International Journal of Biometeorology

, Volume 62, Issue 5, pp 809–822 | Cite as

Assessment of MODIS-derived indices (2001–2013) to drought across Taiwan’s forests

Original Paper

Abstract

Tropical and subtropical ecosystems, the largest terrestrial carbon pools, are very susceptible to the variability of seasonal precipitation. However, the assessment of drought conditions in these regions is often overlooked due to the preconceived notion of the presence of high humidity. Drought indices derived from remotely sensed imagery have been commonly used for large-scale monitoring, but feasibility of drought assessment may vary across regions due to climate regimes and local biophysical conditions. Therefore, this study aims to evaluate the feasibility of 11 commonly used MODIS-derived vegetation/drought index in the forest regions of Taiwan through comparison with the station-based standardized precipitation index with a 3-month time scale (SPI3). The drought indices were further transformed (standardized anomaly, SA) to make them better delineate the spatiotemporal variations of drought conditions. The results showed that the Normalized Difference Infrared Index utilizing the near-infrared and shortwave infrared bands (NDII6) may be more superior to other indices in delineating drought patterns. Overall, the NDII6 SA-SPI3 pair yielded the highest correlation (mean r ± standard deviation = 0.31 ± 0.13) and was most significant in central and south Taiwan (r = 0.50–0.90) during the cold, dry season (January and April). This study illustrated that the NDII6 is suitable to delineate drought conditions in a relatively humid region. The results suggested the better performance of the NDII6 SA-SPI3 across the high climate gradient, especially in the regions with dramatic interannual amplifications of rainfall. This synthesis was conducted across a wide bioclimatic gradient, and the findings could be further generalized to a much broader geographical extent.

Keywords

Drought Leaf water content Normalized Difference Infrared Index (NDII) Seasonal precipitation Spring rainfall Standardized Precipitation Index (SPI) 

Introduction

Concurrence of severe, prolonged droughts and elevated temperature has caused tremendous impacts on terrestrial ecosystems over the recent decade (Overpeck and Udall 2010; Zhao and Running 2010; Dai 2013; Kaptué et al. 2015). Consecutive years of drought events struck both shallow (Anderegg et al. 2015) and deep-rooted (McDowell and Allen 2015) forests and woodlands in western North America, especially California (Belmecheri et al. 2016). Two major droughts (2005 and 2010) over the Amazonian basin pushed the limit of ecological resilience and could potentially neutralize the carbon budget of this largest terrestrial carbon sink of the biosphere (Gatti et al. 2014; Saleska et al. 2017). Recent mega-drought events also substantially reduced forest productivity and ecosystem services in all of Europe (Ciais et al. 2005), the Congolese rainforest of Africa (Zhou et al. 2014), and some regions of Australia (Van Dijk et al. 2013) and East Asia (Saigusa et al. 2010). All of these could result in systems prone to perturbations (e.g., wildfire, insect, and/or pathogen spread) (Allen et al. 2010) and transiting states from original pristine tropical forests to low stature drought endurance woodlands (Fauset et al. 2012; Brando et al. 2014).

Meteorological droughts are recurrent climatic events of a precipitation deficit compared to the long-term average, which can be quantified over a vast region using remotely sensed vegetation and/or drought indices (Quiring and Ganesh 2010; Rhee et al. 2010; Caccamo et al. 2011). Green vegetation time series, such as the Normalized Difference Vegetation Index (NDVI) or the Enhanced Vegetation Index (EVI), can provide spatiotemporally continuous information about vegetation growth and have been widely used to assess the relationships with temperature and/or precipitation over a vast region (Zhang et al. 2010; Vicente-Serrano et al. 2013). The EVI minimizes the constraints of NDVI, such as sensitivity to atmospheric and background effects and insensitivity to dense green vegetation canopies, and responds better to vegetation-climate interactions (Huete et al. 2002). These vegetation indices were developed to monitor the condition of green vegetation in general, but not specifically designed for drought. The Vegetation Condition Index (VCI) was developed based on the maximum and minimum of NDVI time series (Kogan 1994; McVicar and Jupp 1998), which has been applied to detect vegetation drought severity (Quiring and Ganesh 2010). Plant physiological responses to drought can directly reflect land surface temperature (LST). Therefore, Kogan (1997) integrated VCI with Temperature Condition Index (TCI) to develop the Vegetation Health Index (VHI) to portray the temperature-related vegetation stress (Unganai and Kogan 1998; Rhee et al. 2010). The time series of the ratio of land surface temperature (Ts) by NDVI was also suggested as an indicator to assess drought conditions at the country level (McVicar and Bierwirth 2001). A band-depth index, the D 1640 (Caccamo et al. 2011), has been utilized for environmental moisture discrimination (Van Niel et al. 2003). The Normalized Difference Water Index (NDWI) is a ratio-based vegetation index used to enhance sensitivity to leaf water content using near-infrared (NIR) and shortwave infrared (SWIR) spectral reflectance bands, which are highly related to the canopy structure and liquid water absorption, respectively (Gao 1996). Wang et al. (2008) investigated the responsiveness of the NDWI-like indices to leaf water content using the Moderate Resolution Imaging Spectroradiometer (MODIS) band 2 as the reference NIR band, and band 5 (NDWI), 6 (the Normalized Difference Infrared Index-band 6 [NDII6]), or 7 (NDII7) for SWIR. They found that the NDWI was more sensitive than the NDII6 and NDII7 regarding leaf water content from dry to wet soils. Caccamo et al. (2011) showed that the NDII6 was most suitable for drought detection in fire-prone woodlands and forests. The Normalized Multiband Drought Index (NMDI) is composed of MODIS bands 2, 6, and 7, and the performance was superior to those using a single SWIR band based on the manipulation of field spectra (Wang and Qu 2007). The Normalized Difference Drought Index (NDDI) is another drought detection index, derived from the NDVI and NDII7 and was verified as a better indicator of drought in grasslands than the NDVI and NDWI (Gu et al. 2007). Hence, a remotely sensed assessment of vegetation stress may vary mainly due to the substrate complexity and/or vegetation abundance, which can be directly or indirectly modified by regional climate.

The meteorological drought may be effectively evaluated utilizing field meteorological drought indices (Trenberth et al. 2014) such as the Standardized Precipitation Index (SPI, Hirschi et al. 2011) and the Standardized Precipitation Evapotranspiration Index (SPEI, McEvoy et al. 2012; Vicente-Serrano et al. 2012). Both indices can be calculated at different time scales. The SPI only requires precipitation data and has been used to assess drought conditions at different time scales (McKee et al. 1993; Ji and Peters 2003; Hirschi et al. 2011). Ji and Peters (2003) assessed vegetation responses to droughts in the Great Plains in the USA showing that 3-month SPI (SPI3) had the best relationships with the NDVI, and another study conducted in monsoon climate over India suggested that the SPI was a better drought index for the district-wide drought monitoring on the country scale (Pai et al. 2011). The SPEI would require input of precipitation minus potential evapotranspiration (PET). McEvoy et al. (2012) evaluated associations between multiscalar drought indices, SPI and SPEI, with streamflow, lake, and reservoir water surface in arid regions over Nevada and eastern California. They found that the SPEI showed slightly higher correlations over the SPI with water surface variables. Vicente-Serrano et al. (2012) appraised the performance of several drought indices to forecast variations in streamflow, soil moisture, forest growth and crop yield in mid- and high latitude and revealed that the SPEI had better capability to identify drought impacts. However, it is still challenging to utilize SPEI for drought assessment in humid tropical/subtropical climate with dense vegetative surface in which the soil moisture is difficult to measure. Therefore, the World Meteorological Organization (WMO) recommended the SPI as the reference drought index for effectively tracking meteorological drought and risk management (Hayes et al. 2011).

Satellite earth observations have become the mainstream method of monitoring global climate change with the advances in supercomputing technologies in recent years (Nemani et al. 2011). A comprehensive synthesis of the responses of remotely sensed drought indices to various climate-governed biophysical conditions at different time scales is pivotal, but unfortunately lacking in the literature. There is a particular lack for tropical and subtropical humid regions, mainly due to the preconceived notion of the presence of high humidity and high mean annual precipitation (MAP) ≥ 2500 mm year−1. Regardless of the high annual precipitation in Taiwan, the region has suffered increasing droughts since 1960 due to the amplified seasonal rainfall (i.e., drier winter-spring and wetter summer) and regional divergences (wetter northeastern Taiwan and drier central and southern Taiwan) over the past century (Yu et al. 2006; Chen et al. 2009). The ramifications of drought events, especially for the seasonal scale (Doughty et al. 2015), on carbon sequestration in these regions could be even more pronounced where the major terrestrial carbon pools of the earth are (Brienen et al. 2015). Therefore, in this study, we used a humid tropical/subtropical island with detailed meteorological records as a model site to investigate the relationships between a comprehensive set of vegetation drought indices and field precipitation observations across wide bioclimatic gradients. Specifically, our objectives are (i) to examine the correlations between seasonal remotely sensed and field drought indices across precipitation regimes over forest regions and (ii) to assess the feasibility of a comprehensive set of MODIS drought indices to delineate drought conditions, which provide the basic structures used in the following sections: “Materials and methods,” “Results” and “Discussion.”

Materials and methods

Study area

Taiwan, a 36,000-km2 tropical/subtropical mountainous island, is situated between Eurasia and the Pacific Ocean (Fig. 1a). The elevation ranges from sea level to 3952 m a.s.l. in a short horizontal distance (75 km) (Fig. 1b). Southwest and northeast monsoons are the dominant climate systems in summer and winter, respectively (Yen and Chen 2000). MAP is about 2500 mm year−1 with high spatial variability ranging from 1300 to 6700 mm year−1 measured at western coastal (120° 30′ 54.24″ E, 24° 15′ 31.44″ N) and northern mountain (121° 44′ 4″ E, 25° 00′ 15″ N) meteorological stations, respectively. At the regional scale, MAP increases up to 4500 mm year−1 at high elevation and in northeastern Taiwan and drops to 1500 mm year−1 in a western coastal plain (Fig. 1b) (Chang et al. 2014a). More than 75% of MAP falls during the summer growing season (May–October); spring is relatively dry over southwestern Taiwan to the lee side of the Central Mountain Range during the period of northeast monsoons; and the windward northeast Taiwan is relatively wet in winter. Consecutive years below the long-term (1981–2013) average were recorded over 2002−2004 and 2009−2011 (Fig. 2). The mean annual temperature (MAT) of Taiwan is ~ 22 °C in the western flat plain, and it varies through the seasons (± 5 °C) and across topography. According to long-term meteorological records, the MAT of Taiwan increased 0.02 °C per year over the past 33 years, and the rate was accelerated approximately 10 times since 2001 (Fig. 2).
Fig. 1

a Geography of Taiwan. b The distribution of meteorological stations, mean annual precipitation (MAP, contours), and the background of long-term (2001–2013) average of projected green vegetation cover fraction derived from MODIS surface reflectance data (MOD09A1) (Huang et al. 2013). c Elevation contours and land cover type in background

Fig. 2

Mean annual precipitation (MAP) and temperature (MAT) anomalies during the study period (2001−2013, the shaded background) relative to 33-year (1981−2013) averages (MAP = 2590 mm year−1 and MAT = 21.2 °C)

The dominant land uses and vegetation types of the island gradually change along the elevation gradient from urbanized areas and farmland on the coastal plains and tablelands (< 800 m a.s.l.), evergreen broadleaf forests at low and mid elevations (200−2000 m a.s.l.), and mixed and conifer forest at mid and high elevations (> 1100 m a.s.l.). Natural, plantation, and bamboo forests occupy approximately 60%, while farmlands and urbanized areas cover 29 and 6.1% of the land, respectively (Fig. 1c) (Chang et al. 2014a). Since 1991, no logging on natural forests has been allowed in Taiwan, so that there is basically no anthropogenic disturbance. Because of its pronounced bioclimatic and vegetation abundance gradients with a dense meteorological station network (see the next section), Taiwan is an ideal model site for assessing the feasibility of remotely sensed drought indices in the tropical/subtropical humid region.

Data used

Land cover and meteorological data

To illustrate the island-wide pattern of precipitation, we utilized the records from all meteorological stations (n = 390). However, to avoid the uncertainty induced by anthropogenic perturbations, only the stations encompassed by forests (n = 202, Fig. 1b) were used for the analyses. The monthly precipitation (MP) data for the observation period (1981−2013) were acquired from the Data Bank for Atmospheric Research of Taiwan (http://dbar.ttfri.narl.org.tw/) (Fig. S1), and spatial inter-/extrapolation of MP was applied and generated for 2001−2013 using cokriging (ArcGIS v. 10.3, Environmental Systems Research Institute, Inc., CA, USA) combining with precipitation rain gauge data, elevation, and aspect based on a 20-m digital elevation model to take the orographic effect into account. This geostatistical method was used for estimating optimal unbiased environmental variables at an unsampled location and had been used to effectively predict climate-related spatial patterns (Guan et al. 2005; McVicar et al. 2007; Moukana et al. 2013). Chiu et al. (2009) depicted that the precipitation derived from spatial interpolation could be significantly correlated to meteorological data (r > 0.80) with lower prediction errors.

To properly delineate the drought conditions in Taiwan with apparent seasonal rainfall fluctuations (wet summer and dry winter), we used a regionally specific SPI3 (accumulative precipitation for the current month and the two precedent months) as suggested (Chen et al. 2009; Chang et al. 2014b). Please see the supplementary material for the detail definition and calculation of SPI (Figs. S1 and S2).

MODIS vegetation and drought indices

Two tiles (H28V06 and H29V06) of the Terra MODIS 8-day 500-m resolution surface reflectance (MOD09A1) and daytime 1-km resolution LST (MOD11A2) covering the entire study site from 2001 to 2013 were acquired to derive a variety set of vegetation drought index time series (Table 1). The images were projected to the coordinate system WGS-84 UTM 51N using the MODIS Reprojection Tool Web Interface (MRTWeb, https://mrtweb.cr.usgs.gov/), and a maximum value composite procedure was utilized to select the highest quality data from a particular month for analyses (Huete et al. 2002). Therefore, we generated a total of 156 monthly vegetation/drought index images from the 8-day data covering the focal month between 2001 and 2013 from a total of 624 images. To quantify the biophysical background condition, which is directly or indirectly shaped by bioclimate and topography, a straightforward projected green vegetation (GV) cover fraction with values ranging from 0 to 1 was derived from surface reflectance (see Huang et al. 2013 for technical details and supplementary materials). A 13-year (2001−2013) average of GV was used to portray the green vegetation abundance of Taiwan. LST data were resampled to 500 m using the nearest neighbor interpolation to match other data layers (Table 1).
Table 1

Vegetation/drought indices utilized in this study can be derived from the Moderate Resolution Imaging Spectroradiometer (MODIS) reflective (ρ) and/or thermal bands. The subscripts indicate MODIS bands 1 = 620–670 nm, 2 = 841–876 nm, 3 = 459–479 nm, 5 = 1230–1250 nm, 6 = 1628–1652 nm, and 7 = 2105–2155 nm, and the minimum (min) and maximum (max) of the MODIS land surface temperature (LST) product

Vegetation indices

Formula

References

NDVI

(ρ 2 − ρ 1)/(ρ 2 + ρ 1)

 

EVI

2.5 × (ρ 2 − ρ 1)/(ρ 2 + 6 × ρ 1 − 7.5 × ρ 3 + 1)

Huete et al. (2002)

VCI

100(NDVI − NDVImin)/(NDVImax − NDVImin)

Kogan (1994)

Ts/NDVI ratio

LST/NDVI

McVicar and Bierwirth (2001)

D 1640

\( {D}_{1640}=\frac{\rho_6}{\left({\rho}_5\left(1-c\right)+\left(c{\rho}_7\right)\right)} \),

where \( c=\frac{\rho_6-{\rho}_5}{\rho_7-{\rho}_5}\approx 0.4494 \)

Van Niel et al. (2003)

VHI

0.5 × VCI + 0.5 × TCI

where TCI = (LSTmin − LST)/(LSTmax − LSTmin)

Kogan (1994, 1997)

NDWI

(ρ 2 − ρ 5)/(ρ 2 + ρ 5)

Gao (1996)

NDII6

(ρ 2 − ρ 6)/(ρ 2 + ρ 6)

Caccamo et al. (2011)

NDII7

(ρ 2 − ρ 7)/(ρ 2 + ρ 7)

Caccamo et al. (2011)

NMDI

(ρ 2 − (ρ 6 − ρ 7))/(ρ 2 + (ρ 6 − ρ 7))

Wang and Qu (2007), Wang et al. (2008)

NDDI

(NDVI − NDII7)/(NDVI + NDII7)

Gu et al. (2007)

NDVI Normalized Difference Vegetation Index, EVI Enhanced Vegetation Index, VCI Vegetation Condition Index, D 1640 depth of MODIS band 6, VHI Vegetation Health Index, TCI Temperature Condition Index, NDWI Normalized Difference Water Index, NDII6 the Normalized Difference Infrared Index-band 6, NDII7 Normalized Difference Infrared Index-band 7, NMDI Normalized Multiband Drought Index, NDDI Normalized Difference Drought Index

Eleven MODIS-based indices that were commonly used to assess the health or drought conditions of vegetative surfaces can then be calculated using the MODIS surface reflectance and LST data (Table 1). To compare the field-based SPI3 and MODIS-based indices over the vegetated area, the pixel value corresponding to a meteorological station and those of the neighboring pixels (a 3 × 3 window) were averaged to represent the drought indices of a site. The procedure of averaging has been commonly utilized to mitigate errors caused by image misregistration and extreme values (outliers) (Jain et al. 2010; Caccamo et al. 2011; Chmielewski et al. 2014). The final results of monthly maximum NDVI values smaller than 0.3 for forested area were excluded from the following analyses. Such lower value pixels were very likely influenced by clouds and were unsuitable for investigating the associations between vegetation dynamics and climatic conditions (Zhou et al. 2001; Piao et al. 2004). During the study period (156 months with 202 stations), there were a total of 118 pixels excluded which was only accounted less than 0.4% (118/31, 512) and at most 6% (13/202) in a separate month. These data points were also eliminated from calculating other vegetation/drought indices.

Methods

Assessment of the MODIS drought indices

The relationships between monthly satellite drought indices and the SPI3 for each station in forested regions (Fig. 1) were investigated. Geographical patterns of these indices were analyzed by partitioning the stations based on arbitrarily set precipitation (< 2000, 2000−2500, 2500−3000, 3000−3500 and > 3500 mm) and GV (< 0.65, 0.65−0.70, 0.70−0.75, 0.75−0.80, 0.80−0.85 and > 0.85) classes (for the number of meteorological stations for each category, see Table S1). In this study, we applied both original data and the standardized anomalies (SA) of the original data (collectively defined as “drought indices,” hereafter) and selected the one yielding the higher correlation with SPI3. The SA was considered because it could amplify the temporal variation (Caccamo et al. 2011) and has been widely utilized to investigate the relationships between remotely sensed vegetation indices and precipitation (Lloyd-Hughes and Saunders 2002; Chiu et al. 2009; Kaptué et al. 2015). The SA was calculated as follows:
$$ \mathrm{SA}=\frac{x_i-\overline{x}}{\sigma } $$
(1)
Where, x i is the monthly precipitation or MODIS drought index (except VCI and VHI which are already anomalies), and \( \overline{x} \) and σ are the average (2001–2013) of monthly records and standard deviation, respectively. A positive SA suggests greater precipitation or vegetation/drought index than the long-term average and vice versa. In addition, correlations of the selected MODIS drought indices and the SPI3 pairs were superimposed on the map of the mean SA of monthly precipitation spatial coverage to further understand the relationship between drought indices and spatial variations of different precipitation regimes.

Seasonal relationships between satellite and field drought indices

We selected the most suitable drought index that was highly correlated with the SPI3 based upon these aforementioned analyses and further investigated the spatiotemporal relationships between the superior drought index and the SPI3 for some of the months most representative of vegetation stress in the region: the driest, coldest, and hottest months for an average year by referring to the long-term (1981−2013) meteorological records (Figs. 2 and S1), and April, the main spring growing month of the region (Chang et al. 2013). Moreover, the spatial patterns of correlations between the selected satellite drought indices and SPI3 were also evaluated to determine if they were coincident with different precipitation regimes through the observation period (2001–2013).

Results

Satellite drought indices and the SPI3

There were 27.3 ± 22.2% (mean ± standard deviation [sd]) and 35.8 ± 21.4% of station-based SPI3 significantly correlated to original drought indices and those with SA, respectively (r > 0.15, p < 0.05, n = 156), with a maximum of NDII6 (71%) and NDII6 SA (80%) and minimum of NDWI (2%) and NDWI SA (5%) (Fig. 3). NDII indices are the prevailing drought surrogates across biophysical environmental gradients, yielding the most apparent correlations between the SA of MODIS drought indices and the SPI3. There are 55, 9, and 21% of the meteorological stations performing best correlations with SPI3 against SA of NDII6, NDII7 and other indices, respectively (Fig. 3 and Table S1). The majority (66%) of these significant meteorological stations was distributed over central and southwest Taiwan, where the SA of monthly rainfall was pronounced (≤ − 0.1). Meanwhile, the SPI3 derived from 18% of the stations was discordant (r ≤ 0.15, p ≥ 0.05) to any of the temporal dynamics of drought indices or those of SA; most of them (77%) were located in northeast Taiwan, where the SAs of monthly rainfall were greater than 0.0 (Fig. 4). Based on these aforementioned analyses, the NDII6 SA and the commonly used NDVI SA were selected for expediency of the following analysis.
Fig. 3

Mean (n = 202) correlations between the MODIS-derived drought indices (Table 1) and the SPI3. The whiskers indicate the minimum and maximum values; the bottom, top, and middle of each box are the first and third quantiles and median, respectively. Correlations encompassed by the two gray-colored dashed lines are statistically insignificant (p > 0.05). The dots indicate the number of meteorological stations yielding the significant correlations between drought indices and SPI3; the blue lines denote the mean values of correlations between SPI3 and SA of the drought indices

Fig. 4

The maximum correlations between SA of MODIS vegetation/drought indices and the SPI3 for each meteorological station over Taiwan, and mean SA of monthly precipitation (background) during 2001−2013

Seasonal variation of NDII6 SA-SPI3

The relationships between NDII6 SA and SPI3 for the driest and coldest (January), hottest (July), and main spring growing (April) months were investigated to understand the seasonal variations of regional remotely sensed drought conditions and the in situ measurements during 2001–2013. NDII6 SA were significantly correlated to the SPI3 in January (r = 0.49, p < 0.001) and April (r = 0.58, p < 0.001, Fig. 5a) which were better than commonly used NDVI SA (r = 0.34, p < 0.001 for January and 0.24, p < 0.001 for April, Fig. 5b) and other vegetation/drought indices (Table S2) except those for July (r = 0.075, p = 0.102 for NDII6 SA and r = 0.052, p = 0.102 for NDVI SA) (Fig. 5).
Fig. 5

a The relationships between NDII6 SA and SPI3. b The relationships between NDVI SA and SPI3 in the driest and coldest (January), spring growing (April), and the hottest (July) months during the observation period (2001–2013)

The spatiotemporal patterns of NDII6 SA-SPI3 varied for the driest (mean ± sd 101 ± 38.92 mm month−1) and coldest (15.16 ± 0.79 °C) month, January; the hottest month, July (26.35 ± 0.47 °C); and April, based on the long-term (1981−2013) meteorological records. In January, there were 57% (n = 115) of meteorological stations with significant NDII6 SA-SPI3 correlations (r = 0.30−0.90, p < 0.05) distributed over central and southwest Taiwan, where the monthly precipitation SA was less than − 2.5 (Fig. 6a, also see the complete comparisons of all vegetation/drought indices in Table S3). However, there were only 30% (n = 61) of stations with significant NDVI SA-SPI3 correlations (Fig. 6b and Table S3). In April, the main spring growing month, 63% (n = 127) of the stations with significant NDII6-SPI3 correlations (r = 0.30−0.92, p < 0.05) were in the southwest, where the monthly precipitation SA was generally less than − 1.0 (Fig. 6a). In contrast, only 27% (n = 55) of stations had significant NDVI SA-SPI3 correlations, and the pattern was eliminated with April precipitation SA (Fig. 6b). In the hottest month July, there were only 19% (n = 39) and 15% (n = 31) of meteorological stations with pronounced NDII6 SA-SPI3 correlations (r = 0.30−0.44, p < 0.05) and NDVI SA-SPI3 correlations (r = 0.30−0.45, p < 0.05, Table S3). These stations were spread intermittently with no apparent spatial pattern (Fig. 6).
Fig. 6

a The patterns of station-based NDII6 SA-SPI3 correlations and b station-based NDVI SA-SPI3 correlations across mean SA of monthly precipitation in the driest and coldest (January), the spring growing (April), and the hottest (July) months from 2001 to 2013

January and April were selected to further investigate the spatial relationships between NDII6 SA and SPI3 based upon the aforementioned analyses. In general, the distributions of positive (negative) SPI3 were consistent with positive (negative) NDII6 SA during the 13 observation years (data not shown). For simplicity, we selected the climatically extreme years for demonstration (see Table S4 in the supplementary information showing mean [±SD] values of all drought index SA). The patterns of SPI3 were coincident with NDII6 SA in January and April during the driest (2002) and wettest (2005) years, except for some regions located in northeastern Taiwan which had the disharmony of negative SPI3 and positive NDII6 SA pairs (Fig. 7a). In January of the dry years, the SPI3s of 80% (n = 162) of stations were classified as mild drought (mean ± sd = − 0.73 ± 0.25) in 2002, which were associated with both negative NDII6 SA (− 0.03 ± 0.02 in 2002, Fig. 7a) and NDVI SA (− 0.71 ± 0.75 in 2002, Fig. 7b). For April of the dry years, the SPI3s of most stations (≥ 95%, n ≥ 192) labeled as extreme drought (mean ± sd = − 2.33 ± 0.64 in 2002) were geographically related to negative NDII6 SA (− 0.10 ± 0.05 in 2002, Fig. 7a). The drought pattern in April of 2002 was related to the negative NDVI SA (− 0.71 ± 0.99, Fig. 7b). In the wettest year (2005), the SPI3s of most of the stations (≥ 95%, n ≥ 192) were indicated as mild to extreme wet in January (mean ± sd = 0.55 ± 0.33) and April (0.91 ± 0.25), which were consistent with the positive NDII6 SA (0.02 ± 0.02 in January and 0.03 ± 0.04 in April, Fig. 7a and Table S4). The positive SPI3 pattern in January was inconsistent to the negative NDVI SA (− 0.34 ± 0.77), but the positive SPI3 pattern in April was similar to the positive NDVI SA (0.64 ± 0.80, Fig. 7b and Table S4).
Fig. 7

a Spatiotemporal dynamics of the January and April NDII6 SA and SPI3 in the driest (2002) and the wettest (2005) years during the study period (2001−2013). b Spatiotemporal dynamics of the January and April NDVI SA and SPI3 in 2002 and 2005

Discussion

MODIS indices and drought assessment

This study investigates the feasibility of a wide array of drought indices across an approximate 3000-mm MAP gradient. Our results showed that the majority of the indices did not perform well, based on the correlation analysis with the ground-based SPI3 (Fig. 3). We found that all indices involving the visible spectrum (0.4−0.7 μm) and thermal (8−14 μm) bands (Table 1) were, in general, insensitive to the fluctuation of precipitation in this tropical/subtropical region. In the optical region of green vegetation, pigments (mainly chlorophyll a and b, carotenoids, and anthocyanins) in green leaf tissues absorb the majority of solar radiation within the visible range for photosynthesis (Asner and Martin 2009). In the thermal region, the green canopy emits less energy (is cooler) during the daytime, which is due to plant transpiration (Kogan 1997). The thermal response of vegetation to stress conditions is much rapid than the one in other spectral regions mainly due to a disruption of transpiration with the increase of temperature. The change of reflectance in the visible spectral region is not expected to be synchronized with change in temperature. Once drought occurs, spectral reflectance should synchronously increase in the visible region with a reduction in the green vegetation cooling effect, affecting the thermal region due to stress. However, these changes were not detectable in the study area. One possible reason is that the forest ecosystems of Taiwan are structured with multiple layers of plants of different functional types, such as trees, shrubs, and vascular and nonvascular epiphytes, and the stress levels could be different (Jian et al. 2013), thus obscuring satellite drought signals (Huang and Anderegg 2014). Another possibility could be the coarser spatial resolution of MODIS imagery with which the drought conditions might not be detectable without a wide spatial extent.

The only exceptions, with more than 50% of stations yielding statistical significances, are the NIR (0.8−1.3 μm) and SWIR-based drought indices, NDII6 (SWIR1, 1.3−2.0 μm) and NDII7 (SWIR2, 1.3−2.5 μm) (Fig. 3 and Table S1). This suggests that the NIR and SWIR-based spectral indices performed better for evaluating drought conditions than vegetation indices using visible and NIR bands (such as NDVI, EVI, and VCI) and other recently developed indices including NMDI, NDWI, and NDDI. Studies had demonstrated that the NDII6 was not only sensitive to canopy structure and canopy water content in dense vegetation (Townsend et al. 2012; Caccamo et al. 2011), but to soil moisture (Gu et al. 2008) and tree demography (Anderson et al. 2010), which is a superior index for spatiotemporal drought assessment on this humid mountainous island (Taiwan). The NIR region is mainly related to leaf cell structure and canopy thickness (Asner 1998), and healthier vegetation would reflect more energy within this region. In the SWIR region, spectral absorptance occurs due to leaf water content (Kokaly et al. 2009). Reflectance of NIR is quite stable during the drought, but the degrees of the elevation of reflectance within SWIR may vary from species to species (Stimson et al. 2005). In this relatively humid region, the SWIR1 are more sensitive to drought since the performance of NDII6 SA is superior to NDII7 SA (Figs. 3 and 4). This finding is crucial for large-scale monitoring of the drought impact on humid regions (Asner 1998), although this largest terrestrial carbon pool (Kindermann et al. 2008) is recognized as less sensitive to drought (Gatti et al. 2014; Guan et al. 2015; Saleska et al. 2017).

Correlations of ground-measured (the SPI3) and remotely sensed drought conditions across different precipitation regimes can be further enhanced by converting them to a constant scale (SA) (Fig. 4). A more apparent positive–negative trend along the northeast–southwest direction was observed in the SA of MP than the original data (Fig. 4); the correlation between the SPI3 and SA of satellite-derived drought indices (mainly the SA of NDII6 and NDII7) reflects this spatial tendency of dryness. Chen et al. (2009) investigated the trends of rainfall record over the past century in Taiwan and exhibited that 2002 was recognized as the most severe drought in the last century. The SPI3 performed well for characterizing the meteorological drought pattern. Our study demonstrated that the consistency between satellite-derived NDII6 and in situ-based SPI3 is effective in the evaluation of meteorological drought and vegetative water conditions simultaneously in this humid mountainous island with dramatic physical and climate gradients (Fig. 7). Previous studies suggested that comparisons of vegetation and rainfall interactions should be analyzed in a normalized (monthly or annual anomalies) scale rather than an absolute one since the agreement is better between two time series data sets for a given locality (Helldén and Tottrup 2008; Kaptué et al. 2015), which is also confirmed by this regional study.

Sensitivity of NDII6 to seasonal precipitation

Spatiotemporal patterns of the station-based SPI3 were consistent with NDII6 SA (Figs. 5 and 7) during the dry/cold and spring growing months, highlighting the feasibility of MODIS NDII6 to assess seasonal drought across a wide precipitation gradient in this relatively humid tropical/subtropical region. The steeper slope of the regression model between the SPI3 and NDII6 SA (with higher correlation) in April, and their distinct patterns from positive to negative SPI3-NDII6 SA, may indicate the pronounced variations of spring rainfall in this region, which is the precursor to the abundance of forest productivity of that year and is associated with El Niño/La Niña-Southern Oscillation (ENSO) (Chang et al. 2013). Nevertheless, the poor correlations in the summer monsoon season are expected in which the monthly rainfall could reach over 100 mm (Chang et al. 2014b). There is some inconsistency of the positive NDII6 SA and the negative SPI3 pairs in northeastern Taiwan in the dry/cold January of 2002 (Fig. 7). This region still received approximately 2900 mm in the previous 12 months, which should have been sufficient for the basic needs of forests in this region (Guan et al. 2015). The negative SPI might not necessarily indicate an actual drought in some circumstances because the metric is solely based upon meteorological measurement, which could not consider the influence of soil moisture and ground water on plant growth (Bhuiyan et al. 2006).

Many humid tropical/subtropical forests have suffered from extreme drought over the recent decades due to the continually warming climate (Trenberth et al. 2014; Fu 2015). Ecosystems in humid regions are more susceptible to prolonged, recurrent drought events than semiarid regions, which would result in colossal impacts on biogeochemical, hydrological, and energy cycles and endemic species (Saatchi et al. 2013; Zhou et al. 2014). Ecosystems of the tropical/subtropical mountainous island of Taiwan were sensitive to ENSO-induced spring drought (Chen et al. 2003), which could be exacerbated in the foreseeable future (Hanigan et al. 2012; Cai et al. 2015). According to centurial observation and future projection (Hsu and Chen 2002), there was an island-wide warming trend of approximately + 0.1 °C per decade, with the tendency toward a precipitation increase (decrease) in northern (southern) Taiwan and amplified dry and wet seasonal cycles. Elevated temperatures, in concert with the deficit of precipitation, would enhance drought severity (Diffenbaugh et al. 2015), which would have tremendous ramifications on the ecosystem productivity that heavily relies on season rainfall (Chang et al. 2013).

Conclusions

In this study, we utilized 13 years of 11 commonly used MODIS-based drought indices and the SPI3 from a dense meteorological station network to assess the feasibility of satellite monitoring of forest drought conditions in a relatively humid region across an annual precipitation gradient of 1500 to 4500 mm year−1. We found that most drought indices cannot detect the moisture conditions of dense vegetation except for NDIIs, especially NDII6, combining NIR (841−876 nm) and SWIR (1628−1652 nm) spectral regions. With the SA transformation (normalization) of NDII6, we can effectively monitor drought conditions through seasons and across the wide bioclimatic gradient on the driest and coldest month (January), hottest month (July), and the main spring growing month (April). Some inconsistency between NDII6 SA and SPI3 was observed in northeastern Taiwan particularly where the received annual precipitation > 2500 mm even in the driest year, 2002, which might indicate the limitation of SPI3 for drought detection in humid regions that have long preservation of soil moisture. In summary, this study highlights the possibility of large-scale monitoring of forest health in humid regions under ongoing drier climate conditions.

Notes

Acknowledgments

We thank Dr. Tim R. McVicar for his valuable input in improving this manuscript.

Funding information

This study was sponsored by the Ministry of Science and Technology of Taiwan (Grant Numbers MOST 103-2811-M-002-231, 103-2119-M-002-016-, 104-2116-M-002-012, 104-2811-M-002-064, 105-2410-H-002-218-MY3, 105-2811-H-002-024, and 106-2811-H-002-027) and National Taiwan University (106R104516).

Supplementary material

484_2017_1482_MOESM1_ESM.docx (258 kb)
ESM 1 (DOCX 258 kb)

References

  1. Allen CD, Macalady AK, Chenchouni H, Bachelet D, McDowell N, Vennetier M, Kitzberger T, Rigling A, Breshears DD, Hogg EHT, Gonzalez P, Fensham R, Zhang Z, Castro J, Demidova N, Lim JH, Allard G, Running SW, Semerci A, Cobb N (2010) A global overview of drought and heat-induced tree mortality reveals emerging climate change risks for forests. Ecol Manage 259(4):660–684.  https://doi.org/10.1016/j.foreco.2009.09.001 CrossRefGoogle Scholar
  2. Anderegg WRL, Flint A, Huang CY, Flint L, Berry JA, Davis FW, Sperry JS, Field CB (2015) Tree mortality predicted from drought-induced vascular damage. Nat Geosci 8(5):367–371.  https://doi.org/10.1038/ngeo2400 CrossRefGoogle Scholar
  3. Anderson LO, Malhi Y, Aragão LEOC, Ladle R, Arai E, Barbier N, Phillips O (2010) Remote sensing detection of droughts in Amazonian forest canopies. New Phytol 187(3):733–750.  https://doi.org/10.1111/j.1469-8137.2010.03355.x CrossRefGoogle Scholar
  4. Asner GP (1998) Biophysical and biochemical sources of variability in canopy reflectance. Remote Sens Environ 64(3):234–253.  https://doi.org/10.1016/S0034-4257(98)00014-5 CrossRefGoogle Scholar
  5. Asner GP, Martin RE (2009) Spectral and chemical analysis of tropical forests: scaling from leaf to canopy levels. Remote Sens Environ 112:3958–3970CrossRefGoogle Scholar
  6. Belmecheri S, Babst F, Wahl ER, Stahle DW, Trouet V (2016) Multi-century evaluation of Sierra Nevada snowpack. Nat Clim Chang 6:2–3CrossRefGoogle Scholar
  7. Bhuiyan C, Singh RP, Kogan FN (2006) Monitoring drought dynamics in the Aravalli region (India) using different indices based on ground and remote sensing data. Int J Appl Earth Obs Geoinf 8(4):289–302.  https://doi.org/10.1016/j.jag.2006.03.002 CrossRefGoogle Scholar
  8. Brando PM, Balch JK, Nepstad DC, Morton DC, Putz FE, Coe MT, Silverio D, Macedo MN, Davidson EA, Nobrega CC, Alencar A, Soares-Filho BS (2014) Abrupt increases in Amazonian tree mortality due to drought-fire interactions. Proc Natl Acad Sci U S A 111(17):6347–6352.  https://doi.org/10.1073/pnas.1305499111 CrossRefGoogle Scholar
  9. Brienen RJW, Phillips OL, Feldpausch TR, Gloor E, Baker TR, Lloyd J, Lopez-Gonzalez G, Monteagudo-Mendoza A, Malhi Y, Lewis SL, Vásquez Martinez R, Alexiades M, Álvarez Dávila E, Alvarez-Loayza P, Andrade A, Aragão LEOC, Araujo-Murakami A, Arets EJMM, Arroyo L, Aymard C. GA, Bánki OS, Baraloto C, Barroso J, Bonal D, Boot RGA, Camargo JLC, Castilho CV, Chama V, Chao KJ, Chave J, Comiskey JA, Cornejo Valverde F, da Costa L, de Oliveira EA, di Fiore A, Erwin TL, Fauset S, Forsthofer M, Galbraith DR, Grahame ES, Groot N, Hérault B, Higuchi N, Honorio Coronado EN, Keeling H, Killeen TJ, Laurance WF, Laurance S, Licona J, Magnussen WE, Marimon BS, Marimon-Junior BH, Mendoza C, Neill DA, Nogueira EM, Núñez P, Pallqui Camacho NC, Parada A, Pardo-Molina G, Peacock J, Peña-Claros M, Pickavance GC, Pitman NCA, Poorter L, Prieto A, Quesada CA, Ramírez F, Ramírez-Angulo H, Restrepo Z, Roopsind A, Rudas A, Salomão RP, Schwarz M, Silva N, Silva-Espejo JE, Silveira M, Stropp J, Talbot J, ter Steege H, Teran-Aguilar J, Terborgh J, Thomas-Caesar R, Toledo M, Torello-Raventos M, Umetsu RK, van der Heijden GMF, van der Hout P, Guimarães Vieira IC, Vieira SA, Vilanova E, Vos VA, Zagt RJ (2015) Long-term decline of the Amazon carbon sink. Nature 519(7543):344–348.  https://doi.org/10.1038/nature14283 CrossRefGoogle Scholar
  10. Caccamo G, Chisholm LA, Bradstock RA, Puotinen ML (2011) Assessing the sensitivity of MODIS to monitor drought in high biomass ecosystems. Remote Sens Environ 115(10):2626–2639.  https://doi.org/10.1016/j.rse.2011.05.018 CrossRefGoogle Scholar
  11. Cai W, Wang G, Santoso A, McPhaden MJ, Wu L, Jin FF, Timmermann A, Collins M, Vecchi G, Lengaigne M, England MH, Dommenget D, Takahashi K, Guilyardi E (2015) Increase frequency of extreme La Niña events under greenhouse warming. Nat Clim Change 5(2):132–137.  https://doi.org/10.1038/nclimate2492 CrossRefGoogle Scholar
  12. Chang CT, Wang HC, Huang CY (2013) Impacts of vegetation onset time on the net primary productivity in a mountainous island in Pacific Asia. Environ Res Lett 8(4):045030.  https://doi.org/10.1088/1748-9326/8/4/045030 CrossRefGoogle Scholar
  13. Chang CT, Wang SF, Vadeboncoeur MA, Lin TC (2014a) Relating vegetation dynamics to temperature and precipitation at monthly and annual timescales in Taiwan using MODIS vegetation indices. Int J Remote Sens 35(2):598–620.  https://doi.org/10.1080/01431161.2013.871593 CrossRefGoogle Scholar
  14. Chang CT, Wang HC, Huang C (2014b) Retrieving multi-scale climatic variations from high dimensional time-series MODIS green vegetation cover in a tropical/subtropical mountainous island. J Mt Sci 11(2):407–420.  https://doi.org/10.1007/s11629-013-2835-x CrossRefGoogle Scholar
  15. Chen GTJ, Jiang Z, Wu MC (2003) Spring heavy rain events in Taiwan during warm episodes and the associated large-scale conditions. Mon Weather Rev 131(7):1173–1188.  https://doi.org/10.1175/1520-0493(2003)131<1173:SHREIT>2.0.CO;2 CrossRefGoogle Scholar
  16. Chen ST, Kuo CC, Yu PS (2009) Historical trends and variability of meteorological droughts in Taiwan. Hydrol Sci J 54(3):430–441.  https://doi.org/10.1623/hysj.54.3.430 CrossRefGoogle Scholar
  17. Chiu CA, Lin PH, Lu KC (2009) GIS-based tests for quality control of meteorological data and spatial interpolation of climate data. Mt Res Dev 29(4):339–249.  https://doi.org/10.1659/mrd.00030 CrossRefGoogle Scholar
  18. Chmielewski S, Chmielewski TJ, Tompalski P (2014) Land cover and landscape diversity analysis in the West Polesie Biosphere Reserve. Int Agrophys 28:153–162CrossRefGoogle Scholar
  19. Ciais P, Reichstein M, Viovy N, Granier A, Ogée J, Allard V, Aubinet M, Buchmann N, Bernhofer C, Carrara A, Chevallier F, de Noblet N, Friend AD, Friedlingstein P, Grünwald T, Heinesch B, Keronen P, Knohl A, Krinner G, Loustau D, Manca G, Matteucci G, Miglietta F, Ourcival JM, Papale D, Pilegaard K, Rambal S, Seufert G, Soussana JF, Sanz MJ, Schulze ED, Vesala T, Valentini R (2005) Europe-wide reduction in primary productivity caused by the heat and drought in 2003. Nature 437(7058):529–533.  https://doi.org/10.1038/nature03972 CrossRefGoogle Scholar
  20. Dai A (2013) Increasing drought under global warming in observations and models. Nat Clim Chang 3:52–58CrossRefGoogle Scholar
  21. Diffenbaugh NS, Swain DL, Touma D (2015) Anthropogenic warming has increased drought risk in California. Proc Natl Acad Sci USA 112(13):3931–3936.  https://doi.org/10.1073/pnas.1422385112 CrossRefGoogle Scholar
  22. Doughty CE, Metcalfe DB, Girardin CAJ, Amézquita FF, Cabrera DG, Huasco WH, Silva-Espejo JE, Araujo-Murakami A, da Costa MC, Rocha W, Feldpausch TR, Mendoza ALM, da Costa ACL, Meir P, Phillips OL, Malhi Y (2015) Drought impact on forest carbon dynamics and fluxes in Amazonia. Nature 519(7541):78–82.  https://doi.org/10.1038/nature14213 CrossRefGoogle Scholar
  23. Fauset S, Baker TR, Lewis SL, Feldpausch TR, Affum-Baffoe K, Foli EG, Hamer KC, Swaine MD (2012) Drought-induced shifts in the floristic and functional composition of tropical forests in Ghana. Ecol Lett 15(10):1120–1129.  https://doi.org/10.1111/j.1461-0248.2012.01834.x CrossRefGoogle Scholar
  24. Fu R (2015) Global warming-accelerated drying in the tropics. Proc Natl Acad Sci U S A 112(12):3593–3594.  https://doi.org/10.1073/pnas.1503231112 Google Scholar
  25. Gao B (1996) NDWI—a normalized difference water index for remote sensing of vegetation liquid water from space. Remote Sens Environ 58(3):257–266.  https://doi.org/10.1016/S0034-4257(96)00067-3 CrossRefGoogle Scholar
  26. Gatti LV, Gloor M, Miller JB, Doughty CE, Malhi Y, Domingues LG, Basso LS, Martinewski A, Correia CSC, Borges VF, Freitas S, Braz R, Anderson LO, Rocha H, Grace J, Phillips OL, Lloyd J (2014) Drought sensitivity of Amazonian carbon balance revealed by atmospheric measurements. Nature 506(7486):76–80.  https://doi.org/10.1038/nature12957 CrossRefGoogle Scholar
  27. Gu Y, Brown JF, Verdin JP, Wardlow B (2007) A five-year analysis of MODIS NDVI and NDWI for grassland drought assessment over the central Great Plains of the United States. Geophys Res Lett 34:L06407Google Scholar
  28. Gu Y, Hunt E, Wardlow B, Basara JB, Brown JF, Verdin JP (2008) Evaluation of MODIS NDVI and NDWI for vegetation drought monitoring Oklahoma Mesonet soil moisture data. Geophys Res Lett 35(22):L22401.  https://doi.org/10.1029/2008GL035772 CrossRefGoogle Scholar
  29. Guan H, Wilson J, Makhnin O (2005) Geostatistical mapping of mountain precipitation incorporating autosearched effects of terrain and climatic characteristics. J Hydrometeorol 6(6):1018–1031.  https://doi.org/10.1175/JHM448.1 CrossRefGoogle Scholar
  30. Guan K, Pan M, Li H, Wolf A, Wu J, Medvigy D, Lyapustin AI (2015) Photosynthetic seasonality of global tropical forests constrained by hydroclimate. Nat Geosci 8(4):284–289.  https://doi.org/10.1038/ngeo2382 CrossRefGoogle Scholar
  31. Hanigan IC, Butler CD, Kokic PN, Hutchinson MF (2012) Suicide and drought in New South Wales, Australia, 1970–2007. Proc Natl Acad Sci USA 109(35):13950–13955.  https://doi.org/10.1073/pnas.1112965109 CrossRefGoogle Scholar
  32. Hayes M, Svoboda M, Wall N, Widhalm M (2011) The Lincoln declaration on drought indices: universal meteorological drought index recommended. Bull Am Meteorol Soc 92(4):485–488.  https://doi.org/10.1175/2010BAMS3103.1 CrossRefGoogle Scholar
  33. Helldén U, Tottrup C (2008) Regional desertification: a global synthesis. Glob Planet Change 64(3-4):169–176.  https://doi.org/10.1016/j.gloplacha.2008.10.006 CrossRefGoogle Scholar
  34. Hirschi M, Seveniratne SI, Alexandrov V, Boberg F, Boroneant C, Christensen OB, Formayer H, Orlowsky B, Stpanek P (2011) Observational evidence for soil-moisture impact on hot extremes in southeastern Europe. Nat Geosci 4(1):17–21.  https://doi.org/10.1038/ngeo1032 CrossRefGoogle Scholar
  35. Hsu HH, Chen CT (2002) Observed and projected climate change in Taiwan. Meteorol Atmos Phys 79(1-2):87–104.  https://doi.org/10.1007/s703-002-8230-x CrossRefGoogle Scholar
  36. Huang CY, Anderegg WRL (2014) Vegetation, land surface brightness, and temperature dynamics after aspen forest die-off. J Geophys Res Biogeosci 119(7):1297–1308.  https://doi.org/10.1002/2013JG002489 CrossRefGoogle Scholar
  37. Huang CY, Chai CW, Chang CM, Huang JC, Hu KT, Lu ML, Chung YL (2013) An integrated optical remote sensing system for environmental perturbation research. IEEE J Sel Top App Earth Observ Remote Sens 6(6):2434–2444.  https://doi.org/10.1109/JSTARS.2013.2250489 CrossRefGoogle Scholar
  38. Huete A, Didan K, Miura T, Rodriguez EP, Gao X, Ferreira LG (2002) Overview of the radiometric and biophysical performance of the MODIS vegetation indices. Remote Sens Environ 83(1-2):195–213.  https://doi.org/10.1016/S0034-4257(02)00096-2 CrossRefGoogle Scholar
  39. Jain SK, Keshri R, Goswami A, Sarkar A (2010) Application of meteorological and vegetation indices for evaluation of drought impacts: a case study for Rajasthan, India. Nat Hazards 54(3):643–656.  https://doi.org/10.1007/s11069-009-9493-x CrossRefGoogle Scholar
  40. Ji L, Peters AJ (2003) Assessing vegetation response to drought in the northern Great Plains using vegetation and drought indices. Remote Sens Environ 87(1):85–98.  https://doi.org/10.1016/S0034-4257(03)00174-3 CrossRefGoogle Scholar
  41. Jian PY, Hu FS, Wang CP, Chiang JM, Lin TC (2013) Ecological facilitation between two epiphytes through drought mitigation in a subtropical rainforest. PLoS One 8(5):e64599.  https://doi.org/10.1371/journal.pone.0064599 CrossRefGoogle Scholar
  42. Kaptué AT, Prihodko L, Hanan NP (2015) On regreening and degradation in Sahelian watersheds. Proc Natl Acad Sci U S A 112(39):12133–12138.  https://doi.org/10.1073/pnas.1509645112 CrossRefGoogle Scholar
  43. Kindermann GE, McCallum I, Fritz S, Obersteiner M (2008) A global forest growing stock, biomass and carbon map based on FAO statistics. Silva Fenn 42:387–396CrossRefGoogle Scholar
  44. Kogan FN (1994) Application of vegetation index and brightness temperature for drought detection. Adv Space Res 15:91–100CrossRefGoogle Scholar
  45. Kogan FN (1997) Global drought watch from space. Bull Am Meteorol Soc 78(4):621–636.  https://doi.org/10.1175/1520-0477(1997)078%3C0621:GDWFS%3E2.0.CO;2
  46. Kokaly RF, Asner GP, Ollinger SV, Martin ME, Wessman CA (2009) Characterizing canopy biochemistry from imaging spectroscopy and its application to ecosystem studies. Remote Sens Environ 113:S78–S91.  https://doi.org/10.1016/j.rse.2008.10.018 CrossRefGoogle Scholar
  47. Lloyd-Hughes B, Saunders MA (2002) A drought climatology for Europe. Int J Climatol 22(13):1571–1592.  https://doi.org/10.1002/joc.846 CrossRefGoogle Scholar
  48. McDowell NG, Allen CD (2015) Darcy’s law predicts widespread forest mortality under climate warming. Nat Clim Chang 5:699–672CrossRefGoogle Scholar
  49. McEvoy SJ, Huntington JL, Abatzoglou JT, Edwards LM (2012) An evaluation of multiscalar drought indices in Nevada and eastern California. Earth Interact 16(18):1–18.  https://doi.org/10.1175/2012EI000447.1 CrossRefGoogle Scholar
  50. McKee TB, Doesken NJ, Kliest J (1993) The relationship of drought frequency and duration to time series. In: Proceedings of the 8th conference on applied climatology, Boston, USA, pp 179−183Google Scholar
  51. McVicar TR, Bierwirth PN (2001) Rapidly assessing the 1997 drought in Papua New Guinea using composite AVHRR imagery. Int J Remote Sens 22(11):2109–2128.  https://doi.org/10.1080/01431160120728 CrossRefGoogle Scholar
  52. McVicar TR, Jupp DLB (1998) The current and potential operational uses of remote sensing to aid decides on drought exceptional circumstances in Australia: a review. Agric Syst 57(3):399–468.  https://doi.org/10.1016/S0308-521X(98)00026-2 CrossRefGoogle Scholar
  53. McVicar TR, Van Niel TG, Li LT, Hutchinson MF, Mu X, Liu Z (2007) Spatially distributing monthly reference evapotranspiration and pan evaporation considering topographic influences. J Hydrol 338(3-4):196–220.  https://doi.org/10.1016/j.jhydrol.2007.02.018 CrossRefGoogle Scholar
  54. Moukana JA, Asaue H, Koike K (2013) Cokriging for modeling shallow groundwater level changes in consideration of land use/land cover pattern. Environ Earth Sci 70(4):1495–1506.  https://doi.org/10.1007/s12665-013-2235-0 CrossRefGoogle Scholar
  55. Nemani R, Votava P, Michaelis A, Melton AM, Milesi C (2011) Collaborative supercomputing for global change science. EOS 92(13):109–110.  https://doi.org/10.1029/2011EO130001 CrossRefGoogle Scholar
  56. Overpeck J, Udall B (2010) Dry times ahead. Science 328(5986):1642–1643.  https://doi.org/10.1126/science.1186591 CrossRefGoogle Scholar
  57. Pai DS, Sridhar L, Guhathakurta P, Hatwar HR (2011) District-wide drought climatology of the southwest monsoon season over India based on standardized precipitation index (SPI). Nat Hazards 59(3):1797–1813.  https://doi.org/10.1007/s11069-011-9867-8 CrossRefGoogle Scholar
  58. Piao SL, Fang JY, Ji W, Guo QH, Ke JH, Tao S (2004) Variation in a satellite-based vegetation in relation to climate in China. J Veg Sci 15(2):219–226.  https://doi.org/10.1658/1100-9233(2004)015[0219:VIASVI]2.0.CO;2 CrossRefGoogle Scholar
  59. Quiring SM, Ganesh S (2010) Evaluating the utility of the vegetation condition index (VCI) for monitoring meteorological drought in Texas. Agric For Meteorol 150(3):330–339.  https://doi.org/10.1016/j.agrformet.2009.11.015 CrossRefGoogle Scholar
  60. Rhee J, Im J, Carbone GJ (2010) Monitoring agricultural drought for arid and humid regions using multi-sensor remote sensing data. Remote Sens Environ 114(12):2875–2887.  https://doi.org/10.1016/j.rse.2010.07.005 CrossRefGoogle Scholar
  61. Saatchi SS, Asefi-Najafabady S, Malhi Y, Aragão LEOC, Anderson LO, Myneni RB, Nenami R (2013) Persistent effects of a severe drought on Amazonian forest canopy. Proc Natl Acad Sci U S A 110(2):565–570.  https://doi.org/10.1073/pnas.1204651110 CrossRefGoogle Scholar
  62. Saigusa N, Ichii K, Murakami H, Hirata R, Asanuma J, Den H, Han SJ, Ide R, Li SG, Ohta T, Sasai T, Wang SQ, Yu GR (2010) Impact of meteorological anomalies in the 2003 summer on gross primary productivity in East Asia. Biogeosciences 7(2):641–655.  https://doi.org/10.5194/bg-7-641-2010 CrossRefGoogle Scholar
  63. Saleska SR, Didan K, Huete AR, da Rocha HR (2017) Amazon forests green-up during 2005 drought. Science 318:612CrossRefGoogle Scholar
  64. Stimson HC, Breshears DD, Ustin SL, Kefauver SC (2005) Spectral sensing of foliar water conditions in two co-occurring conifer species: Pinus edulis and Juniperus monosperma. Remote Sens Environ 96(1):108–118.  https://doi.org/10.1016/j.rse.2004.12.007 CrossRefGoogle Scholar
  65. Townsend PA, Singh A, Foster JR, Rehberg NJ, Kingdon CC, Eshleman KN, Seagle SW (2012) A general Landsat model to predict canopy defoliation in broadleaf deciduous forests. Remote Sens Environ 119:255–265.  https://doi.org/10.1016/j.rse.2011.12.023 CrossRefGoogle Scholar
  66. Trenberth KE, Dai A, van der Schrier G, Jones PD, Barichivich J, Briffa KR, Sheffield J (2014) Global warming and changes in drought. Nat Clim Chang 4:17–22CrossRefGoogle Scholar
  67. Unganai LS, Kogan FN (1998) Drought monitoring and corn yield estimation in southern Africa from AVHRR data. Remote Sens Environ 63(3):219–232.  https://doi.org/10.1016/S0034-4257(97)00132-6 CrossRefGoogle Scholar
  68. Van Dijk AIJM, Beck HE, Crosbie RS, de Jeu RAM, Liu YY, Podger GM, Timbal B, Viney NR (2013) The Millennium drought in southeast Australia (2001–2009): natural and human causes and implications for water resources, ecosystems, economy, and society. Water Resour Res 49(2):1040–1057.  https://doi.org/10.1002/wrcr.20123 CrossRefGoogle Scholar
  69. Van Niel TG, McVicar TR, Fang H, Liang S (2003) Calculating environmental moisture for per-field discrimination of rice crops. Int J Remote Sens 24(4):885–890.  https://doi.org/10.1080/0143116021000009921 CrossRefGoogle Scholar
  70. Vicente-Serrano SM, Beguería S, Lorenzo-Larcruz J, Camarero JJ, López-Moreno JI, Azorin-Molina C, Revuelto J, Morán-Tejeda E, Sanchez-Lorenzo A (2012) Performance of drought indices for ecological, agricultural, and hydrological applications. Earth Interact 16: no. 10Google Scholar
  71. Vicente-Serrano SM, Gouveia C, Camarero JJ, Begueria S, Trigo R, Lopez-Moreno JI, Azorin-Molina C, Pasho E, Lorenzo-Lacruz J, Revuelto J, Moran-Tejeda E, Sanchez-Lorenzo A (2013) Response of vegetation to drought time-scales across global land biomes. Proc Natl Acad Sci U S A 110(1):52–57.  https://doi.org/10.1073/pnas.1207068110 CrossRefGoogle Scholar
  72. Wang L, Qu JJ (2007) NMDI: a normalized multi-band drought index for monitoring soil and vegetation moisture with satellite remote sensing. Geophys Res Lett 34(20):L20405.  https://doi.org/10.1029/2007GL031021 CrossRefGoogle Scholar
  73. Wang L, Qu JJ, Hao X, Zhu Q (2008) Sensitivity studies of the moisture effects on MODIS SWIR reflectance and vegetation water indices. Int J Remote Sens 29(24):7065–7075.  https://doi.org/10.1080/01431160802226034 CrossRefGoogle Scholar
  74. Yen MC, Chen TC (2000) Seasonal variation of the rainfall over Taiwan. Int J Climatol 20(7):803–809.  https://doi.org/10.1002/1097-0088(20000615)20:7<803::AID-JOC525>3.0.CO;2-4 CrossRefGoogle Scholar
  75. Yu PS, Yang TC, Kuo CC (2006) Evaluating long-term trends in annual and seasonal precipitation in Taiwan. Water Resour Manage 20(6):1007–1023.  https://doi.org/10.1007/s11269-006-9020-8 CrossRefGoogle Scholar
  76. Zhang X, Goldberg M, Tarpley D, Friedl MA, Morisette J, Kogan F, Yu Y (2010) Drought-induced vegetation stress in southwestern North America. Environ Res Lett 5(2):024008.  https://doi.org/10.1088/1748-9326/5/2/024008 CrossRefGoogle Scholar
  77. Zhao M, Running SW (2010) Drought-induced reduction in global terrestrial net primary production from 2000 through 2009. Science 329(5994):940–943.  https://doi.org/10.1126/science.1192666 CrossRefGoogle Scholar
  78. Zhou LM, Tucker CJ, Kaufmann RK, Slayback D, Shabanov NV, Myneni RB (2001) Variations in Northern Vegetation activity inferred from satellite data of vegetation index during 1981 to 1999. J Geophys Res 106(D17):20069–20083.  https://doi.org/10.1029/2000JD000115 CrossRefGoogle Scholar
  79. Zhou L, Tian Y, Myneni RB, Ciais P, Saatchi S, Liu YY, Piao S, Chen H, Vermote EF, Song C, Hwang T (2014) Widespread decline of Congo rainforest greenness in the past decade. Nature 509(7498):86–90.  https://doi.org/10.1038/nature13265 CrossRefGoogle Scholar

Copyright information

© ISB 2017

Authors and Affiliations

  • Chung-Te Chang
    • 1
  • Hsueh-Ching Wang
    • 1
    • 2
  • Cho-ying Huang
    • 1
  1. 1.Department of GeographyNational Taiwan UniversityTaipeiTaiwan
  2. 2.Graduate School of Disaster ManagementCentral Police UniversityTaoyuanTaiwan

Personalised recommendations