Monitoring vegetative drought dynamics in the Brazilian semiarid region

Agricultural and Forest Meteorology 214–215 (2015) 494–505

Contents lists available at ScienceDirect

Agricultural and Forest Meteorology journal homepage: www.elsevier.com/locate/agrformet

Monitoring vegetative drought dynamics in the Brazilian semiarid region A.P.M. Cunha ∗ , R.C. Alvalá, C.A. Nobre, M.A. Carvalho Brazilian Centre for Monitoring and Warning of Natural Disasters, São José dos Campos, SP, Brazil

a r t i c l e

i n f o

Article history: Received 13 August 2015 Received in revised form 9 September 2015 Accepted 18 September 2015 Keywords: Drought indices Drought monitoring Vegetative drought Vegetation index Semiarid

a b s t r a c t Drought is a complex natural phenomenon that can lead to reduced water supplies and can consequently have substantial effects on agriculture and socioeconomic activities that cause social crises and political problems. Different drought indicators are used for identifying droughts. This work explored the applicability of a near-real time drought monitoring methodology using Terra-MODIS Normalized Difference Vegetation Index (NDVI) and land surface temperature (LST) products. This approach is called the Vegetation Supply Water Index (VSWI), which integrates land surface reflectance and thermal properties. The results indicate that during a major drought event from 2012 to 2013, approximately 85% of the Brazilian semiarid region was affected. The number of days of soil moisture deficit, which was derived from a simple water balance model and the daily interpolated precipitation, were used to verify the results. A correlation analysis of VSWI, precipitation and soil moisture deficit shows that VSWI is closely related to rainfall and soil water content, especially under dry conditions, and indicates that the use of VSWI can be a suitable near-real time drought monitoring approach. The evaluation of the 2012–2014 drought considering the VSWI index highlighted two major characteristics of vegetation response to drought conditions, i.e., the recovery and memory effects of vegetation. © 2015 Elsevier B.V. All rights reserved.

1. Introduction Drought is basically defined as an extended period of months or years, in which precipitation is less than the annual average, resulting in water scarcity. Generally droughts are classified as either a meteorological drought (lack of precipitation over a region for a period of time), hydrological drought (deficiencies in surface and subsurface water supplies), agricultural drought (deficiency in water availability for crop or plant growth) or socioeconomic drought (failure of water resources systems to meet water demands, which impacts human activities both directly and indirectly) (Wilhite, 2000; Yang, 2010; Son et al., 2012; Udmale et al., 2014). Although precipitation deficiencies are important, agricultural drought severity is usually more closely associated with deficiencies in soil moisture. The areas affected by drought evolve gradually as the symptoms of moisture stress in plants often develop slowly. These impacts of drought on vegetation are here referred as vegetative drought (Rulinda et al., 2012). Several studies have documented that uneven temporal distributions of precipitation and rising temperatures have caused vegetation shifts, and

∗ Corresponding author. E-mail address: [email protected] (A.P.M. Cunha). http://dx.doi.org/10.1016/j.agrformet.2015.09.010 0168-1923/© 2015 Elsevier B.V. All rights reserved.

these studies have shown a newly directional change pattern of vegetation communities in response to climate change and the consequent dryness (Zhou et al., 2011, 2013a,b, 2014). Drought is considered to be among those natural disasters that can cause the most serious global economic and social losses (Carolwicz, 1996) and affects more people than any other natural disasters (Keyantash and Dracup, 2002). 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). In recent decades, droughts have increased in frequency and intensity over much of the planet and can be related to climate change (Marengo et al., 2009; Zhou et al., 2011). The percentage of area that is affected by drought has doubled from the 1970s to the early 2000s (Nagarajan, 2009). In Brazil, this phenomenon has occurred mainly in the semiarid area of Northeast Brazil (SANEB) due to uneven precipitation in space and time. Seasonal droughts usually occur in the winter and spring and have a significant impact on agricultural harvests. The severe droughts that are caused by climactic variations harm the growth on plantations and cause serious social problems because a large number of people who inhabit the region truly live in a situation of extreme poverty (Marengo et al., 2009). More than 80% of the agricultural establishments in

A.P.M. Cunha et al. / Agricultural and Forest Meteorology 214–215 (2015) 494–505

Northeast Brazil consist of smallholder farmers that practice subsistence agriculture (IBGE, 2009). Thousands of subsistence farmers have seen their livelihoods wither away during drought episodes. Other consequences of large droughts are starvation, malnutrition, misery and migration to urban centers (rural exodus). Thus, the livelihood and food security of millions of smallholder farmers are exposed to profound risk from drought. In Brazil, from 1980 to 2010 approximately 15% of natural disasters were related to drought conditions in the Northeast Region (NEB). Drought affects more people than any other natural disaster; since 1980 more than 50 million people have been affected by drought (Sapir and Below, 2014). The recent drought of 2012/2013 reached approximately 1300 municipalities and affected approximately 10 million people in Brazil. Drought is, on average, Brazil’s most costly natural hazard, primarily because it causes hard impacts on agriculture and livestock production. For example, the 2012/2013 drought resulted in economic losses of US $1.6 billion for the 10 most important crops (beans, rice, corn, cotton, bananas, sugar cane, cassava, soybeans and coffee), US $1.5 billion due to cattle mortality and costs of greater than US $1.5 billion in insurance claims, according to the Brazilian Institute of Geography and Statistics (IBGE). The quantification of drought is usually determined by remotely sensed spectral indices and water balance simulations. Drought indices are particularly useful for monitoring the impact of climate variability on vegetation because the spatial and temporal identification of drought episodes is extremely complex. A number of drought indices, including meteorological (Wilhite and Glantz, 1985), remote sensed, hydrological and other indicators, have been used to measure drought impacts (Palmer, 1965, 1968; Gibbs and Maher, 1967; Shafer and Dezman, 1982; Kogan, 1990, 2002; McKee et al., 1993; Keyantash and Dracup, 2004; Bhuiyan et al., 2006; Yagci et al., 2011; Zhou et al., 2012; Du et al., 2013; Yang et al., 2013; Abbas et al., 2014; Nichol and Abbas, 2015). Traditional methods of drought assessment and monitoring rely on rainfall data (e.g., the Palmer drought severity index (PDSI) and Standardized Precipitation Index (SPI)). However, in a region where the density of meteorological stations as well as the temporal scale of the data are insufficient, it is impossible to monitor drought using indices that are based on rainfall data. In contrast, satellite-sensor data are consistently available and can be used to detect the onset of drought, its duration and magnitude across large areas (Thiruvengadachari and Gopalkrishna, 1993). Remote sensing has proven to be a powerful tool for evaluating the temporal and spatial aspects of drought conditions (Johnson et al., 1993; Peters et al., 2002). Moderate Resolution Imaging Spectroradiometer (MODIS) data play an increasingly important role in drought monitoring and assessment (Wan et al., 2004) because of their associated rich spectral information, high temporal repeat cycle and convenient means of data access. The Normalized Difference Vegetation Index (NDVI), which provides a general measure of the state and health of vegetation, was one of the first remote sensing-based indicators that was used for drought detection and monitoring. Many studies have reported relationships between vegetation indices, rainfall and soil moisture (Davenport and Nicholson, 1993; Herrmann et al., 2005; Liu et al., 2013; Ibrahim et al., 2015). This is an important reason why NDVI is widely used in agricultural drought monitoring (Henricksen and Durkin, 1986; Tucker and Choudhury, 1987; Tucker, 1989; Gutman, 1990). On the other hand, vegetation cover condition, as sampled by vegetation index, is a relatively slow response variable that typically adjusts only after notable crop damage has already occurred. In contrast, land surface temperature (LST) derived from thermal infrared (TIR) information can be considered to be a rapid response variable. LST is a good indicator of the energy balance at the Earth’s surface because it is one of the key parameters in the physics of land-surface processes on regional and

495

global scales. Researchers have concluded that the combination of vegetation and temperature conditions is a good indicator of soil moisture content. A large number of vegetation health and drought indices are based on the LST–NDVI space such as Vegetation Health Index (VHI, Kogan, 1990, 1997), Temperature-Vegetation Drought Index (TVDI, Sandholt et al., 2002), Vegetation Supply Water Index (VSWI, Carlson et al., 1990, 1994), and Drought Severity Index (DSI, Mu et al., 2013). These drought indices are often applied for arid or semi-arid regions, and their use in humid–semi humid regions are limited (Rhee et al., 2010). The NDVI–LST relationship, which characterizes moisture and thermal conditions and the entirety of vegetation health, has been used successfully for early drought detection and the estimation of crop and pasture production losses for winter wheat in the USA (Kogan, 1997). The potential for stress exists when the water stored in soil is insufficient to sustain the current growth. In the vegetation covering areas, LST can be considered equal to the temperature of vegetation canopy (Liu et al., 2013). The canopy temperature response can occur even when the plants are green because stomata closure to minimize water loss by transpiration results in a decreased latent heat flux (Berliner et al., 1984; Carlson et al., 1994; Yang and Merchant, 1997). Because of the territorial expansion of the Northeast region of Brazil, it is necessary to develop methods for large-scale vegetative drought assessment. To direct the emergency actions of the government that are taken to mitigate the effects of drought, it is crucial to determine an appropriate and user friendly index that reflects the direct impact of drought on livestock and subsistence agriculture. Thus, the main objective of this study is to evaluate possible indicators to monitor the impacts of drought on vegetation over the NEB. In this paper, the Vegetation Supply Water Index (VSWI) performance is compared with precipitation data and soil water deficit. This study focuses on the diagnosis of a remote sensing indicator that is responsive to short-term environmental changes because early warning capabilities are limited in current drought monitoring systems.

2. Materials and methods The study area (Fig. 1a) is located in the equatorial zone (1–21◦ S, 32–49◦ W) and covers an area of 1,800,555 km2 , which represents approximately 20% of Brazil’s territory. The limits of the study area were defined by the “Superintendence for the Development of the Northeast (SUDENE)”. The semiarid area of Northeast Brazil covers an area of 980,323 km2 and consists of 1133 municipalities and a population of approximately 22 million people (approximately 12% of the national population). These numbers make the Brazilian semiarid region the most populated semiarid region in the world. In the SANEB, rural areas in the interior are generally used for subsistence agriculture that is primarily comprised of beans, manioc, potatoes and other crops (Cavalcanti et al., 1999). Most of the study area is covered by mixed grasslands–croplands (Fig. 1b). Other land cover types are caatinga (closed and open shrublands) and savanna (not shown). In 2010, the total area of pasture and agricultural activities was 1,024,621 km2 , which represents 57% of the NEB territory (Vieira et al., 2013). Northeast Brazil is characterized by a variable and irregular spatio-temporal distribution of precipitation. The rainfall ranges from less than 800 mm/year in the semiarid interior to more than 1500 mm/year in the rainy climatic zone that is mainly on the east coast (Fig. 2a). Different rainfall regimes have been identified in the NEB (Fig. 2b). (i) In the South-Southwest sector, the main rainy season is from November to February, and the rainfall is associated with cold fronts traveling from the South Region of Brazil. (ii) In the

496

A.P.M. Cunha et al. / Agricultural and Forest Meteorology 214–215 (2015) 494–505

Fig. 1. (a) Location of Northeast Brazil and (b) distribution of mixed grasslands–croplands areas in the study region.

Fig. 2. (a) Mean annual rainfall 1970–2012 and (b) wet seasons regions (FMAM: February to May; JFMA: January to April; DJFM: December to March; NDJF: November to February; AMJJ: April to July).

A.P.M. Cunha et al. / Agricultural and Forest Meteorology 214–215 (2015) 494–505

497

composited land surface temperature (LST) products (MOD11A2, MYD11A2, 1 km × 1 km) from 2002 to 2014 in the study area were acquired from the U.S. National Aeronautics and Space Administration (NASA) Land Processes Distributed Active Archive Center (LP DAAC). MODIS, which is carried on the Terra-Aqua satellite, is a cost-effective sensor that covers the globe at least once a day. MODIS, which is a successor of the Advanced Very High Resolution Radiometer (AVHRR), is the primary sensor in the NASA Earth Observing System (EOS) program for monitoring the terrestrial ecosystem (Justice et al., 2002) and has several advances over AVHRR (Thenkabail et al., 2004). MODIS is more sensitive to changes in vegetation dynamics (Huete et al., 2002) and has been found to be a more accurate and versatile instrument for monitoring global vegetation conditions than AVHRR (Gitelson and Kaufman, 1998; Justice et al., 2002). Composite MODIS data have a temporal resolution of 8 days and are available from 2000 onwards. The 8-day, 7-band data are made available by the USGS EROS DAAC after corrections for molecular scattering, ozone absorption and aerosols. The data are also adjusted to nadir (sensor looking straight down) and standard sun angles using bidirectional reflectance (BRDF) models (Vermote et al., 2002; Justice et al., 2002). The products have been processed for atmospheric and geometric corrections.

Fig. 3. Spatial distribution of location of the weather stations.

Northern sector (semiarid zone), the rainfall occurs from February to May, January to April and December to May, and the rainfall is associated with the southward movement of the Intertropical Convergence Zone (ITCZ). (iii) In the East sector, the main rainy season occurs from April to August and is mainly caused by greater temperature differences between the ocean and nearby land. 2.1. Data 2.1.1. Precipitation The data that were used in this study include in situ meteorological data and remotely sensed data. The meteorological data were acquired from different sources, including weather stations (Data Collection Platform) from the National Institute for Space Research (INPE), National Institute of Meteorology (INMET), National Center for Monitoring and Early Warning of Natural Disasters (CEMADEN) and the State Centers of Meteorology. Fig. 3 shows the spatial distribution of the 1947 weather stations for which we had data. Rigorous quality inspection was carried out for the daily precipitation data. The interpolation of the data from the weather stations into a regular grid with 5 km resolution was performed using a technique called “kriging” that was developed by (Matheron, 1969) and has been primarily applied in hydrology and other earth science disciplines for the spatial interpolation of various physical quantities given a number of spatially distributed measurements. The spatially interpolated data were also used as an input to a water balance model that was developed by PROCLIMA/INPE (Souza et al., 2001) and that is used to estimate soil moisture and the number of days of soil moisture deficit with a spatial resolution of 5 km. 2.1.2. NDVI and LST The 16-day MODIS NDVI time series product (MOD13A2, MYD13A2, 1 km × 1 km) and an 8-day average value of the

2.1.2.1. Acquisition and processing MODIS data. MODIS images are available on a daily basis, but their use involves considerable extra processing. Time series of MODIS imagery provide near real-time, continuous and relatively high-resolution data. These images can used to assess the development of drought and its severity in regions with scarce and inaccurate on-the-ground meteorological observations. The daytime MOD11A2 and MYD11A2 products consist of 12 Science Data Sets (SDSs), including 8-day composite LST and the quality of each LST pixel. The LST values in Kelvins are encoded in a 16-bit unsigned integer that ranges from 7500 to 65,535. To derive the actual temperature values, a multiplication factor of 0.02, as stated in MODIS product manual, was used. The LST data from two consecutive periods were averaged to generate composite LST for the same periods as the NDVI to match the 16-day NDVI composite product (MOD13A2 and MYD13A2). NDVI values that were less than or equal to zero were excluded from further analysis because pixel values less than or equal to zero are assumed to represent either cloud contaminated imagery or the presence of a water body. Moreover, NDVI and LST images were filtered using a mask of the water bodies that are in the NEB region. Finally, digital maps of NDVI and LST products were generated for the entire research area over the course of 10 years (2002–2014).

2.2. Drought-monitoring products 2.2.1. Soil moisture deficit Soil moisture condition is an important indicator for evaluating drought, reflects recent precipitation and antecedent conditions and indicates agricultural potential and available water storage (Boken, 2005). Soil moisture conditions are very important in agriculture because they are used directly to assess the irrigation needs for a variety of crops. In this study, the number of days of soil moisture deficit (NDD), which is derived from a simple water balance model, is one of the indicators that was used to characterize the drought. Days of soil moisture deficit measures the number of days that a plant’s growth is restricted by insufficient moisture in the soil. The NDD in each pixel is calculated from the soil moisture data.

498

A.P.M. Cunha et al. / Agricultural and Forest Meteorology 214–215 (2015) 494–505

Soil moisture is calculated daily from the water balance in the soil (Souza et al., 2001; Rossato et al., 2005): Qs(t) = prec(t) − evap(t) − DP(t) + qs0

(1)

The model water balance components are soil water content at time t (Qs(t)), precipitation (prec), actual evapotranspiration (evap), drain depth (DP) and soil water content for the previous day (qs0 ). This soil water balance model developed for Brazil (Rossato et al., 2005), uses a new approach that takes into account the spatial variability of soil characteristics. The maximum soil water storage was derived from the field capacity and the wilting point, using a pedotransfer function (PTF). PTF allows the estimation of soil hydraulic properties from basic soil data (such as texture, particlesize distribution, organic C content and bulk density). Tomasella developed specific PTFs for the prediction of soil water retention curves for the tropical soils of Brazil (Tomasella and Hodnett, 1998; Tomasella et al., 2000, 2003, 2008). Once we have the measured data, including the temperature, wind and radiation, the FAO-56 Penman–Monteith method (Allen et al., 1998) was applied to determine reference evapotranspiration. The vegetation parameters needed for the Penman–Monteith method were those obtained from the SSiB model (“Simplified Simple Biosphere”, Xue et al., 1991), as cited by Dorman and Sellers (1989). Calibrated vegetation parameters for semiarid Brazilian (stress conditions) which were used, are described in Cunha et al. (2009) and Cunha et al. (2013) (plant height, root depth, vegetation roughness length, surface resistance, albedo, etc.). Thus, the reference evapotranspiration was calculated according to the land cover present in the study area. For calculating the water balance, the reference evapotranspiration was “transformed” into actual evapotranspiration, as suggested by the FAO method (Doorenbos and Pruitt, 1977), i.e., considering that the amount of water that plants transpire depends on the water storage in soil. The soil water content was calculated by combining data from weather stations with a soil database in a geo-referenced environment. The soil water deficit is considered to be low when the soil moisture falls below a critical value. The critical value is defined as the soil moisture at the point where soils are “extremely drier than normal” and drought conditions are likely to be evident. It was assumed that the critical soil matric potential threshold was −60 kPa (Feddes et al., 1988). Therefore, NDD is the sum of days during which the water loss is higher than the soil water retention. Additional details about these methodologies can be found in Souza et al. (2001). 2.2.2. Vegetation Supply Water Index (VSWI) calculation Growing crops need continuous supplies of soil water to ensure harvest. Rainfall and irrigation are the main sources of soil water in agricultural fields. When the soil water supply is sufficient for growing crops, evapotranspiration from agricultural fields is high, which leads to the observation of low surface temperatures in satellite remote sensing images. During a drought period, the soil water supply is insufficient to meet the normal demands for growing crops. Consequently, the stoma on crop leaves tend to close to decrease water loss from the canopy, which leads to an apparent increase of temperature in the fields. Therefore, by using the relationship between the canopy temperature change and the soil water supply in the fields, it is possible to develop an approach for drought monitoring (Gao et al., 2008; Wu et al., 2015). VSWI has been widely applied as a drought indicator based on the philosophy of this approach. The VSWI approach, which is an additive combination of NDVI and the thermal data (land surface temperature – LST), was established to detect vegetation stress, moisture and drought-affected areas (Zhou et al., 2013a,b). VSWI can be used to monitor the onset, change, development and intensity of drought and the extent of the impact on vegetation. VSWI is

calculated by the ratio of the average 8-day LST product versus the smoothed high-quality NDVI value in the MODIS composted data: VSWIijk =

0.02(LSTijk )

(2)

0.001(NDVIijk )

For pixel i in period j for year k. Land surface temperature can be considered similar to canopy surface temperature when ground covered with crop (Chen et al., 2014). High VSWI values indicate a high canopy temperature and low vegetation index and therefore may indicate stressed vegetation condition. Small VSWI values indicate a low canopy temperature and high vegetation index, which represent unstressed vegetation. While VSWI is characterized by varying moisture and the thermal conditions of vegetation, it represents overall vegetation health. In the present study, the VSWI that is used for drought monitoring was modified. According to the literature, drought is apparent when the value of the VSWI threshold exceeds 60, and the severity of drought will decline when the values are less than 60. However, this threshold can vary as a function of different factors, including the vegetation cover types, local weather conditions and soil type. Thus, to avoid the use of a threshold that may contain errors, in this study we considered the use of VSWI anomaly percentage (VSWIanom% ). VSWIanom% was calculated in each pixel to assess the changes of the index with regard to mean conditions (VSWIanom% ). The VSWIanom% is used to compare a period of the index (VSWIijk ) with the average for several years (13 years), which reflects the deviation of the VSWI from the normal range: VSWIanom% =

VSWIijk − VSWIijk VSWIijk

× 100

(3)

A positive anomaly percentage indicates different levels of vegetation stress, and a negative anomaly percentage means a favorable condition for vegetation. The second approach relates to the application of a vegetation filter. Thus, to assess drought impacts using the VSWI index over pasture and agricultural lands (Fig. 1b), the data were extracted based on land use information from the Land Cover Land Use map of the study area. This map was derived from a Landsat 7 equipped with Enhanced Thematic Mapper Plus (15 m) and Landsat 5 Thematic Mapper (30 and 60 m) mosaic (Vieira et al., 2013). A total of 162 satellite images were used to prepare the vegetation cover map. The map presents the location and distribution of major vegetation types and non-vegetated land surface formations for the Northeast Brazil Region, which includes the semiarid region. 2.3. Comparative analysis of the spatial and temporal characteristics drought indicators Four representative hydrological years were selected as a basis to compare the MODIS image-derived and meteorologicalmeasured drought indices: 2010–2011 as a wet year and 2011–2012, 2012–2013 and 2013–2014 as dry years at different intensities. Scatter plots, Pearson correlation analyses and a correlation matrix for the VSWI and meteorological data were developed. Average values of VSWI over the 5 km × 5 km square were used for the statistics calculation to match the spatial resolution of the interpolated meteorological data. 3. Results 3.1. NDVI–LST relationships as an indicator of drought NDVI for vegetative cover generally range from 0.1 to 0.8. The values depend on the canopy architecture, density and

A.P.M. Cunha et al. / Agricultural and Forest Meteorology 214–215 (2015) 494–505

499

Fig. 4. Long-term annual averages of NDVI for the three years: (a) 2011, (b) 2012, and (c) 2013. (d)–(f) Long-term averages of LST and (g)–(i) density scatter plots of LST versus NDVI for the same sub-periods. Yellow and red colors represent higher concentration of pixels. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

500

A.P.M. Cunha et al. / Agricultural and Forest Meteorology 214–215 (2015) 494–505

vegetation moisture. High values (between 0.4 and 0.8) are associated with greener and denser vegetative cover, whereas small values (between 0.2 and 0.3) represent low vegetation cover. Actually, NDVI values of semiarid areas, as well as those of the Brazilian semiarid area, present low values (Fig. 4) and have been demonstrated to be strongly dependent on plant water availability in wet months. The semiarid region is characterized by the presence of the natural vegetation that is known as caatinga. Caatinga vegetation is composed of shrubs and small trees that are usually thorny and deciduous and lose their leaves in the early dry season (lower NDVI). During the dry season, the NDVI values in caatinga vegetation are low, while in the rainy months the values are higher due to the response of the vegetation. Previous studies (Santos and Shimabukuro, 1993; Gao et al., 2003; Ferreira et al., 2005) have shown seasonal variability of vegetation in semiarid or caatinga vegetation where NDVI values have shown large annual amplitudes. On the coast outside of the semiarid boundary, high values of NDVI are found in regions that are covered by broadleaf evergreen trees and deciduous forest (not shown). Fig. 4 shows the spatial distribution of the annual average NDVI and LST. Among the three analyzed years, 2011 presented the highest NDVI values and lowest LST values (Fig. 4a, b) and correspond to a wet year. During 2012, the inverse relation was observed for most of the LST/NDVI space, i.e., when the surface temperature increased, the NDVI value decreased (Fig. 4c, d), which represents a higher level of vegetation stress (less water left on the soil for plant transpiration). During dry conditions, rising leaf temperatures are good indicators of plant moisture stress and precede the onset of drought. The same behavior is observed for 2013 (Fig. 4e, f). These results are corroborated by the study performed by Karnieli et al. (2006, 2010), who demonstrated that it is the most common behavior in drought environments (Karnieli et al., 2006; Julien and Sobrino, 2009). Karnieli et al. (2010) showed that during the mid-growing season in the US Great Plains and Southwest, 80% of the cropland areas and 68% of the pasture lands are characterized by strong negative correlations between LST and NDVI. The averages of NDVI and LST for each year (2011–2013) were used to create a density scatter plot. Fig. 4g–i is scatter plots of NDVI and LST data for the entire study area, croplands and grasslands, respectively. In the scatter plots, the yellow and red colors represent higher concentration pixels. Fig. 4 shows a high negative correlation between NDVI and LST. High NDVI values are reached at low LST values, which can be easily explained by the fact that their vegetation is water limited and not temperature limited (Goward et al., 2002; Karnieli et al., 2010). Because of the inclination angle of the Earth, the radiation level throughout the entire study area is high during whole year, and radiation is most likely not a limiting factor for vegetation growth. In the scatter plot for 2011 (Fig. 4g), it can be observed that the higher concentration pixels are between 0.45 and 0.7 (NDVI) and 299.65 K and 310.95 K. The concentration is larger for a smaller range of NDVI values and higher temperatures for 2013 and especially for 2012 (0.25–0.5 for the NDVI and 306.15 K and 315.85 K for temperature). From the scatter space, drier areas were characterized by low NDVI and high temperature (Kogan, 2000; Karnieli et al., 2010). 3.2. Drought assessment using VSWI To calculate the VSWI, the daily MODIS LST was averaged to a 8-day composite using the same day of year for the composite of MODIS NDVI (Terra and Aqua platforms). The time series data of the VSWI were used to analyze the drought occurrence trends. A high value of VSWI means that the canopy temperature is high or that the vegetation density is low, leading to a severe drought

Fig. 5. Long-term average of VSWI within hydrological year (October to September).

(Fig. 5). In the long-term average of VSWI, the highest values are concentrated in the semiarid region, which can be associated with the distribution of the annual precipitation in this region (Fig. 2a). Climatologically, this region is characterized by an average annual rainfall that does not exceed 800 mm. Furthermore, this semiarid region is also predominantly covered with grassland and partly sparse vegetation. These types of vegetation usually show low NDVI values throughout the year. Throughout the long-term average, the VSWI based on the vegetation stress indicates high stress values in regions with sparse vegetation and bare soil, which is found in some parts of the study area (not shown). The bare soil region and some dune areas in the study area never exhibit wet conditions. The sandy soil that is exposed to the direct solar radiation cannot retain water, and therefore no photosynthesis activity is observed. However, it is noteworthy that the VSWI method is applicable for monitoring drought effect in vegetation covering regions. When used to the regions with bare land, the VSWI method is less efficient, since the higher soil background temperature would severely interfere in the assessment (Son et al., 2012). Fig. 6 presents the spatial and temporal pattern of VSWI for four hydrological years over the study area. As can be observed on the VSWI anomalies map for 2010–2011, the majority of the study area was almost drought-free. The lowest VSWI values were found against a long-term average (negative VSWI anomaly percentage), primarily in the northern sector of the region (Fig. 6e). The VSWI anomalies in the semiarid region reached approximately −8%, which indicates a wet condition. This condition agrees with the mean annual rainfall during the hydrological year 2010–2011, with an accumulation of approximately 900 mm in the region. In addition, 21 days of soil moisture deficit during the rainy season were found (Fig. 6i, n). Overall, the spatial distribution of NDD follows the rainfall pattern, and regions with higher NDD are directly associated with those with the lowest cumulative precipitation.

A.P.M. Cunha et al. / Agricultural and Forest Meteorology 214–215 (2015) 494–505

501

Fig. 6. Long-term annual averages of VSWI, normalized VSWI anomalies, number of days of soil moisture deficit and mean annual rainfall for the four hydrological years: 2010–2011, 2011–2012, 2012–2013 and 2013–2014.

Signs of droughts impacts (positive anomalies) appeared locally in some areas in the western region and in the center of the study area. Some of these positive anomaly percentage points (Fig. 6e) that were located on the western part of the study area correspond to agricultural areas. It can be noted that the general situation of ground surface during the hydrological year 2010–2011 was very suitable for agricultural cropping and livestock production in most parts of the semiarid region. Fig. 6 indicates that the study area experienced drought conditions during the hydrological year 2011–2012. According to the VSWI index, approximately 50% of the Northeast Region suffered vegetative drought conditions, especially in the semiarid region. Based on the annual average, 2011–2012 was characterized by

high VSWI values in most part of the study area (Fig. 6b). Positive VSWI anomalies reached 9% in the semiarid region (Fig. 6f). This result is also consistent with the observed cumulative precipitation and NDD fields. Approximately 54 days of soil moisture deficit were observed, with a cumulative precipitation of approximately 500 mm in the region (Fig. 6j, o). Signs of drought intensified during the hydrological year 2012–2013, when the vegetation experienced stress and a loss of health. Note in Fig. 6g that most parts of the study area were subjected to a spatially extensive drought (approximately 65% of the NEB and 85% of the SANEB, Fig. 7), primarily in the central and northern part of the semiarid region. The positive VSWI anomalies reached 15.4%. Approximately 60% of the study area

502

A.P.M. Cunha et al. / Agricultural and Forest Meteorology 214–215 (2015) 494–505

Fig. 7. Percentage of grassland and/or croplands areas affected by droughts according to VSWI index in NEB and SANEB Regions (hydrological years 2004–2014).

consists of extensive livestock farming and agriculture, therefore the occurrence of drought results in reductions in livestock production and/or crop yield. Finally, the impact of the drought became less intense during the hydrological year 2013–2014, especially in the northern part of the semiarid region; it was reduced to 40% of the NEB. The VSWI anomalies reached 7.7% on average (Fig. 6h), which followed the cumulative precipitation of 700 mm and the NDD of 30 in the rainy season (Fig. 6m, q). 3.3. Validation of drought-monitoring products To identify which indicators (precipitation and NDD) were most strongly related to the spatial distributions of VSWI in the different rainy seasons, a Pearson correlation analysis was computed for each pixel in the study area, stratified by sub-period and with VSWI as the dependent variable. The results of the analysis are summarized in Table 1. The analysis was performed only for lands that were covered by pastures and/or crops, and areas of natural vegetation, urban areas and water bodies were excluded from the analysis. Those areas were excluded because their

inclusion could interfere with the meteorological and hydrological conditions and maintain normal vegetation health in spite of poor rainfall or water-stress, which can cause a large response delay. Overall, the correlations between VSWI, precipitation and the NDD time series had different characteristics in the five sub-regions and rainfall regime (Table 1). Note that in the November–February period, the VSWI variance, which is explained by the precipitation and the number of days of soil moisture deficit, is lower than in the other four sub-periods. The highest correlation was found between the precipitation and NDD for the rainy seasons of all three years. This pattern was expected because the precipitation data were used to calculate NDD (direct dependence). In this case, it is necessary to consider that errors in the precipitation data can propagate into the NDD or even errors that are associated with the estimation of the evapotranspiration because the interpolated data were also used to estimate it. With regard to the VSWI index, there is a significant linear correlation with the precipitation during the JFMA wet season of 2013 (0.78), and lowest correlation was observed for the FMAM wet season of 2011 (0.12).

Table 1 Average temporal Pearson correlation coefficient (r2 ), computed between VSWI, precipitation (Prec) and soil moisture deficit (NDD) derived from simple water balance model over grasslands–croplands areas. 2011 VSWI

2012 Prec

DJFM VSWI Prec NDD

1 0.37 0.31

1 0.82

NDJF VSWI Prec NDD

1 0.40 0.37

1 0.82

JFMA VSWI Prec NDD

1 0.35 0.40

1 0.84

FMAM VSWI Prec NDD

1 0.16 0.12

1 0.69

1 −0.25 0.24

1 −0.75

AMJJ VSWI Prec NDD

2013

NDD

VSWI

Prec

1

1 −0.60 0.53

1 −0.86

1

1 −0.30 0.43

1 −0.54

1

1 −0.75 0.68

1 −0.89

1

1 −0.60 0.59

1 −0.88

1

1 −0.59 0.63

1 −0.83

NDD

VSWI

Prec

NDD

1

1 −0.53 0.53

1 −0.89

1

1

1 −0.16 0.25

1 −0.48

1

1

1 −0.78 0.77

1 −0.89

1

1

1 −0.72 0.69

1 −0.89

1

1

1 −0.23 0.13

1 −0.48

1

DJFM, December to March; NDJF, November to February; JFMA, January to April; FMAM, February to May; AMJJ, April to July; NDD, number of days of soil moisture deficit; Prec, accumulated precipitation; VSWI, Vegetation Supply Water Index.

A.P.M. Cunha et al. / Agricultural and Forest Meteorology 214–215 (2015) 494–505

4. Discussions Drought is an ecological phenomenon that manifests itself in the reduction of agricultural production and causes social crises and political problems. Different drought indicators are used to identify droughts. However, the impacts of the choice of a certain indicator for drought characterization (e.g., severity, frequency and duration of droughts) are not fully understood. Generally, the drought measuring indicators are not well correlated to one another. Therefore, it is quite common that when one drought index identifies drought at a particular place, another drought index indicates a normal condition at the same place and time. Overall, the results demonstrate that the empirical LST–NDVI relationship (VSWI) can be effectively exploited as an indicator of spatial temporal characteristics of water stress conditions in the whole study area. Comparing the two years of intense drought (2011–2012 and 2012–2013), it can be noted that the positive VSWI anomalies were more intense in 2012–2013 than in 2011–2012 (Fig. 6f, g). On the other hand, based on the precipitation and NDD indicators, the hydrological year 2011–2012 presented more extreme dry condition characteristics than the hydrological year 2012–2013. For the hydrological year 2012–2013, the cumulative precipitation was approximately 600 mm, whereas the NDD was 42 in the rainy season (Fig. 6l, p). One explanation for these inconsistencies might be the “memory effect” of vegetation. In arid and semiarid lands, rainfall is seasonal, which affects vegetation directly in the same season and indirectly in the immediate next season through the “memory effect” (Martiny et al., 2005; Philippon et al., 2005) and “recovery effect” (the difficulty of vegetation to recover from previous drought conditions, Martiny et al., 2005). The memory effect is defined as the capacity of semiarid ecosystems to benefit from water surpluses, which indicates that there clearly exists a memory of the past precipitation at a time lag of one-year (Schwinning and Sala, 2004; Martiny et al., 2005). Drought can reduce resiliency, rendering plants to be more vulnerable to a recurring disturbance. Therefore the mean VSWI level of one year preceded by drought may also be higher, such as that seen in the hydrological year 2012–2013 that was preceded by a dry hydrological year (2011–2012). Therefore vegetation can be durably affected by a drought if the drought is preceded by another dry year (recovery effect); in addition, rainfall in the immediately previous season also influences the thermal condition of the vegetation. Furthermore, the hydrological year 2010–2011 was a wet year that may have stimulated the photosynthetic activity of the vegetation during the following year; hence the intensity of the drought was less severe in 2011–2012 than in 2012–2013. In other words, vegetation can still develop even with low rainfall if the previous year was wet (memory effect). This “memory effect” was also shown by Philippon et al. (2005) in the Sahel and by Martiny et al. (2005) in three semiarid regions in Africa. In summary, the hydrological year 2011–2012 might have been less impacted by drought because it was preceded by a wetter year, and the hydrological year 2012–2013 might have been more impacted by drought because it was preceded by a drier hydrological year. The impacts of the vegetation stress for the 2011–2014 drought condition were also felt outside of the semiarid boundary. This is an indication that the entire Northeast Brazil region is experiencing severe water stress, including the western extremity of the region, where the climate is equatorial and the precipitation ranges from 1000 to 1200 mm per year (Marengo et al., 2009). In general, the highest correlation between VSWI, NDD and precipitation occurs under dry conditions (there is less soil moisture

503

availability), and a very weak association between the rainfall and vegetation condition occurs under those rainy conditions. Because hydrometeorological data such as precipitation and soil moisture content that are collected by surface observation stations often possess poor spatial resolution, especially in remote regions with difficult access. Thus, it is important to note that the rainfall data were interpolated at a spatial resolution of 5 km, and it is therefore necessary to consider that there may be inconsistencies in the precipitation data if the weather stations have a low spatial density. 5. Conclusions Considering the geographical complexity and irregularity of the climate of Northeast Brazil, the objective of this study was to examine the spatial and temporal characteristics of vegetative drought in the Northeast Region using the remote sensing-based VSWI. The results have demonstrated that: • The empirical LST–NDVI relationship can be effectively exploited as an indicator of spatial temporal characteristics of water stress conditions in the Brazilian semiarid region. VSWI index efficiently identified vegetated areas (crop/pasture) that were affected by the 2012 and 2013 drought. It is highlighted that the use of VSWI anomaly percentage is recommended once the same index values might not denote the same drought severity, and different values might have the same drought severity. VSWI is timedependent and usually region-specific, which reflects different physiological adaptations and partially different ecological factors. Furthermore, VSWI is strongly recommended for use during plant growing seasons (wet seasons). • The results also highlight interannual persistence effects in vegetation dynamics, as shown by the complex mechanisms of the recovery and memory effects. • The study showed that the semiarid region is more affected by drought stress conditions; however, during the 2012–2013 droughts, areas outside of the semiarid boundary also presented drought impacts. This finding indicates that the behavior of drought is very erratic. This has large implications for planners and policy makers who are actively engaged in drought mitigation and preparedness. Due to the different sources of information and principles that are used for drought indices, monitoring results that are obtained from indicators often have certain differences. It can be determined that any single index is not sufficient for precisely depicting drought characteristics. Thus, the combined use of different indicators at the same time or indices that integrate various sources of information may achieve results that are more consistent with the actual situation. References Abbas, S., Nichol, J.E., Qamer, F.M., Xu, J., 2014. Characterisation of drought development through remote sensing: a case study in Central Yunnan, China. Remote Sens. 6 (6), 4998–5018. Allen, R.G., Pereira, L.S., Raes, D., Smith, M., 1998. Crop Evapotranspiration – Guidelines for Computing Crop Water Requirements. FAO – Food and Agriculture Organization of the United Nations, Rome. Berliner, P., Oosterhuis, D.M., Green, G.C., 1984. Evaluation of the infrared thermometer as a crop stress detector. Agric. For. Meteorol. 31, 219–230. Boken, V.K., 2005. Agricultural drought and its monitoring and prediction: some concepts. In: Boken, V.K., Crackenll, A.P., Heathcote, R.L. (Eds.), Monitoring and Predicting Agricultural Drought: A Global study. Oxford University Press, New York. Bhuiyan, C., Singh, R.P., Kogan, F.N., 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, 289–302. Carlson, T.N., Perry, E.M., Schmugge, T.J., 1990. Remote estimation of soil moisture availability and fractional vegetation cover for agricultural fields. Agric. For. Meteorol. 52, 45–69.

504

A.P.M. Cunha et al. / Agricultural and Forest Meteorology 214–215 (2015) 494–505

Carlson, T.N., Gillies, R.R., Perry, E.M., 1994. A method to make use of thermal infrared temperature and NDVI measurements to infer surface soil water content and fractional vegetation cover. Remote Sens. Rev. 9, 161–173. Carolwicz, M., 1996. Nature hazards need not lead to natural disasters. EOS 77 (16), 149–153. Cavalcanti, N.B., Oliveira, C.A.V., Brito, L.T.L., Resende, G.M., 1999. Avaliac¸ão do uso de técnicas de captac¸ão de água de chuva na região semi-árida do Nordeste brasileiro. Rev. Bras. Eng. Agríc. Ambient 3 (3), 403–407. Chen, Y.Y., You, Y.S., Tang, Y.H., 2014. Study and apply on model of monitoring and assessing early and mid-summer drought multi-parameter – a case study in Chongqing. In: International Conference on Water Resource and Environmental Protection: WREP2014. Cunha, A.P.M.A., Alvala, R.C.S., Correia, F.W.S., Kubota, P.Y., 2009. Calibration of the ‘Simplified Simple Biosphere Model – SSiB’ for the Brazilian Northeast. AIP Conf. Proc. 1100, 641. Cunha, A.P.M.A., Alvalá, R.C.S., Sampaio, G., Shimizu, M.H., Costa, M.H., 2013. Calibration and validation of the integrated biosphere simulator (IBIS) for Brazilian semi-arid region. J. Appl. Meteorol. Climatol. 52, 2753–2770. Davenport, M.L., Nicholson, S.E., 1993. On the relation between rainfall and the Normalized Difference Vegetation Index for diverse vegetation types in East Africa. Int. J. Remote Sens. 14, 2369–2389. Doorenbos, J., Pruitt, W.O., 1977. Guidelines for Predicting Crop Water Requirements, FAO-ONU, Rome, Irrigation and Drainage Paper no. 24 (rev.)., pp. 144. Dorman, J.L., Sellers, P.J.A., 1989. Global climatology of albedo, roughness length and stomatal resistance for atmospheric general circulation models as represented by the Simple Biosphere Model (SiB). J. Clim. Appl. Meteorol. 28, 833–855. Du, L., Tian, Q., Yu, T., Meng, Q., Jancso, T., Udvardy, P., Huang, Y., 2013. A comprehensive drought monitoring method integrating MODIS and TRMM data. Int. J. Appl. Earth Obs. Geoinf. 23, 245–253. Feddes, R.A., Kabat, P., Van Bakel, P.J.T., Bronswijk, J.J.B., Halbertsma, J.M., 1988. Modelling soil water dynamics in the unsaturated zone: state of art. J. Hydrol. 100, 69–111. Ferreira, O.E., Beltrão, N.E.M., König, A., 2005. Efeitos da aplicac¸ão de água residuária e nitrogênio sobre o crescimento e produc¸ão do algodão herbáceo. Rev. Bras. Oleagin. Fibro. 9 (01/03), 893–902. Gao, X., Huete, A.R., Didan, K., 2003. Multisensor comparisons and validation of MODIS vegetation indices at the semiarid Jornada Experimental Range. IEEE Trans. Geosci. Remote Sens. 4 (10). Gao, M., Qin, Z., Zhang, H., Lu, L., Zhou, X., Yang, X., 2008. Remote sensing of agro-droughts in Guangdong province of China using MODIS Satellite Data. Sensors 8 (8), 4687–4708. Gibbs, W.J., Maher, J.V., 1967. Rainfall Deciles as Drought Indicators. Bureau of Meteorology Bulletin No. 48. Commonwealth of Australia, Melbourne. Gitelson, A., Kaufman, Y., 1998. MODIS NDVI optimization to fit the AVHRR data series – spectral considerations. Remote Sens. Environ. 66 (3), 343–350. Goward, S.N., Xue, Y., Czajkowski, K.P., 2002. Evaluating land surface moisture conditions from the remotely sensed temperature/vegetation index measurements: an exploration with the simplified simple biosphere model. Remote Sens. Environ. 79, 225–242. Gutman, G., 1990. Towards monitoring droughts from space. J. Clim. 3, 282–295. Henricksen, B.L., Durkin, J.W., 1986. Growing period and drought early warning in Africa using satellite data. Remote J. Sens. 7, 1581–1608. Herrmann, S.M., Anyamba, A., Tucker, C.J., 2005. Recent trends in vegetation dynamics in the African Sahel and their relationship to climate. Glob. Environ. Change 15, 394–404. Huete, A., Didan, K., Miura, T., Rodriguez, E.P., Gao, X., Ferreira, L.G., 2002. Overview of the radiometric and biophysical performance of the MODIS vegetation indices. Remote Sens. Environ. 83, 195–213. IBGE, 2009. O censo agropecuário 2006 e a agricultura familiar no Brasil. MDA, Brasília. Ibrahim, Y.Z., Balzter, H., Kaduk, J., Tucker, C.J., 2015. Land degradation assessment using residual trend analysis of GIMMS NDVI3g soil moisture and rainfall in sub-Saharan West Africa from 1982 to 2012. Remote Sens. 7, 5471–5494. Johnson, G.E., Achutuni, V.R., Thiruvengadachari, S., Kogan, F.N., 1993. The role of NOAA satellite data in drought early warning and monitoring: selected case studies. In: Wilhite, D.A. (Ed.), Drought Assessment, Management and Planning: Theory and Case Studies. Kluwer Academic Publishers. Julien, Y., Sobrino, J.A., 2009. Global land surface phenology trends from GIMMS database. Int. J. Remote Sens. 30, 3495–3513. Justice, C., Townshend, J., Vermote, E., Masuoka, E., Wolfe, R., Saleous, N., Roy, D., Morisette, J., 2002. An overview of MODIS, its data processing and products for terrestrial science applications. Remote Sens. Environ. 83 (1–2), 3–15. Karnieli, A., Bayasgalan, M., Bayarjargal, Y., Agam, N., Khudulmur, S., Tucker, C.J., 2006. Comments on the use of the Vegetation Health Index over Mongolia. Int. J. Remote Sens. 27, 2017–2024. Karnieli, A., Agam, N., Pinker, R.T., Anderson, M., Imhoff, M.L., Gutman, G.G., Panov, N., Goldberg, A., 2010. Use of NDVI and land surface temperature for drought assessment: merits and limitations. J. Clim. 23, 618–633. Keyantash, J., Dracup, J.A., 2002. The quantification of drought: an evaluation of drought indices. Bull. Am. Meteorol. Soc. 83, 1167–1180. Keyantash, J., Dracup, J.A., 2004. An aggregate drought index: assessing drought severity based on fluctuations in the hydrologic cycle and surface water storage. Water Resour. Res. 40, W09304, http://dx.doi.org/10.1029/ 2003WR002610.

Kogan, F.N., 1990. Remote sensing of weather impacts on vegetation in non-homogeneous areas. Int. J. Remote Sens. 11 (8), 1405–1419. Kogan, F.N., 1997. Global drought watch from space. Bull. Am. Meteorol. Soc. 78, 621–636. Kogan, F.N., 2000. Contribution of remote sensing to drought early warning. In: Wilhite, D.A., Sivakumar, M.V.K., Wood, D.A. (Eds.), Early Warning Systems for Drought Preparedness and Drought Management. Proc. Expert Group Meeting, 5–7 September 2000, Lisbon, Portugal. World Meteorological Organization, Geneva, pp. 86–100. Kogan, F.N., 2002. World droughts in the new millennium from AVHRR-based Vegetation Health Indices. Eos Trans. Am. Geophys. Union 83 (48), 562–563. Liu, J., Wu, J., Wu, Z., Liu, M., 2013. Response of NDVI dynamics to precipitation in the Beijing–Tianjin sandstorm source region. Int. J. Remote Sens. 34, 5331–5350. Marengo, J.A., Schaeffer, R., Pinto, H.S., Zee, D.M.W., 2009. Mudanc¸as climáticas e eventos extremos no Brasil. FBDS, Rio de Janeiro. Martiny, N., Richard, Y., Camberlin, P., 2005. Interannual persistence effects in vegetation dynamics of semi-arid Africa. Geophys. Res. Lett. 32, L24403, http:// dx.doi.org/10.1029/2005GL024634. McKee, T.B., Doesken, N.J., Kleist, J.,1993. The relation of drought frequency and duration to time scales. In: Proceedings of the Eighth Conference on Applied Climatology. American Meteorological Society, Boston, pp. 179–184. Matheron, G., 1969. Le krigeage universel. Technical Report 1. Paris School of Mines. Cah. Cent. Morphol. Math., Fontainebleau. Mu, Q., Zhao, M., Kimball, J.S., McDowell, N.G., Running, S.W., 2013. A remotely sensed global terrestrial drought severity index. Bull. Am. Meteorol. Soc. 94, 83–98. Nagarajan, R., 2009. Drought Assessment. Springer, Dordrecht, Netherlands. Nichol, J.E., Abbas, S., 2015. Integration of remote sensing datasets for local scale assessment and prediction of drought. Sci. Total Environ. 505, 503–507. Palmer, W.C., 1965. Meteorological Drought. Research Paper No. 45. U.S. Department of Commerce Weather Bureau, Washington, DC. Palmer, W.C., 1968. Keeping track of crop moisture conditions, nationwide: the new Crop Moisture Index. Weatherwise 21, 156–161. Peters, A.J., Walter-Shea, E.A., Jii, L., Vina, A., Hayes, M., Svoboda, M.D., 2002. Drought monitoring with NDVI-based standardized vegetation index. Photogramm. Eng. Remote Sens. 68 (1), 71–75. Philippon, N., Mougin, E., Jarlan, L., Frison, P.-L., 2005. Analysis of the linkages between rainfall and land surface conditions in the West African monsoon trough CMAP, ERS-WSC, and NOAA-AVHRR data. J. Geophys. Res. 110, D24115, http://dx.doi.org/10.1029/2005JD006394. Rhee, J., Im, J., Carbone, G.J., 2010. Monitoring agricultural drought for arid and humid regions using multi-sensor remote sensing data. Remote Sens. Environ. 114, 2875–2887. Rossato, L., Tomasella, J., Alvalá, R.C.S., 2005. Avaliac¸ão da Umidade do Solo no ˜ (1982/83). Rev. Bras. Agromet. 13 (1), Brasil durante o Episódio El Nino 143–153. Rulinda, C.M., Dilo, A., Bijker, W., Stein, A., 2012. Characterising and quantifying vegetative drought in East Africa using fuzzy modelling and NDVI data. J. Arid Environ. 78, 169–178. Sandholt, I., Rasmussen, K., Andersen, J., 2002. A simple interpretation of the surface temperature/vegetation index space for assessment of surface moisture status. Remote Sens. Environ. 79, 213–224. Santos, J.R., Shimabukuro, Y.E., 1993. O sensoriamento remoto como indicador das fenofases dos cerrados brasileiros: estudo de caso com dados AVHRR/NOAA. Simpósio Brasileiro de Sensoriamento Remoto, 8, 1993, Curitiba. Anais, 2. MCT/INPE, São José dos Campos, pp. 249–257. Sapir, D.G., Below, R., 2014. EM-DAT: International Disaster Database. Université Catholique de Louvain, Brussels, Belgium www.emdat.be. Schwinning, S., Sala, O.E., 2004. Hierarchy of responses to resource pulses in arid and semi-arid ecosystems. Oecologia 141, 211–220. Shafer, B.A., Dezman, L.E., 1982. Development of a Surface Water Supply Index (SWSI) to assess the severity of drought conditions in snowpack runoff areas. In: Proceedings of the Western Snow Conference, Fort Collins, CO, pp. 164–175. Son, N.T., Chen, C.F., Chen, C.R., Chang, L.Y., Minh, V.Q., 2012. Monitoring agricultural drought in the Lower Mekong Basin using MODIS NDVI and land surface temperature data. Int. J. Appl. Earth Obs. Geoinf. 18, 417–427. Souza, S.S., Tomasella, J., Gracia, M.G., Menezes, P.C.P., Amorin, M.C., Pinto, C.A.M., 2001. O Programa de Monitoramento Climático em Tempo Real na área de atuac¸ão da SUDENE: PROCLIMA. Bol. Soc. Bras. Meteor. 25, 15–24. Tannehill, I.R., 1947. Drought, Its Causes and Effects. Princeton University Press, Princeton, NJ, USA. Thiruvengadachari, S., Gopalkrishna, H.R., 1993. An integrated PC environment for assessment of drought. Int. J. Remote Sens. 14, 3201–3208. Thenkabail, P.S., Gamage, N., Smakhin, V., 2004. The Use of Remote Sensing Data for Drought Assessment and Monitoring in South West Asia. IWMI Research Report # 85. IWMI, Colombo, Sri Lanka, 25 pp. Tomasella, J., Hodnett, M.G., 1998. Estimating soil water retention characteristics from limited data in Brazilian Amazonia. Soil Sci. 163, 190–202. Tomasella, J., Hodnett, M.G., Rossato, L., 2000. Pedotransfer functions for the estimation of soil water retention in Brazilian soils. Soil Sci. Soc. Am. J. 64, 327–338. Tomasella, J., Pachepsky, Y., Crestana, S., Rawls, W.J., 2003. Comparison of two techniques to develop pedotransfer functions for water retention. Soil Sci. Soc. Am. J. 67, 1085.

A.P.M. Cunha et al. / Agricultural and Forest Meteorology 214–215 (2015) 494–505 Tomasella, J., Hodnett, M.G., Cuartas, L.A., Nobre, A.D., Waterloo, M.J., Oliveira, S.M., 2008. The water balance of an Amazonian micro-catchment: the effect of interannual variability of rainfall on hydrological behavior. Hydrol. Proc. 22, 2133–2147. Tucker, C.J., Choudhury, B.J., 1987. Satellite remote sensing of drought conditions. Remote Sens. Environ. 23, 243–251. Tucker, C.J., 1989. Comparing SMMR and AVHRR data for drought monitoring. Int. J. Remote Sens. 10, 1663–1672. Udmale, P., Ichikawa, Y., Manandhar, S., Ishidaira, H., Kiem, A., 2014. Farmers × perception of drought impacts, local adaptation and administrative mitigation measures in Maharashtra State, India. Int. J. Disaster Risk Sci. 10, 250–269. Vermote, E.F., Saleous, N., Justice, C., 2002. Atmospheric correction of the MODIS data in the visible to middle infrared: first results. Remote Sens. Environ. 83 (1–2), 97–111. Vieira, R.M.S.P., Cunha, A.P.M.A., Alvalá, R.C.S., Carvalho, V.C., Ferraz Neto, S., Sestini, M.F., 2013. Land use and land cover map of a semiarid region of Brazil for meteorological and climatic models. Rev. Bras. Meteorol. 28, 129–138. Wan, Z., Zhang, Y., Zhang, Q., Li, Z., 2004. Quality assessment and validation of the MODIS global land surface temperature. Int. J. Remote Sens. 25, 261–274. Wilhite, D.A., Glantz, M.H., 1985. Understanding the drought phenomenon: the role of definitions. Water Int. 10 (3), 111–120. Wilhite, D.A., 2000. Drought as a natural hazard: concepts and definitions. In: Wilhite, D.A. (Ed.), Drought: A Global Assessment, vols. 1 and 2. Routledge Publishers, London (chapter 1). Wu, J., Zhou, L., Mo, X., Zhou, H., Zhang, J., Jia, R., 2015. Drought monitoring and analysis in China based on the Integrated Surface Drought Index (ISDI). Int. J. Appl. Earth Obs. Geoinf. 41, 23–33. Xue, Y., Sellers, P.J., Kinter, J.L., Shukla, J., 1991. A simplified biosphere model for global climate studies. J. Climate 4, 345–364. Yagci, A.L., Deng, M., Di, L., Han, W., Peng, C., 2011. Vegetation index based technique for global agricultural drought monitoring. In: 2011 5th

505

International Conference on Recent Advances in Space Technologies (RAST), 9–11 June, pp. 137–141, http://dx.doi.org/10.1109/RAST.2011.5966808. Yang, W., Merchant, J.W., 1997. Impacts of upscaling techniques on land cover representation in Nebraska, USA. Geocarto Int. 12 (1), 27–39. Yang, W., (MSc thesis) 2010. Drought Analysis Under Climate Change by Application of Drought Indices and Copulas. Civil and Environmental Engineering, Portland State University. Yang, S., Xu, Y., Gao, Y., 2013. Summer drought assessment based on Vegetation Supply Water Index in Chongqing. Adv. Intell. Syst. Res., http://dx.doi.org/10. 2991/rsete.2013.231. Zhou, G.Y., Wei, X.H., Wu, Y.P., Liu, S., Zhang, D., Zhang, Q., Liu, J., Meng, Z., Wang, C., Chu, G., Liu, S., Tang, X., Liu, X., 2011. Quantifying the hydrological responses to climate change in an intact forested small watershed in Southern China. Glob. Change Biol. 17, 3736–3746. Zhou, G.Y., Peng, C.H., Li, Y.L., Liu, S., Zhang, Q., Tang, X., Liu, J., Yan, J., Zhang, D., Chu, G., 2013a. A climate change-induced threat to the ecological resilience of a subtropical monsoon evergreen broadleaved forest in south China. Glob. Change Biol. 19, 1197–1210. Zhou, G.Y., Houlton, B.Z., Wang, W., Huang, W., Xiao, Y., Zhang, Q., Liu, S., Cao, M., Wang, X., Wang, S., Zhang, Y., Yan, J., Liu, J., Tang, X., Zhang, D., 2014. Substantial reorganization of China’s tropical and subtropical forests: based on the permanent plots. Glob. Change Biol. 20, 240–250, http://dx.doi.org/10. 1111/gcb.12385. Zhou, L., Zhang, J., Wu, J., Zhao, L., Liu, M., Lü, A., Wu, Z., 2012. Comparison of remote sensed and meteorological data derived drought indices in Mid-Eastern China. Int. J. Remote Sens. 33, 1755–1779. Zhou, L., Wu, J., Zhang, J., Leng, S., Liu, M., Zhao, L., Zhang, F., Shi, Y., 2013b. The integrated surface drought index (ISDI) as an indicator for agricultural drought monitoring: theory, validation, and application in Mid-Eastern China. IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens. 6, 1254–1262.