Establishing and assessing the Integrated Surface Drought Index (ISDI) for agricultural drought monitoring in mid-eastern China

https://doi.org/10.1016/j.jag.2012.11.003Get rights and content

Abstract

Accurately monitoring the temporal, spatial distribution and severity of agricultural drought is an effective means to reduce the farmers’ losses. Based on the concept of the new drought index called VegDRI, this paper established a new method, named the Integrated Surface Drought Index (ISDI). In this method, the Palmer Drought Severity Index (PDSI) was selected as the dependent variable; for the independent variables, 12 different combinations of 14 factors were examined, including the traditional climate-based drought indicators, satellite-derived vegetation indices, and other biophysical variables. The final model was established by fully describing drought properties with the smaller average error (relative error) and larger correlation coefficients. The ISDI can be used not only to monitor the main drought features, including precipitation anomalies and vegetation growth conditions but also to indicate the earth surface thermal and water content properties by incorporating temperature information. Then, the ISDI was used for drought monitoring from 2000 to 2009 in mid-eastern China. The results for 2006 (a typical dry year) demonstrate the effectiveness and capability of the ISDI for monitoring drought on both the large and the local scales. Additionally, the multiyear ISDI monitoring results were compared with the actual drought intensity using the agro-meteorological disaster data recorded at the agro-meteorological sites. The investigation results indicated that the ISDI confers advantages in the accuracy and spatial resolution for monitoring drought and has significant potential for drought identification in China.

Highlights

► We established a new drought monitoring method called Integrated Surface Drought Index (ISDI) based on the concept of the drought index called VegDRI. ► ISDI integrated multi-source data based on data mining technology. ► ISDI has high accuracy for drought monitoring in large-area as well as on the local scale. ► ISDI has good potentials for drought identification in China.

Introduction

Drought is an important disaster, and its impacts on agriculture are enormous. The drought events also have huge harm to economies, societies and environments (Wilhite, 2002). In recent decades, the impacts of drought have escalated in response to population increase, environmental degradation, industry development, and fragmented government authority in water and natural resources management (Wilhite, 2002). China is a region that is prone to natural disasters. The grain loss caused by drought accounted for 60% of all grain losses caused by meteorological disasters, resulting in 58% or more of economic losses (Li et al., 1999). The frequent occurrence of drought, coupled with the impact of global warming, poses an increasingly severe threat to the Chinese agricultural production (Ma et al., 2004).

Drought differs from other natural disasters, such as floods, typhoons, and earthquakes (Wilhite, 2000). First, the effects of drought often accumulate slowly over a considerable period of time and may linger for years after the termination of the event, and both the onset and end of drought are difficult to determine (Tannehill, 1947). Second, the absence of a precise and universally accepted definition of drought adds to the confusion regarding drought research and identification (Dracup et al., 1980, Gibbs, 1975, Wilhite et al., 1987, Wilhite and Glantz, 1985). Agricultural drought refers to a period with declining soil moisture content and consequent crop failure (Son et al., 2012, Boken et al., 2005). Considering the temporal and spatial complexity of a drought, it is difficult to accurately and quantitatively identify the onset, end, and duration of drought. During the latter part of 20th century, scientists established various drought indices based on different discipline perspectives and their own understanding of the definition of drought. The drought monitoring indices based on traditional meteorological data were developed first (Wilhite and Glantz, 1985). The meteorological indices offer advantages in quantitatively characterizing drought and applicability in different regions. The Standardized Precipitation Index (SPI) based on simple principles has been widely used all over the world, which allows for monitoring and assessment of drought at different timescales, and shows good comparability between different areas (Guttman, 1999, McKee et al., 1993). The Palmer Drought Severity Index (PDSI) is established based on a simplified water balance principle. It takes into account the antecedent precipitation and water supply and demand, involving calculations related to evapotranspiration, soil water supply, runoff, and surface soil water loss (Palmer, 1965, Alley, 1984, Guttman, 1997). The PDSI is suitable not only for monthly drought monitoring but also for weekly drought detection (Bayarjargal et al., 2006, Dai and Trenberth, 1998, WMO, 1975).

Remote sensing technology makes it possible for retrieval of soil moisture, and vegetation conditions across large areas. The Moderate Resolution Imaging Spectroradiometer (MODIS) data plays an increasingly important role in drought monitoring and assessment (Wan et al., 2004), owing to the associated rich spectral information, short revisit cycle, and convenient data access means. The Normalized Difference Vegetation Index (NDVI) is the most widely used indicator of vegetation growth conditions and vegetation coverage, which has been successfully used to estimate vegetation biomass and assess environmental conditions (Chen et al., 2012, Bannari et al., 1995, Justice et al., 1985, Rasmussen, 1998, Tucker and Sellers, 1986); this is why NDVI is widely used in agricultural drought monitoring (Gutman, 1990, Henricksen and Durkin, 1986, Tucker, 1989, Tucker and Choudhury, 1987). Vegetation Condition Index (VCI) is developed based on NDVI time-series data (Kogan, 1990). Compared to NDVI, VCI can better reflect the relationships between the vegetation growth conditions and the precipitation and can minimize the interference of other environmental factors when used to monitor regional agricultural drought during the growth season (Liu and Kogan, 1996, Wang et al., 2001a, Wang et al., 2001b). Percent of Average Seasonal Greenness (PASG) is another index based on historical remote sensing vegetation index sequence (Brown et al., 2008). PASG evaluates vegetation growth conditions by calculating the percentage between the greenness in specific period and the average historical greenness over the same period (Tadesse et al., 2005).

Surface temperature is also an indicator of drought. Drought induces water deficit, which would reduce transpiration and lead to the rise of surface temperature, while relative low temperature stands for the normal healthy status of vegetation under the same condition of vegetation coverage (Jackson et al., 1981, Wan and Dozier, 1996). The Temperature Condition Index (TCI) is developed based on the principle mentioned above and has been widely used in drought monitoring (Kogan, 1995). Vegetation growth conditions will be affected by drought. There will be a decrease in the vegetation index, such as NDVI, and an increase in the canopy temperature because of the stomata closure to minimize water loss by transpiration (McVicar and Jupp, 1998). Thus, the slope of Ts (Surface Temperature)/NDVI can be used to assess the vegetation drought level. The application of the Ts-NDVI method in drought monitoring has been investigated by many researchers (Berliner et al., 1984, Mottram et al., 1983, Pinter et al., 1979). The Vegetation Supply Water Index (VSWI) has been developed based on the theory mentioned above (Bayarjargal et al., 2006, Gillies and Carlson, 1995, Gillies et al., 1997, Price, 1990). Using the Ts/NDVI method, McVicar (2001) rapidly assessed the 1997 drought in Papua New Guinea and validated the effectiveness of this method in drought monitoring (McVicar and Bierwirth, 2001).

Drought is a complex natural disaster. However, each drought index has its own advantages and weaknesses in drought monitoring. Almost all the drought indices are based on specific geographical and temporal scales; it is difficult to spread its applicability all over the world. Because of the meteorological drought indices using discrete, point-based meteorological measurements collected at weather station locations, the results have restricted level of spatial precision for monitoring drought patterns. Remote sensing technology provides alternative data for operational drought monitoring, with advanced temporal and spatial characteristics (Misshra and Singh, 2010). However, additional information still needs to be incorporated so as to thoroughly explain the anomaly in vegetation caused by drought. Besides, to achieve a more accurate description of drought characteristics, drought intensity differences caused by vegetation type, temperature, elevation, manmade irrigation, and other factors under the same water condition must be considered (Kallis, 2008, Zhang et al., 2009). The integration of traditional meteorological data, remotely sensed drought indices, together with information on elevation, vegetation type, and man-made irrigation, provides a promising approach to better characterize the spatial extent and intensity of drought. This research method becomes an urgent problem for further drought investigation.

The Classification And Regression Tree (CART) confers unique advantages in establishing a drought index compared with the traditional statistical regression techniques. It can handle a variety of data types (e.g. nominal, interval, and ratio data), and data without a normal distribution (non-parametric) and hierarchical relationships among variables (Brown et al., 2008). CART can also process large data volumes efficiently and has transparent, interpretable model outputs (De’ath and Fabricius, 2000). Tadesse et al. (2005) first proved the effectiveness of data-mining methods in drought risk management, identification, and monitoring. Using the regression tree model, the Vegetation Drought Response Index (VegDRI) was established (Brown et al., 2008). VegDRI uses PDSI as the dependent variable and incorporates different types of independent variables, including meteorological indices, remotely sensed indices, and biophysical data. Presently, this method has been used for near-real time drought monitoring throughout the United States. However, the VegDRI is still in a stage of improvement, and it needs further refinement. For example, the VegDRI mainly utilizes drought indices calculated from precipitation and vegetation conditions to detect the spatial extent and intensity of drought without other environmental variables, such as land surface temperature. The characteristics of drought indices and methods of optimal inputs selection for an integrated drought index need to be investigated in future research (VegDRI, 2011).

The objective of this paper is to establish a new integrated drought monitoring method named Integrated Surface Drought Index (ISDI) based on the concept of VegDRI. By selecting the best combination of variables with better accuracy, this method integrates the land surface water and thermal environment conditions, vegetation growth conditions, and biophysical information. This paper uses the ISDI to derive drought-detected results by considering the spatial extent and intensity of drought in mid-eastern China from 2000 to 2009. The results of the ISDI are then evaluated in terms of the spatial and temporal aspects to verify the advantage of the newly integrated drought monitoring model with respect to the accuracy, spatial resolution, and the application potential for drought identification in China.

Section snippets

Study area

Mid-eastern China was chosen as the study area (Fig. 1), covering 16% of the whole country (31–44.5°N, 109–123°E). This area includes 11 provinces and is mainly located in the semi-arid and arid area that span middle latitudes and in the semi-humid and semi-arid area of the warm temperate zone (Zheng and Li, 2008). Historically, severe droughts have frequently occurred in mid-eastern China because of the uneven spatial distribution of precipitation caused by the East Asian monsoon climate. The

Remote sensed data processing

The Terra MODIS LST and NDVI products were processed to remove cloud-contaminated pixels using the quality control documents before used as inputs to calculate drought indices. The NDVI data sets are 16-day MVC products. However, there are still noise in the NDVI time series caused by cloud contamination and atmospheric variability, which greatly affects the accuracy of the vegetation dynamics (phenology) measurements and drought indices. In this paper, a simple, but robust, method based on the

Precision of the integrated model construction

The specific construction schemes of the integrated drought index are shown in Table 5.

The 12 schemes in Table 5 can be divided into 3 groups according to the number of independent variables in the model: schemes (1)–(6) with 7 independent variables, schemes (7)–(10) with 8 independent variables and schemes (11) and (12) with 9 independent variables. Comparison of the 3 groups of scheme shows that schemes (7)–(10) have the highest overall accuracy, with the correlation coefficient larger than

Conclusions

This research comprehensively monitored drought characteristics in terms of vegetation growth condition, surface water and thermal environment, and biophysical information. The ISDI was established using data-mining technology, which uses the PDSI as a dependent variable and eight other factors as independent variables based on the traditional meteorological drought data, remotely sensed data, and biophysical data. Previous studies have demonstrated that the land surface temperature has a

Acknowledgments

This work was funded by the National Natural Science Foundation of China through the project 'Investigation of Regional Agricultural Drought Monitoring Approach Based on Simulation of Crop Growth Process (41171403)’ and the CRSRI Open Research Program (CKWV2012320/KY). The authors would like to thank Ruijing Jia for her helpful editorial support and the reviewers of the manuscript for their helpful comments. We are also grateful to Dr. Aifeng Lü, Bin He, Lin Zhao, and Xinyu Mo for their useful

References (65)

  • A. Bannari et al.

    A review of vegetation indices

    Remote Sensing Reviews

    (1995)
  • V.K. Boken et al.

    Monitoring and Predicting Agricultural Drought: A Global Study

    (2005)
  • J. Brown et al.

    The Vegetation Drought Response Index (VegDRI): a new integrated approach for monitoring drought stress in vegetation

    GIScience Remote Sensing

    (2008)
  • J. Chen et al.

    Scale correction of two-band ratio of red to near-infrared using imagery histogram approach: a case study on Indian remote sensing satellite in Yellow River Estuary

    IEEE Journal of Selected Topics in Applied Earth Observation

    (2012)
  • A. Dai et al.

    Global variations in droughts and wet spells: 1900–1995

    Geophysical Research Letters

    (1998)
  • J.A. Dracup et al.

    On the definition of droughts

    Water Resources Research

    (1980)
  • G. De’ath et al.

    Classification and regression trees: a powerful yet simple technique for ecological data analysis

    Ecology

    (2000)
  • W.J. Gibbs

    Drought – its definition, delineation and effects

    (1975)
  • R.R. Gillies et al.

    Thermal remote sensing of surface soil water content with partial vegetation cover for incorporation into climate models

    Journal of Applied Meteorology

    (1995)
  • R.R. Gillies et al.

    A verification of the 'triangle’ method for obtaining surface soil water content and energy fluxes from remote measurements of the Normalized Difference Vegetation Index (NDVI) and surface radiant temperature

    International Journal of Remote Sensing

    (1997)
  • G.G. Gutman

    Towards monitoring droughts from space

    Journal of Climate

    (1990)
  • N.B. Guttman

    Comparing the Palmer Drought Index and the Standardized Precipitation Index

    Journal of American Water Resource Association

    (1997)
  • N.B. Guttman

    Accepting the standardized precipitation index: a calculation algorithm

    Journal of American Water Resource Association

    (1999)
  • B.L. Henricksen et al.

    Growing period and drought early warning in Africa using satellite data

    International Journal of Remote Sensing

    (1986)
  • R.D. Jackson et al.

    Canopy temperature as a crop water stress indicator

    Water Resources Research

    (1981)
  • C.O. Justice et al.

    Analysis of the phenology of global vegetation using meteorological satellite data

    International Journal of Remote Sensing

    (1985)
  • G. Kallis

    Droughts

    Annual Review of Environment and Resources

    (2008)
  • F.N. Kogan

    Remote sensing of weather impacts on vegetation in non-homogeneous areas

    International Journal of Remote Sensing

    (1990)
  • F.N. Kogan

    Droughts of the late 1980 in the United States as derived from NOAA polar-orbiting satellite data

    Bulletin of American Meteorological Society

    (1995)
  • F.N. Kogan

    Global drought watch from space

    Bulletin of American Meteorological Society

    (1997)
  • K. Li et al.

    China Drought Research and Disaster mitigation measures

    (1999)
  • W.T. Liu et al.

    Monitoring regional drought using the Vegetation Condition Index

    International Journal of Remote Sensing

    (1996)
  • Cited by (122)

    View all citing articles on Scopus
    View full text