A spatiotemporal analysis of droughts and the influence of North Atlantic Oscillation in the Iberian Peninsula based on MODIS imagery
- 498 Downloads
Drought is among the least understood natural hazards and requires particular notice in the context of climate change. While the Mediterranean climate is by itself prone to droughts, a rise of temperatures and alteration of rainfall patterns already render the southern parts of continental Portugal and Spain highly susceptible to desertification. Precipitation in the Iberian Peninsula is mainly controlled by the large-scale mode of North Atlantic Oscillation (NAO) and is distributed with elevated variability over the cold months. Most drought studies of this region rely on meteorological data or apply information on vegetation dynamics, such as the Normalised Differenced Vegetation Index (NDVI), to indirectly investigate droughts. This paper evaluates the influence of the NAO winter index on the spatiotemporal occurrence of droughts in the Iberian Peninsula during the spring and summer seasons (March to August) for the years 2001–2005, 2007 and 2010. We applied the Vegetation Temperature Condition Index (VTCI) to identify local droughts. VTCI is a remote sensing drought index developed for reflecting soil moisture conditions in agricultural areas and combines information on land surface temperature (LST) and NDVI. As such, VTCI overcomes the shortcomings of NDVI in terms of drought monitoring. We derived biweekly information on LST and NDVI from MODIS/Terra and produced VTCI–NAO correlation maps at a confidence level of at least 90 % based on the VTCI time series. The results reflect a typical Mediterranean pattern in most parts of Iberia that is highly influenced by relief. Spring seasons are marked by great variability of precipitation, while summers persistently become dry, particularly in the south. NAO exerts its greatest influence in April and June, clearly delineating high correlation areas in the northwest and southeast with reverse patterns between the spring and early summer months. Due to the impact on water availability, the spring months are important for plant growth. At the same time, agricultural lands were found with types of land cover less resilient to droughts. The knowledge acquired in studies like the one reported here is therefore likely to be used in drought warning models for agriculture in spring.
Drought is a natural disaster that can exert serious harm on the society, the environment and the economy (Ji and Peters 2003; Santos et al. 2010). About half of the terrestrial surfaces on earth are located in areas prone to drought, and more dramatically, almost all major agricultural lands extend in such regions (USDA 1994). Of all environmental disasters within the twentieth century, droughts entailed the greatest detrimental impact (Mishra and Singh 2010). Although mainly driven by precipitation deficits, drought is sensitive to preconditions such as land cover/use, type of soil, elevation and human-induced activities, when an increased demand of water supply from agriculture, industry and tourism may result in water scarcity (Pereira and Paulo 2004; Paredes et al. 2006). Hence, drought lacks a universally accepted definition. The most general characterisation is perhaps the one that considers the phenomenon as a recurring natural disaster caused by a temporary deficit in water supply. Drought may reduce soil moisture, streams and groundwater resources, deteriorate water quality, damage vegetation cover, degrade ecosystems and agricultural lands and also affect human beings (Mishra and Singh 2010; Rojas et al. 2011). This leads to the identification of five types of drought: meteorological, hydrological, agricultural, ecological and socioeconomic.
The remote sensing drought index applied in this paper corresponds to soil moisture conditions and indirectly reflects precipitation patterns. As such, it describes the hydrological or agricultural type of water scarcity (Choi et al. 2013). Nevertheless, we consider drought a natural hazard driven by atmospheric factors from synoptic to large-scale modes. The predictability of drought among atmospheric disasters is lowest due to its sensitivity to a number of environmental influences (Mishra and Singh 2010). The Iberian Peninsula regularly suffers water deficits that cause serious damage particularly to non-irrigated agriculture during drought years, degrading the environment and enhancing soil erosion (Vicente-Serrano 2006; Costa and Soares 2012). Since climate change and the rise of surface temperature have been recognised as a major threat of the twenty-first century, the drought issue belongs to one of the most explosive environmental topics for the research community, the public worldwide as well as in Portugal and Spain (Mishra and Singh 2010; Costa and Soares 2012).
The Iberian climate is strongly influenced by the atmospheric circulation pattern of the North Atlantic Oscillation (NAO). The NAO index compares the occurrence of high barometric pressure zones around the Azores and low barometric pressure systems in the North Atlantic (Iceland) and is expressed as the difference of normalised sea level pressure between both regions. The NAO determines the climatic variability in the North Atlantic throughout the year where it controls the direction and intensity of storm tracking paths, which is decisive for the precipitation regime in Iberia, because low pressure systems from the Atlantic account for the major source of winter rainfall (Hurrell 1995; Goodess and Jones 2002; Trigo et al. 2002, 2004; Vicente-Serrano and Trigo 2011). The impact of NAO on vegetation dynamics and phenology in Europe and the Mediterranean Basin has been studied several times (e.g. Trigo et al. 2002; Gouveia and Trigo 2011). Vicente-Serrano and Heredia-Laclaustra (2004) detected a positive trend of vegetation production given as the sum of Normalised Differenced Vegetation Index (NDVI) from 1982–2000 in the north and a negative trend in the south of Iberia, where at the same time, the NAO influence was found higher (Rodríguez-Puebla et al. 1998; Gouveia et al. 2008). This observation also coincides with the findings of Martín-Vide and Fernández (2001), who describes a major influence of NAO on precipitation in the southwest of Iberia during the winter months, while the other regions rely on other teleconnection patterns such as the Polar Pattern or the Scandinavian Pattern (Rodríguez-Puebla et al. 1998). The influence of NAO also extends to other parts of Europe. In their correlation analysis between NAO and vegetation greenness (NDVI), Gouveia et al. (2008) demonstrated that in Iberia, water accessibility from NAO-induced precipitation highly impacts vegetation growth in spring, while in northeast Europe, the influence of NAO on winter temperature is more decisive for a rapid start into the vegetation period. NAO is therefore the major large-scale mode that controls winter precipitation and impacts NDVI trends particularly in the northwest, south and southwest of Iberia.
Traditional drought monitoring methods rely on meteorological or hydrological data that are collected at single sampling stations (Santos et al. 2009; Caccamo et al. 2011). The majority of drought studies for the western Mediterranean applies such information (e.g. Trigo et al. 2002; Vicente-Serrano 2006; Costa and Soares 2009, 2012; Santos et al. 2010; Gouveia and Trigo 2011; Martins et al. 2012). The area-wide estimation of punctual accessible drought values is accomplished by spatial interpolation, by geostatistical interpolation, for example, and is susceptible to uncertainties especially over climatically and topographically complex terrain (Rhee et al. 2010; Caccamo et al. 2011). Remote sensing data that cover area-wide terrestrial surfaces over an extended period have therefore acquired interest in drought assessment (Martín-Vide and Fernández 2001; Vicente-Serrano and Heredia-Laclaustra 2004; Bayarjargal et al. 2006). In areas with a low density of sample stations, they may be the only source of available information (Caccamo et al. 2011).
Among remote sensing indices, the NDVI is the most widely used for drought analysis. NDVI was introduced by Rouse et al. (1974) to catch the difference between red and near-infrared solar radiation emitted by green plants and reflects the condition of vegetation and its dynamics over the year (Vicente-Serrano and Heredia-Laclaustra 2004). NDVI is correlated to precipitation with a time gap of 1–2 months (Wang et al. 2001) or 3 months as reported by Ji and Peters (2003) depending on the kind of vegetation and soil type. The latter authors argued that NDVI is a good indicator of vegetation moisture conditions and can be used as indirect information for drought monitoring if related to a long-term NDVI series. NDVI indeed became part of numerous remote sensing drought indices, such as the Vegetation Condition Index (VCI; Kogan 1990, 1995), the Normalised Difference Water Index (NDWI; Gao 1996), the Vegetation Temperature Index (VTI; Kogan 1997), the Vegetation Health Index (VHI; Kogan 2000) and the Temperature Vegetation Dryness Index (TVDI; Sandholt et al. 2002) to mention some of them and was successfully used for identifying droughts on regional or local scales (Peters et al. 1993; Nicholson et al. 1998; Gonzalez-Alonso et al. 2000; Liu and Negron-Juarez 2001; Salinas-Zavala et al. 2002; Ji and Peters 2003). However, NDVI has shortcomings that may cause problems when solely used for drought monitoring. The mentioned deficits are its lagged time response to cumulative rainfall (Davenport and Nicholson 1993; Wang et al. 2001; Gouveia et al. 2008), its sensitivity towards environmental preconditions, such as the type of soil, climate and vegetation (Singh et al. 2003; Vicente-Serrano 2007), the fact that the NDVI signal saturates before the full biomass is reached (Carlson et al. 1990; Carlson and Ripley 1997) and that the relationships among vegetation parameters are often non-linear (Choudhury et al. 1994; Vicente-Serrano and Heredia-Laclaustra 2004). In accordance to this discussion, a number of authors suggested the use of land surface temperature (LST) as a second variable for a more complex analysis as the drought related one (Kogan 1995; Park et al. 2004; Wan et al. 2004; Ghulam et al. 2007; Rhee et al. 2010; Wu et al. 2012; Lazzarini et al. 2013). LST is a parameter of the energy state on the earth’s surface (Wan et al. 2004) and serves as an indicator for evapotranspiration, soil moisture and vegetation water stress (Karnieli et al. 2010). Practically, satellite-derived LST measures the temperature of the ground’s skin over bare soil surfaces, of the vegetation canopy surface over densely vegetated areas and a mixture of all over sparsely vegetated terrain (Parida et al. 2008). LST can grow rapidly with increasing water stress due to changes in vegetation condition, vegetation cover and soil moisture. The physical fact that leaf surface temperature of green plants increases during dry conditions explains why LST raises slightly before vegetation cover decreases (Wan et al. 2004) and justifies its use for drought monitoring. LST was used, together with NDVI, in some of above stated drought indices. The LST–NDVI relationship was investigated several times in terms of vegetation cover, burnt areas, soil moisture, drought and climate change (Dall’Olmo and Karnieli 2002; Sandholt et al. 2002; Karnieli and Dall’Olmo 2003; Wan et al. 2004; Jang et al. 2006; Julien et al. 2006; Stisen et al. 2007). A strong negative correlation between LST and NDVI was revealed, and in the scatter plot, the LST–NDVI slope was found to become steeper under dry conditions (Nemani et al. 1993; Prihodko and Goward 1997; Goward et al. 2002). Karnieli et al. (2010) found that in mid and low latitudes of North America, solar radiation is the principal factor driving the LST–NDVI relationship, which affirms LST as a useful variable for drought monitoring.
The Vegetation Temperature Condition Index (VTCI) has been proposed for examining agricultural droughts (Wang et al. 2001; Wan et al. 2004). The VTCI approach is based on the theory that LST–NDVI space creates a triangle (Moran et al. 1994; Gillies et al. 1997; Wan et al. 2004). VTCI combines the variables in a more complex way than the ratio LST–NDVI, which better adapts to the needs of investigating such a complex phenomenon as drought in large regions with many land cover/use characteristics and high climatic variability. Moreover, it overcomes the disadvantages of NDVI, which faces a lagged vegetation response to droughts and makes it less suitable for real-time drought monitoring. Wang et al. (2001) evidenced that on a regional scale (144.1 × 309.1 km2), the VTCI reflects the simulated soil moisture content. Wan et al. (2004) showed in their study for the Great Plains (1001 × 853 km2) that VTCI is related to not only recent rainfall but also cumulative rainfalls and enhances real-time drought monitoring.
The study at hand intends to complement the drought monitoring initiatives on the Iberian Peninsula (Pereira and Paulo 2004; Vicente-Serrano and Heredia-Laclaustra 2004; Vicente-Serrano 2006, 2007; Santos et al. 2007, 2010; Costa and Soares 2009, 2012; Gouveia et al. 2009; Gouveia and Trigo 2011; Costa et al. 2012; Martins et al. 2012; Paolo et al. 2012). We chose the VTCI for spatial and temporal identification of droughts, which so far was never applied for a drought analysis in this region. It directly retrieves soil moisture conditions while taking advantage of compound information on NDVI and LST. Also, until now, the VTCI was only applied for examining droughts in agricultural-dominated areas, but accommodating land classes ranging from arid to alpine environments the Iberian Peninsula comprises far more than low vegetation cover. We concentrated on the recent decade and chose the years 2001 to 2005. Because of the distinct positive (negative) NAO winter index in 2007 (2010), we added the mentioned years to the time series. Two VTCI maps per month were produced at time intervals of 16 days from March to August for all years. Hence, a detailed seasonal analysis was preferred over a longer inter-annual assessment. An object of further interest was the influence of NAO winter index on spatiotemporal drought occurrences in the Iberian Peninsula during spring and summer. The study aspires to the development of a local drought warning system for agriculture based on NAO.
2 Study area
Human activity over a long period (fire, pasturing, agriculture, forest plantations etc.) determines the manifestation of land cover in the Iberian Peninsula today. Natural vegetation types are scarce. Nevertheless, the climatic diversity, relief, a low population density compared to European context and a loss of agricultural land in the last three decades all engender high environmental diversity (Lobo et al. 1997). The vegetation is characterised by an Atlantic region (deciduous vegetation) in the north and northwest as well as a Mediterranean region (sclerophyllous vegetation) in the south, east and centre of the Iberian Peninsula. Orographic plant communities complement the vegetation in areas of higher elevation.
3 Data and preprocessing
3.1 Satellite products
NAO winter index (Gibraltar and Reykjavik) for the years 2001–2005, 2007 and 2010 (average values from Dec, Jan, Feb and Mar)
NAO winter index
3.2 Digital elevation model and land cover map
A digital elevation model was acquired at the Earth Science Data Interface of the Global Land Cover Facility from http://glcfapp.glcf.umd.edu:8080/esdi/esdi_index.jsp. It was downloaded as a SRTM30 GTOPO30 mosaic. This is a near-global digital elevation model with a resolution of 30 arc seconds. For the land cover analysis, we used the CORINE land cover map 2006 (CLC2006) from the European Environmental Agency (EEA 2013). We reclassified the map and obtained 14 of originally 44 classes pursuant to the third level description. All maps were transformed to the prevalent projected coordinate system in Spain to UTM30 N with datum WGS84.
The NAO is defined as the normalised pressure difference between a station on the Azores (Ponta Delgada) and one in Iceland (Reykjavik). An extended version using a station in the southwest of the Iberian Peninsula proved to be useful for the winter half of the year (Hurrell 1995). The NAO winter index is computed as the mean NAO value for the months of December, January, February and March (Osborn et al. 1999). The Climatic Research Unit (CRU) at the University of East Anglia is a major institution for NAO research and stores the index values from 1821 to 2000 (http://www.cru.uea.ac.uk/cru/data/nao/). The most recent updates are gathered on the website of Tim Osborn, Reader at CRU. We obtained the Jones et al. (1997) NAO winter index that considers the stations at Gibraltar and Reykjavik (http://www.cru.uea.ac.uk/timo/datapages/naoi.htm). The winter index from 2006/07 was distinctly positive, and the winter index from 2009/10 is one of the greatest negative values within the 190-year record (Table 1).
4.1 VTCI definition
4.2 Calculation of cold and warm edge
The determination of the cold (Eq. 2) and warm (Eq. 3) edge is crucial for calculating the VTCI. VTCI is a time-dependent and region-specific drought index that performs better during the plant growing season (Wan et al. 2004). We followed the strategy of Patel et al. (2011) and computed the coefficients for each observation date separately. For the computation, we produced a table indicating the LST–NDVI pair for each pixel. According to Patel et al. (2011), we sorted the columns by NDVI intervals (0.01) and extracted the maximum and minimum LST for each interval. Then, we improved their approach by also extracting the exact NDVI value for each maximum and minimum LST. We chose the maximum and minimum LST with the corresponding (exact) NDVI value to calculate the cold and warm edge function via the Ordinary Least Squares (OLS) method. For obtaining a better coefficient of determination (R2), we manually excluded extreme values and tails. Negative NDVI values were generally neglected. The coefficients a and a′ are the intercepts, and the coefficients b and b′ are the slope of the warm and cold edge, respectively. The average R2 for all observation dates of the warm edge is 0.82, which implies that the variability of LST for dry pixels is well explained by NDVI. The slope of the cold edge approximates a horizontal line, whereby R2 settles around 0. On the cold edge, the intercept offers more significant information than the slope. Figure 2 shows the linear regression lines for Jul 1 2004 including the equations (Eq. 2 and Eq. 3) and R2.
The procedure described above was used to compute each of the 83 VTCI maps considered in this study. Every pixel in the maps presents a proper VTCI value.
4.3 VTCI interpretation
VTCI ranges between 0 and 1 with low values indicating severe vegetation stress and a value of 1 indicating no vegetation stress. In the literature, a VTCI threshold of 0.4 or 0.45 is used to indicate the start of vegetation stress due to drought (Wang et al. 2001; Sun et al. 2008; Patel et al. 2011). We considered pixels below a VTCI of 0.4 as dry pixels. Unlike previous studies, our study area accommodates a variety of habitats exhibiting soil moisture conditions from arid to wet; thus, we established equal drought categories utilising the entire VTCI spectrum. The four drought categories are very dry (VTCI 0–0.2), dry (VTCI 0.2–0.4), moderate (VTCI 0.4–0.6) and wet (VTCI 0.6–1). An equal interval classification was considered justified because the literature does not yet offer a definition of VTCI levels.
4.4 Correlation analysis
We computed the Pearson’s correlation coefficient between NAO and VTCI pixel-wise using the different NAO winter indices of the observation years and the VTCI maps of one observation date during the entire study period. The point correlation analysis was repeated for all observation dates (Mar 1, Mar 2,…, Aug 2). In this manner, we produced 12 correlation maps of VTCI versus NAO for every observation date, where every pixel was assigned a Pearson’s correlation coefficient r. According to Gouveia et al. (2008), pixels with a statistically significant (linear) correlation (significance levels of 10 % and less) will be named North Atlantic High Correlation pixels (NHCP).
4.5 Map production
All monthly or seasonal maps were obtained as an average from single observation maps. If pixels of one date were not available due to cloudy conditions, we omitted those pixels and considered only the values of the other VTCI map(s). If a pixel was empty in all source maps, it also remained blank in the average map.
5 Results and discussion
5.1 VTCI time series
Land mass and relief play an important role in Iberian local VTCI manifestations. Major elevation systems also keep favourable soil moisture content during the warm season even in the traditionally dry parts in the south and east of the Iberian Peninsula, while sheltered areas within the main circulation direction consequently exhibit (very) dry conditions together with some wet phases in spring. The cloud blocking effect of mountains causes enhanced precipitation amounts and responds with a higher VTCI, whereas a rapid change from a high VTCI to a minor value may occur within a short distance. A sudden alternation from dry to moderate soil conditions is exemplified inland of the central Portuguese coast in March and April (Fig. 3), where the elevated areas of the Central System abruptly enter the coastal plain on a north–south orientated line and catch moisture content superior to the flat areas. The major basins (Figs. 1 and 3) constitute the most important lands for agricultural production in the Iberian Peninsula (Moratiel et al. 2011), but invariably extend in the lee of mountains or are located in climatically dry parts of the territory (Guadalquivir Basin, Tagus Basin). The Duero Basin in the northwest of the Iberian Peninsula represents the only plain exhibiting moderate conditions for vegetation in spring with some variability throughout the rest of the year. On the other hand, the blocking situation of the Pyrenees, Cantabrian Range and Iberian System turns the Ebro Basin in the northeast of Spain into Europe’s most northern semi-arid region (Vicente-Serrano 2007) that demonstrates great instability between dry and wet phases in spring. In the time series, the beginning of May is often identified as the moistest period in the Ebro Basin and expresses the accumulated precipitation amounts from March to April that affect VTCI with a lagged time response in the last spring month (Vicente-Serrano 2007).
Drought categories show the distribution of dry and non-dry land at a higher grade of detail than average VTCI values. The course of drought categories over the study period on a monthly scale permits the detection of differences within the months (Fig. 4, lower panel). The Mediterranean climatic cycle is clearly identifiable by the increase of very dry areas during the summer months and a simultaneous reduction of moderate soil moisture conditions. Together with wet phases in early spring (2003, 2004) or late spring (2002, 2005, 2010), the dry category dominates surfaces of moderate soil moisture throughout the year and ranges between 45 and 60 %. The moderate category varies between 30 and 40 % with peaks of around 50 % in spring. In such dry years as 2001 and 2007, the gap between dry and moderate surfaces increases considerably, while the dry year 2005 was specifically marked by a low soil moisture level in March and the highest percentage (17.5 %) of very dry areas in August. The wet category is normally greatest in mid-spring, thus reflecting humid spring situations. Wet conditions in summer are seen particularly in the north and northwest of the Iberian Peninsula where sufficient rainfall supports enhanced vegetation activity (Fig. 3).
5.2 NAO versus VTCI
5.3 Role of NAO on drought
The correlation findings relate to local VTCI in the time series. April, in fact, was relative dry in the centre-northwest (Duero Basin) during the NAO+ years (as VTCI decreased) and relatively humid in the NAO− years (as VTCI increased). The year 2010 with a distinct negative NAO winter index registered the driest March in the northwest but became significantly humid in April reflecting the alteration in correlation pattern. March is also very likely to become dry on the Balearic Islands during NAO− (observed 2001, 2010). The year 2007, a strong NAO+ year, registered the wettest situation in the southeast, which coincides with the high positive correlation in this region. June and July recorded specifically wet conditions in the centre-northwest and dry situations in the southeast and east during NAO+ (2002, 2003 and 2007), while 2010 (strong NAO− index) registered a dry situation in the centre-northwest at the end of June and July and a specifically wet August in the east and northeast.
The correlation of NAO versus a vegetation index (NDVI) produces different results compared to the correlation of NAO towards the drought index VTCI. Gouveia et al. (2008) found a persistent negative correlation between NAO and NDVI implying low (high) vegetation activity in spring and summer when NAO winter index was positive (negative). According to the authors, the NAO influence was the most notable in April, when in NAO+ years, vegetation activity was reduced compared to NAO− years. NAO versus VTCI correlation results in a clearly dipolar situation between April and June/July. In those 2 months, the northwest and southeast of the Iberian Peninsula exhibit an evidently opponent correlation one to each other, whereas the type of correlation reverses in the course of the seasons. Both indices, NDVI and VTCI, indentify April as drier (moister) during NAO+ (NAO−) years (Fig. 7). Vicente-Serrano and Heredia-Laclaustra (2004) describe a high correlation between NAO and the sum of NDVI for the study years 1982–2002 in the southwest and a non-significant correlation in the north. Contrarily, the southwest exhibits a less significant correlation between NAO and VTCI, whereas, as previously mentioned, the northwest is strongly correlated. Due to drought impact, the authors describe the southwest as an area with stable or slightly decreasing vegetation activity. The persistent occurrence of dry situations even during spring could be a reason for the weak correlation between VTCI and NAO in this region. In the previously mentioned studies, the authors investigated different study periods (1982–2002 and 1982–2000, respectively) based on NDVI, which could explain the different results. At this point, we also wish to stress that those studies examined the impact of climatic variables on large-scale vegetation dynamics, in order to predict vegetation trends. We applied the VTCI to identify local drought patterns and to estimate the impact of NAO on the occurrence of locally severe soil moisture conditions and therefore also used data of higher spatial resolution.
The high spatial and temporal variability of wet periods in spring prevents showing a clear correlation pattern especially in May and March. During these months, one area can be wet in one year but dry in the next. The prevailing Mediterranean climate with dry summers causes a clearer NAO influence pattern during the warm season, as the lack of precipitation during winter is decisive for a rapid increase in the accumulated precipitation deficit (Santos et al. 2007). The importance of April and other spring months is shown by the high impact on vegetation activity. The prominent role of water availability as a limiting factor for vegetation growth during spring is replaced by other determinants, such as low temperature and frosts in the remaining seasons. Droughts that occur in autumn and winter are therefore less influential than in spring and summer (Vicente-Serrano 2007). We believe that the extracted knowledge particularly for areas of high correlation between VTCI and NAO in spring could be useful for an early warning or forecasting system of drought in agricultural management.
5.4 Land cover
We analysed the distribution of pixels for 12 different land cover types pursuant to our four drought categories (not shown). Agricultural lands (non-irrigated agricultural land, permanent crops, heterogeneous agricultural areas and permanently irrigated agricultural land) account for the land cover types most vulnerable to drought, which is indicated by a dominating dry category (50–80 %), a weak moderate category (10–45 %) and a negligible wet category throughout the year as well as a gaining very dry category in summer (up to 18 % for permanent crops and 16 % for non-irrigated agricultural lands). A missing vegetation layer makes those agricultural lands specifically sensitive to drought during growing period in early spring when the root system is still poorly developed and soil easily runs dry. Unfavourable conditions in spring restrain cereals and root crops from healthy plant development and thus are liable to diminished harvests (Xoconostle-Cazares et al. 2010).
Of all land uses, forests constitute the land cover most resilient to drought. In the inter-annual profile of all forest types, the moderate class consistently has greater share than the dry one. Mixed forests appear most robust, but mainly extend in the humid areas of the Iberian Peninsula. Coniferous forests, on the other hand, exhibit favourable condition even in drier zones. Other than the native oak species, pine trees are able to cope well with the environmental conditions of a Mediterranean climate. Only broad-leaved forests become partly dry during summer.
For low vegetation types such as natural grasslands, sclerophyllous vegetation and sparsely vegetated areas (around 15 % of the Iberian Peninsula’s surface), we detected a great fluctuation of soil moisture conditions within two subsequent observation dates in the VTCI time series, especially in spring (Fig. S1–S14). Low vegetation types react more sensitively to drought than high vegetation and are often located in areas with the highest aridity (Vicente-Serrano et al. 2006). These ecotypes compound species that developed strategies to outlast regular dry periods and are able to quickly expand when water is available because they promptly respond to spatiotemporal changes in soil moisture (Le Houerou 1996; Bonifacio et al. 1993; Sannier and Taylor 1998; Vicente-Serrano 2007), which explains the fluctuation in the VTCI maps. Because it is a near real-time drought index, VTCI proves advantageous for the spatiotemporal monitoring of permanently altering drought conditions.
The presented results are in line with findings of other authors, who argued that the effects of drought on vegetation are diverse and depend on the land cover type, the vegetation characteristics (mean NDVI), the month, the timeframe of the episode and the climatic conditions (Vicente-Serrano 2007; Gouveia et al. 2009). The frequent occurrence of drought enhances fire frequencies, soil erosion, degrades vegetation and is harmful for agriculture and economy (Wilhite et al. 2007; Costa and Soares 2012). Iberia’s land cover types are to a great extent human induced. Wide areas of the landscape are dominated by agricultural lands and forms of low vegetation that often evolved as a degradation of forests. It is not unexpected that agricultural areas (specifically non-irrigated agricultural lands) constitute the land classes most afflicted by droughts. They were artificially created in areas separate from the environmental conditions that would support the growth of the respective vegetation and do not endure periods of drought. Bennie and Hensley (2001) reported that farms in regions with an average annual rainfall of less than 600 mm need to have a larger surface area than those in humid regions to achieve an equal output of crop harvest. In terms of economical value, agricultural lands constitute meaningful land cover types. Thus, rising temperatures, decreasing amounts of rainfall and the intensification of agriculture in many parts of Spain during the second half of the last century will force the promotion of efficiency in irrigation methods and a change in crop models (Ceballos et al. 2004; Martín-Rosales et al. 2007; Lasanta and Vicente-Serrano 2012). Calculating the impacts of drought on agriculture is therefore crucial for determining consequences in water management, particularly in the context of climate change.
5.5 NAO and land cover
On a local scale, as represented by the Iberian Peninsula, the NAO–VTCI correlation pattern contributes important knowledge about the spatiotemporal variability of soil moisture conditions, especially in spring. Considering that NAO, as the major atmospheric circulation mode, controls rainfall events in the Western Mediterranean, this knowledge can be used to develop local drought warning tools for water management and agriculture based on a certain NAO winter index. At this point, we want to stress that such atmospheric circulation patterns as the NAO is not the only factor that acts on drought. The nature and quality of soil substrate, the intensive use of agricultural land and application of irrigation are human-induced influences that have an impact on the availability of water for plants (Gouveia et al. 2008).
This study investigates the occurrence of spatiotemporal drought patterns in the Iberian Peninsula within the last decade using Vegetation Temperature Condition Index (VTCI). The role of North Atlantic Oscillation on local drought occurrences was subject to further study. Despite its link to vegetation activity (NDVI), the VTCI was specifically developed to locate areas that provoke vegetation stress associated with drought. The VTCI results reflect a Mediterranean annual cycle in major parts of the terrain and an Atlantic-influenced section in the north and northwest. Relief highly impacts the establishment of dry areas, whereas the conditions may vary within a short distance. We also made evident that the application of VTCI enables the spatiotemporal detection of drought as a highly variable phenomenon and facilitates the identification of soil moisture conditions at a small (local) scale.
VTCI constitutes a less-probed drought index. For a region-wide use not limited to agricultural areas, an additional analysis of the calculation process, namely the establishment of the cold and warm edge that highly influence the index, would sustain its robustness.
Understanding the NAO influence on drought occurrences is particularly interesting for the spring season when vegetation activity is high and dependent on sufficient rainfall. The clear correlation patterns in March, April and June coincide with observations in the VTCI time series and are useful for understanding the spatiotemporal variability of droughts in relation to the large-scale NAO mode. We gave the example of the Duero Basin, where April is very likely to be dry (wet) in NAO+ (NAO−) as VTCI decreases (increases). The correlation findings may trigger the development of an advance warning system for soil moisture conditions, especially for spring months, which would be of great interest for drought management in agriculture. Therefore, further analysis should be applied to improve the knowledge of the relationship between NAO and VTCI. Although the computation of the VTCI is time consuming, the coverage of several decades could deliver important spatiotemporal information on soil moisture conditions for local analysis. We also suggest testing a NAO winter index that only considers December, January and February as an index by the beginning of March is already of interest for agricultural endeavour.
- AEMET (2012) Website of the Spanish Meteorological Agency. www.aemet.es. Accessed February 2013
- EEA (2013) European Environmental Agency—CORINE Land Cover map 2006. http://www.eea.europa.eu/data-and-maps/data/corine-land-cover-2006-clc2006-100-m-version-12-2009. Accessed March 2013
- EUROSTAT (2013) Website of Eurostat, a Directorate-General of the European Commission about European Statistics. http://epp.eurostat.ec.europa.eu/portal/page/portal/eurostat/home/. Accessed March 2013
- Gillies RR, Carlson TN, Cui J, Kustas WP, Humes KS (1997) A verification of the ‘triangle’ method for obtaining surface soil water content and energy fluxes from remote measurement of the Normalized Difference Vegetation Index (NDVI) and surface radiant temperature. Int J Remote Sens 18:3145–3166CrossRefGoogle Scholar
- Gonzalez-Alonso F, Calle A, Casanova JL, Vazquez A, Cuevas JM (2000) Operational monitoring of drought in Spain using NOAA–AVHRR satellite images. Proceedings of 28th International Symposium on Remote Sensing of Environment 27–31 March, Cape Town, South AfricaGoogle Scholar
- Gouveia C, Trigo RM (2011) The Impacts of the NAO on the Vegetation Activity in Iberia. In: Vicente-Serrano SM, Trigo RM (eds) Hydrological, socioeconomic and ecological impacts of the North Atlantic Oscillation in the Mediterranean Basin. Springer Verlag, Berlin-Heidelberg-New York, pp 113–128. doi:10.1007/978-94-007-1372-7 CrossRefGoogle Scholar
- Kogan FN (2000) Contribution of remote sensing to drought early warning. In: Wilhite DA, Sivakumar MVK, Wood DA (eds) Early warning systems for drought preparedness and drought management. Proc. Expert Group Meeting, 5–7 September, Lisbon (Portugal). World Meteorological Organization, Geneve, pp 86–100Google Scholar
- Martín-Vide J, Fernández D (2001) El ídice NAO y la precipitación mensual en la España peninsular. Invest Geogr 26:41–58Google Scholar
- Pereira LS, Paulo AA (2004) Droughts: concepts indices and prediction. In: Hamdy A, Trisorio-Liuzzi G (eds) Options méditerranéennes, series B, studies and research, water management for draught mitigation in the Mediterranean, vol 471. International Centre for Advanced Mediterranean Agronomic Studies, Bari, pp 13–144Google Scholar
- Rodríguez-Puebla C, Nieto S (2010) Trends of precipitation over the Iberian Peninsula and the North Atlantic oscillation under climate change conditions. Int J Climatol 30(12):1807–1815Google Scholar
- Rouse JW, Hass RH, Schell JA, Deering DW (1974) Monitoring vegetation systems in the Great Plains with ERTS. In: Freden SC, Mercanti EP, Becker MA (eds) Proceedings of the 3rd Earth Resources Technology Satellite-1 symposium, Volume 1: Technical Presentations, p 309−317. Washington DC. NASA SP-351Google Scholar
- Solano R, Didan K, Jacobson A, Huete A (2010) MODIS Vegetation Index Users’ Guide (MOD13 Series). Collection 5, Vegetation Index and Phenology Lab, University of Arizona (http://vip.arizona.edu) 1 – 42
- USDA (1994) Major world crop areas and climatic profiles. World Agricultural Outlook Board. US Dep Agric Agric Handb 664:157–170Google Scholar
- Vicente-Serrano SM, Trigo RM (2011) Introduction. In: Vicente-Serrano SM, Trigo RM (eds) Hydrological. Socioeconomic and Ecological Impacts of the North Atlantic Oscillation in the Mediterranean Region. Springer-Verlag, Berlin-Heidelberg-New York, pp 1–8. doi:10.1007/978-94-007-1372-7 Google Scholar
- Wan Z (2007) MODIS Land Surface Temperature Products Users’ Guide, Collection 5. ICESS. University of California, Santa Barbara, pp 1–30Google Scholar
- Wang PX, Li X, Gong JY, Song C (2001) Vegetation temperature condition index and its application for drought monitoring. IEEE 2001 International Geoscience And Remote Sensing Symposium (IGARSS), 9-13 July 2001. The University of New South Wales, SydneyGoogle Scholar