Research paper
The roles of NDVI and Land Surface Temperature when using the Vegetation Health Index over dry regions

https://doi.org/10.1016/j.gloplacha.2020.103198Get rights and content

Highlights

  • Drought monitoring with VHI is historically estimated as the plain average between the VCI and TCI terms;

  • It is shown that over global drylands the importance of VCI is larger than TCI;

  • It is shown that drought events are more likely to be detected when using VHI with larger weights given to VCI.

Abstract

Dry lands are expected to cover about half of the terrestrial surface in the near future due to climate change. Drought events, which are recurrent over dry lands, are also projected to increase in both frequency and severity. There is a strong need to better monitor droughts over dry regions, and satellite-based indicators such as the Vegetation Health Index (VHI) have been operationally used worldwide in the last two decades. VHI is traditionally defined as the simple average of two components, the Temperature Condition Index (TCI) and the Vegetation Condition Index (VCI) respectively derived from information on the thermal and visible bands. However, the weights of VCI and TCI depend on landcover because of the different contributions of moisture and temperature to the vegetation cycle. By systematically comparing VHI with the Standardised Precipitation-Evapotranspiration Index (SPEI), a multi-scalar drought index, we demonstrate that is possible to disentangle the role played by VCI and TCI on vegetation health. Here we propose a methodology that allows estimating optimal weights for the two components of VHI and we show that VHI is persistently dominated by VCI over dry lands. Results obtained indicate that severe drought episodes over dry lands are better identified when using the proposed methodology. This may be an asset for operational monitoring, paving the way to more efficient social and political responses aiming to mitigate drought impacts. This work is also expected to contribute to the development of optimal sets of VCI and TCI weights that take into consideration expected changes in the land surface based on information from future climate scenarios.

Introduction

Drought is a recurrent phenomenon associated with natural climate variability in response to external drivers (e.g. solar radiation) and interactions within the climatic system (e.g., El Niño-Southern Oscillation, ENSO; (Barlow et al., 2001; Janicot et al., 1996; Panisset et al., 2018)). Drought ranks among the most devastating natural hazards, debilitating agriculture, economy, and ecosystems, which can lead to widespread human and animal deaths (Cook et al., 2007; Ding et al., 2011; Godfray et al., 2010; Hillier and Dempsey, 2012; Klos et al., 2009; Wilhite et al., 2007). Recently, the concern that drought events may be increasing in frequency and severity due to climate change has grown worldwide (Dai, 2012; Diffenbaugh et al., 2015; Mann and Gleick, 2015; Spinoni et al., 2017, Spinoni et al., 2018; Trenberth et al., 2013; Vicente-Serrano et al., 2014).

In the last decade there has been a growing interest within the scientific community towards the understanding of drought events with the aim of improving monitoring strategies all over the globe (Bolten et al., 2010; Gu et al., 2008; Martínez-Fernández et al., 2016; Park et al., 2016; Quiring, 2009; Sheffield et al., 2014). However, drought episodes are complex to define and, therefore, to monitor. Indeed, one may define drought as an event characterized by below normal levels of water availability over a period in a given region (Wilhite and Glantz, 1985). However, these events may affect different natural and socioeconomic sectors at different time scales, leading to many distinct drought definitions – e.g., meteorological, hydrological, agricultural and socioeconomic droughts (Wilhite, 2005). Meteorological drought is usually defined based on the degree of dryness when compared with the normal value and duration of the dry period (1–2-month time-scales). In turn, agricultural drought considers the characteristics of the meteorological drought (in terms of precipitation, evapotranspiration, and soil moisture deficits) together with soil and crop characteristics (anywhere between 1 and 6-month time-scales). Finally, hydrological drought is defined at longer time scales, when low water supply in streams, reservoirs and groundwater levels become relevant (time-scales from 6 to 48 months or more).

Traditional methods of drought monitoring rely on indices derived from ground-based point observations of variables like temperature and precipitation among others. These spatially limited variables may be difficult to obtain in near-real time and, depending on the type of observation, may be inaccurate and occasionally temporally inconsistent (Sheffield et al., 2012). Examples of such indices are the Palmer Drought Severity Index (PDSI) (Palmer, 1965), the Standardised Precipitation Index (SPI) (Mckee et al., 1993), or the Standardised Precipitation-Evapotranspiration Index (SPEI) (Vicente-Serrano et al., 2010). The latter was developed considering the effect of temperature introduced in the evapotranspiration term, which makes SPEI a useful index to monitor and characterise drought events in a changing climate. Indeed, indices such as PDSI and SPI were designed assuming that the variability of precipitation is much larger than that of temperature and neglecting temporal trends in temperature and their implications to drought assessment. However, several studies have shown that in relatively recent recent events (e.g. 2003 and 2010 in Europe), drought stress was exacerbated by extreme high temperatures and record-breaking heat waves (Barriopedro et al., 2011; Díaz et al., 2006; Rebetez et al., 2006). In recent decades, satellite remote sensing of the Earths' surface and atmosphere has been a widely used tool to monitor drought (Lakshmi, 2017). Indeed, satellite imagery does not suffer from many of the ground-based observations problems since (1) observations, both from polar and geostationary satellites, comprise large areas of the Earths' surface; (2) there is a strong near-real-time operational factor linked to agencies that deploy and operate Earth Observation satellites (e.g., the European Organisation for the Exploitation of Meteorological Satellites – EUMETSAT –, the National Oceanic and Atmospheric Administration – NOAA –, or the National Aeronautics and Space Administration – NASA); (3) products are required to be comprehensively validated and further disseminated to the public only if the errors are below a previously established threshold. Problems may, however, arise related with cloud cover, low temporal resolution of polar orbiting satellites and temporal stability, which may be due to satellite orbital-drift, the use of multiple platforms/ instruments over time, or instrument calibration stability (Li et al., 2013). Furthermore, the length of the time-series may be smaller than ground-based observations because of the relatively small lives of some satellite series. Nevertheless, the current development and improvement of accurate satellite-derived Climate Data Records (CDRs) of several variables such as Land Surface Temperature (LST) (Duguay-Tetzlaff et al., 2015), top-of-atmosphere radiance (Urbain et al., 2017), surface albedo (Lattanzio et al., 2015), or Normalised Difference Vegetation Index (NDVI) (Tucker et al., 2005), among others, spanning decades of satellite information, may mitigate those temporal-related problems by implementing different corrections and methods to the data. In fact, CDRs turn satellite information into a valuable tool to study past droughts and mitigate future catastrophes by continuously improving monitoring techniques (Lakshmi, 2017).

One of the most widely used remote sensing drought indicators is the Vegetation Health Index (VHI) (Kogan, 1997, Kogan, 2001). Several authors have used VHI for different applications that may directly impact society, such as assessing cereal yield losses (Bokusheva et al., 2016; Prasad et al., 2006; Ribeiro et al., 2018; Unganai and Kogan, 1998) or evaluating malaria vector development (Rahman et al., 2011) among others. This index relies on two sub-indices, each based on observations gathered within different windows of the electromagnetic spectrum: the Vegetation Condition Index (VCI) based on the visible and near-infrared portions, which characterises the moisture condition of the vegetation and is generally estimated with NDVI; and the Temperature Condition Index (TCI) based on the thermal infrared window, which characterises vegetation stress steered by temperature conditions. Although the first set of applications made use of top-of-atmosphere brightness temperatures to estimate TCI (Kogan, 1997, Kogan, 2001), LST removes atmospheric effects and limits the impact of emissivity variability on TCI (Bento et al., 2018a, Bento et al., 2018b). VHI is designed as the weighted sum of the VCI and TCI components. Since moisture and temperature contributions to the vegetation cycle depend on vegetation type and this information is not easily available, the common practice is to compute VHI using equal weights of 0.5 for both VCI and TCI. However, as recently shown by Bento et al., 2018a, this issue may be tackled by correlating VHI with a non-satellite based multi-scalar drought index such as SPEI, that also integrates the effect of precipitation and temperature. Indeed, results over a study area covering the Mediterranean region indicate that the weights typically deviate from the traditionally adopted value of 0.5. Results also show that regions with drier vegetation, which are more sensitive to the lack of water than to thermal stress (Gouveia et al., 2008; Karnieli et al., 2010; Lambin and Ehrlich, 1996; Nemani et al., 2003), are consistently characterized by larger contributions of VCI.

Drier regions (drylands and semiarid regions) cover more than one-third of the global terrestrial surface and are home to several different and relevant ecosystems. It has been shown that these areas are sensitive to climate change and anthropogenic land use (Huang et al., 2016; Vicente-Serrano et al., 2015). Therefore, improved drought monitoring and community water use policies (Kahil et al., 2015) are of great importance to prevent future socio-economic disasters related to water scarcity over dry lands.

This work focuses on two distinct, but ultimately related, topics: (1) to estimate the weights of VCI and TCI over global dry regions; and (2) to design an optimal single weight value to be used over drylands. Ultimately, this means that if similar weights are systematically related to these regions, then it is possible to use this representative weight over drylands independently of the user's application approach (e.g. based on different satellite spatial resolutions or in-situ measurements) and obtain a more spatially comprehensive monitoring of drought events with this newly introduced SPEI-based VHI methodology (as proposed in Bento et al., 2018a). The main objective of this study is accordingly to assess the possibility of achieving a simple and sufficiently accurate weight that allows an efficient VHI-based monitoring of current drought events over dry lands. The achievement of this task may pave the way to future works aiming at developing sets of similar weights considering the effects of climate change, namely the increase on terrestrial area of these regions.

Section snippets

Regional setting

The present study is developed at the global scale, covering dry regions over land surfaces, as defined by values between 0.03 and 0.5 of the aridity index from CGIAR – CSI Global-Aridity and Global PET database (Zomer et al., 2006, Zomer et al., 2008). Dry regions accordingly encompass the arid and semiarid band proposed by UNESCO, 1979. Fig. 1 shows the spatial distribution of the regions used for this study, still encompassing a wide latitudinal range. Markers A1 to A8 (with details provided

Satellite-based data

The TCI component is estimated using LST as derived from the dataset developed at Princeton University. Originally consisting of hourly LST data compiled over a 0.5° × 0.5° spatial grid and spanning the period between 1979 and 2009 (Coccia et al., 2015; Siemann et al., 2016), this is a merged dataset, comprising satellite LST from the High-Resolution Infrared Radiation Sounder (HIRS) and from the National Centers for Environmental Prediction (NCEP) Climate Forecast System Reanalysis (CFSR). The

Assessment of VCI and TCI contributions

Fig. 2 shows, over dry lands and on a pixel by pixel basis, the month where the correlation between VHI0.5 and SPEI is maximum. In turn, Figs. 3a,b present histograms of the relative frequencies of each month (with significant correlation) for Northern (NH) and Southern (SH) Hemispheres, respectively. The difference between hemispheres is clear, NH having a peak in frequency by the end of northern spring/summer while SH peaks in austral summer and autumn.

The obtained spatial distribution of

Conclusions

The Vegetation Health Index (VHI) is one of the most used indices to monitor and characterise drought events all over the globe. However, VHI is traditionally estimated as the plain average between its two contributing terms, VCI and TCI. Although this is a reasonable approach since the contributions of the two terms are generally unknown, it has the handicap of not being sensitive to different vegetation growth limiting factors. Indeed, some types of vegetation are more susceptible to the lack

Declaration of Competing Interest

None.

Acknowledgements

Funding: This work was supported by Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP), Brazil/FCT Project Brazilian Fire – Land – Atmosphere System (1389/2014); national funds through the Portuguese Foundation for Science and Technology (FCT), Portugal, under project IMDROFLOOD (WaterJPI/004/2014); and EUMETSAT in the framework of the Satellite Application Facility on Land Surface Analysis (LSA SAF). Renata Libonati was supported by Conselho Nacional de Desenvolvimento Científico e

References (87)

  • A.K. Prasad et al.

    Crop yield estimation model for Iowa using remote sensing and surface parameters

    Int. J. Appl. Earth Obs. Geoinf.

    (2006)
  • J.M. Salmon et al.

    Global rain-fed, irrigated, and paddy croplands: A new high resolution map derived from remote sensing, crop inventories and climate data

    Int. J. Appl. Earth Obs.

    (2015)
  • J. Spinoni et al.

    Pan-European seasonal trends and recent changes of drought frequency and severity

    Glob. Planet. Change

    (2017)
  • R.M. Trigo et al.

    The intense 2007-2009 drought in the Fertile Crescent: Impacts and associated atmospheric circulation

    Agric. For. Meteorol.

    (2010)
  • L.S. Unganai et al.

    Drought monitoring and corn yield estimation in Southern Africa from AVHRR Data

    Remote Sens. Environ.

    (1998)
  • R.J. Zomer et al.

    Climate change mitigation: a spatial analysis of global land suitability for clean development mechanism afforestation and reforestation

    Agric. Ecosyst. Environ.

    (2008)
  • C.T. Agnew

    Using the SPI to identify drought

    Drought Netw. News

    (2000)
  • A. Anyamba et al.

    Interannual variability of ndvi over africa and its relation to el niño/southern oscillation

    Int. J. Remote Sens.

    (1996)
  • A. Anyamba et al.

    From El Niño to La Niña: Vegetation response patterns over east and southern Africa during the 1997-2000 period

    J. Clim.

    (2002)
  • M. Barlow et al.

    ENSO, pacific decadal variability, and U.S. summertime precipitation, drought, and stream flow

    J. Clim.

    (2001)
  • M. Barlow et al.

    Drought in Central and Southwest Asia: La Niña, the warm pool, and Indian Ocean precipitation

    J. Clim.

    (2002)
  • M. Barlow et al.

    A review of drought in the Middle East and Southwest Asia

    J. Clim.

    (2016)
  • D. Barriopedro et al.

    The hot summer of 2010: redrawing the temperature record map of Europe

    Science

    (2011)
  • S. Beguería et al.

    A multiscalar global drought dataset: the SPEI base: a new gridded product for the analysis of drought variability and impacts

    Bull. Am. Meteorol. Soc.

    (2010)
  • V.A. Bento et al.

    Contribution of Land Surface Temperature (TCI) to Vegetation Health Index: a comparative study using clear sky and all-weather climate data records

    Remote Sens.

    (2018)
  • J.D. Bolten et al.

    Evaluating the utility of remotely sensed soil moisture retrievals for operational agricultural drought monitoring

    IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens.

    (2010)
  • B. Bonsal et al.

    Historical comparison of the 2001/2002 drought in the Canadian Prairies

    Clim. Res.

    (2007)
  • G. Churkina et al.

    Contrasting climatic controls on the estimated productivity of global terrestrial biomes

    Ecosystems

    (1998)
  • A. Dai

    Increasing drought under global warming in observations and models

    Nat. Clim. Chang.

    (2012)
  • J. Díaz et al.

    The impact of the summer 2003 heat wave in Iberia: how should we measure it?

    Int. J. Biometeorol.

    (2006)
  • N.S. Diffenbaugh et al.

    Anthropogenic warming has increased drought risk in California

    Proc. Natl. Acad. Sci.

    (2015)
  • Y. Ding et al.

    Measuring economic impacts of drought: a review and discussion

    Disaster Prev. Manag. An Int. J.

    (2011)
  • A. Duguay-Tetzlaff et al.

    Meteosat land surface temperature climate data record: Achievable accuracy and potential uncertainties

    Remote Sens.

    (2015)
  • H.C.J. Godfray et al.

    Food security: the challenge of feeding 9 billion people

    Science

    (2010)
  • C. Gouveia et al.

    The North Atlantic Oscillation and European vegetation dynamics

    Int. J. Climatol.

    (2008)
  • Y. Gu et al.

    Evaluation of MODIS NDVI and NDWI for vegetation drought monitoring using Oklahoma Mesonet soil moisture data

    Geophys. Res. Lett.

    (2008)
  • I. Harris et al.

    Updated high-resolution grids of monthly climatic observations - the CRU TS3.10 Dataset

    Int. J. Climatol.

    (2014)
  • D. Hillier et al.

    A dangerous delay: the cost of late response to early warnings in the 2011 drought in the Horn of Africa

    Oxfam Policy Pract. Agric. Food L.

    (2012)
  • J. Huang et al.

    Global semi-arid climate change over last 60 years

    Clim. Dyn.

    (2016)
  • S. Janicot et al.

    Sahel droughts and Enso dynamics

    Geophys. Res. Lett.

    (1996)
  • A. Karnieli et al.

    Use of NDVI and land surface temperature for drought assessment: Merits and limitations

    J. Clim.

    (2010)
  • G. Karp

    Afghanistan: Six Years of Soviet Occupation

    (1985)
  • R.J. Klos et al.

    Drought impact on forest growth and mortality in the Southeast USA: an analysis using Forest Health and monitoring data

    Ecol. Appl.

    (2009)
  • Cited by (50)

    • A new composite index for global soil plant atmosphere continuum drought monitoring combing remote-sensing based terrestrial water storage and vapor pressure deficit anomalies

      2022, Journal of Hydrology
      Citation Excerpt :

      VCI and RCI represent the vegetation or agricultural drought and hydrological drought respectively. VsmCI and RCI were calculated according to eq.1, while VCI was calculated according to Bento et al., 2020. In addition, two cases, which represent the seasonal drought (the 2010 Russian drought) and prolonged drought (the 2012–2016 California drought) that occurred in regions with dense stations were applied to examine the application performance of WDDI.

    View all citing articles on Scopus
    View full text