the surface. The soil moisture was calculated by the method of weighing and drying, that is, under the constant temperature of 100–105°C, bake to the dry state, and calculate the moisture content of the excavated sample by the weight before and after drying. Because the surface humidity retrieved by satellite is the comprehensive humidity of various surface features, the average humidity was used to represent the actual humidity within the observation range. The method and position of the observatory we used are similar to those used by Huo et al. (2009).

A visual examination of Figure 3 shows that the intensity of drought depicted and general spatial patterns similar. In Xinjiang Uygur Autonomous Region of China, the VDSI-based drought status is in accordance with actual measurements, namely, the most severe drought area was Gurbantunggut Desert areas and the central portion of the Taklimakan. Especially in the most arid areas of Xinjiang, the improved 2 km spatial resolution image of VDSI is readily apparent, showing considerably higher detail at a local scale area. In a word, drought status can be divided into two parts: the drought in the North is mainly concentrated in the Gurbantunggut Desert, and the drought in the South is mainly concentrated in the Tarim River Basin. In the Southern part, the drought is serious in the center and light in the surrounding area, and many villages and towns are distributed in the light area. The main reason is strong sunshine and less precipitation. The Tarim River Basin is a severely arid region, although it is fed by inland rivers. This is mainly because water supply cannot fully meet ecological water demand. Most of the remaining areas are moderate, and the area of the non-drought region is very small.

The Takilimakan Desert is approximately 400 km in width in the South–Northwest direction and approximately 1,000 km in length in the East–West direction; it is located at the center of the Tarim Basin and it is the tenth largest desert in the world. Its total area is 330,000 km². The Gurbantunggut Desert is the second biggest desert in China. It is located at the center of the Junggar Basin with an area of about 48,800 km². The lowest soil moisture areas, shaded areas in Figure 3, are these two deserts across the study area. There are several cities, including Karamay, Shihezi, and Aksu, scattered throughout these areas which have higher soil moisture (light areas in Figure 3).

Table 1 shows the correlation between the soil moisture data from measurements and the VDSI by means of the Pearson correlation coefficient.

Figure 4 shows the scattered diagram between measured soil moisture and VDSI. According to Figure 4, the data points can be divided into three areas: the first area is between 0 and 10%, and the distribution area is mainly desert; the second area is between 10 and 15%, and the distribution area is mainly on both sides of the Tarim River mainstream; the third area is between 15 and 30%, and the distribution area is mainly residential or water areas. Figure 4 shows that the VDSI value is slightly higher than the measured value, probably because the RS monitoring depth is relatively shallow (about 10 cm below the ground), while the actual measurement is generally about 20 cm below the ground surface, so the VDSI value is slightly higher than the measured value.

The grade standards of VDSI were established based on Xinjiang's Statistical yearbooks, agricultural statistical yearbooks, drought statistical data, fieldwork data, and RS image data, as is shown in Table 2.

Figure 5 depicts the spatial distribution of drought in Xinjiang. Figure 5 map has classified particularly severe, severe, moderate, mild, and normal drought conditions for the same location depicted in Figure 3. As discussed earlier, Figure 3 is generated by the VDSI indicator to determine the overall level of drought status for a value. On the contrary, Figure 5 represents a single aspect of drought stress on five grade standards. Figure 5 classifies North-central Xinjiang as being in severe drought grades (indicated by the red color label in Figure 5) and classifies South-central Xinjiang as being in particularly severe drought (indicated by the white color label). This intuitively reflects that the drought level of the Taklimakan Desert is one level higher than that of Gurbantunggut Desert. Particularly, severe drought occurred on the South side of the Tarim River where the Takilimakan Desert hinterland is located; this area is approximately 19.6 km². The Eastern part of Xinjiang also has some particularly severe drought areas. The severe drought-grade region was located at the North side of the Gurbantunggut Desert and the Tarim River. This area was around 47 km²; the moderate drought-grade region was distributed at the Southern, central, and Eastern edges of Xinjiang. It was about 48.4 km²; the normal and mild drought regions were mainly located around larger cities with areas of 21.2 and 29.7 km², respectively.

The drought index is used to monitor, identify, and classify drought conditions. They allow for quantitative evaluation of the spatial extent, duration, and intensity of anomalous climatic conditions. Further analysis of satellite RS images in different months of the year shows that among the three different types of sandy land (fixed sandy, semi-fixed, and mobile land), the fixed sand land has the largest change in soil moisture content, and the mobile sand land has the smallest change in soil moisture. The mobile sandy land has the largest soil moisture, followed by the semi-fixed sandy land and the fixed sandy land. The climatic conditions change periodically with seasons, and the water content of sandy land also changes with time (Feng et al. 2001; Du et al. 2007; Tracy & Corcoran 2014; Sposito et al. 2019). In May, September, and October, the soil moisture of the sandy land is relatively high, which is related to the fact that the rainfall in this area is mainly concentrated over spring and autumn, which further indicates that rainfall is the main supply mode of soil moisture in sandy land. In July and August, the water content of sandy land is relatively low, which is due to the small rainfall, high temperature, and large evaporation, resulting in a large loss of soil water, but not timely and effective recharge. Therefore, when afforestation is carried out in the arid and desertification areas, the spring and autumn periods with high soil moisture content should be selected.

Due to the high temporal resolution of MODIS image data, but the low spatial resolution, the following aspects should be considered to further improve the accuracy and quality in RS drought monitoring: the fusion technology of multi-source RS data should be considered in the follow-up research work; the ground survey is the measured soil moisture within 1–20 cm of the ground, and the RS inversion depth is the moisture content within 10 cm under the ground surface. The relationship between the two different depths needs further accurate calculation; as the transit time of the satellite in the sky above the study area is only a few minutes, and the actual measurement time on the ground is generally between 9:00 and 16:00, the time out of sync will also cause errors, so the later field investigation should try to keep the same time with the transit time of the satellite in the sky. Further studies are necessary to develop the VDSI to enhance its applicability to different seasons, scales, and regions.

Differences among statistical methods and cloud cover in various agricultural departments caused some errors between statistical data and VDSI results in regard to some specific drought areas. Overall, VDSI still presented reliable reference data for the droughts status across Xinjiang. In future research, the change in surface temperature and the influence of different resolution scales on the inversion results will be studied with regard to the development of the model.

The outcomes of this study suggest, in line with the most recent applications of remote soil moisture, that VDSI has good potential in monitoring drought, especially in the arid and desertification areas. With the forthcoming availability of microwave, optical, and thermal data streams, this study supports the development of innovative trails for enhanced drought monitoring.

A new drought indicator was presented based on the inverse of the MODIS image data, which in turn is related to the soil moisture content. The Xinjiang case study results indicated that VDSI actually reflects the drought status. Our primary conclusions can be summarized as follows.

• (1)

  A new RS monitoring model of VDSI which is appropriate for Xinjiang Uygur Autonomous Region of China was proposed by using the composite information of multiple bands of MODIS satellite image data and considering the spectral reflection characteristics of water. It is of great importance to the operational process of drought monitoring.

• (2)

  The VDSI RS monitoring model can precisely reflect the spatial change characteristics of drought in the study area. The VDSI model was utilized to monitor the drought state on both sides of Tarim River and the edge of Gurbantonggut desert. Through the correlation analysis and verification with the VDSI value and in situ soil moisture obtained at the same time, Confidence Coefficient was higher than 0.79. The linear correlation was significant.

• (3)

  The drought area of each drought class is mainly consistent with the fluctuation trend of statistical data. The comparison between the drought area obtained by RS monitoring and the drought area obtained by field survey shows that VDSI is a responsible reference for agricultural drought application.

The authors appreciate Dr Xiangwu Kang from the Institute of Scientific and Technical Information of China for valuable comments and suggestions for this paper, and Dr Duanyang Xu and Ming Hou for providing the field survey data. We would also like to give thanks to Eric Holt from the UNL Writing Center for correcting the English expressions. The authors thank all anonymous reviewers and the editor for their constructive comments, which significantly improved the quality of this article.

This work was supported by the National Natural Science Foundation of China (Grant No. 42261144749); National Natural Science Foundation of China (Grant Nos. 41877232, 41672255, and 41790444); National Key R&D Program of China (2018YFE0103800); the Fundamental Research Funds for the Central Universities, CHD (300102299302); Key Laboratory Open Project Fund of State Key Laboratory of Loess and Quaternary Geology, Institute of Earth Environment, CAS (Grant No. SKLLQG1909).

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

The authors declare there is no conflict.