Keywords

5.1 Introduction

Regarding the WMO (Wilhite 2006), droughts relate to water deficiency in a particular region, and their severity has potential repercussions in diverse contexts (Serda 2013). Recent studies indicate that droughts’ frequency and severity appear to be increasing in some zones due to climate variability and change (Podestá et al. 2016). Also, its occurrence is cyclical and is related to the El Niño Southern Oscillation ENSO (van der Zee Arias et al. 2012), which influences both the irregularity of the precipitation cycle and the spatiotemporal magnitude of impact at scales. Regional and local, limiting the natural resilience capacity of the territory.

Relate to Diaz et al. 2020a, spatiotemporal methodologies applied to droughts have been developed. These have managed to generate a better understanding of spatiotemporal dynamics, strengthening their monitoring and impact reduction (Cai 2014). Several approaches have been proposed to describe the spatiotemporal development of drought. According to Bacanli et al. 2009; and Dracup et al. 1980, propose methodologies that apply to force a significant number of meteorological variables in the drought indexes calculation based on time series disaggregation of relating it to the spatiality of all zone, establishing a unique phenomenon in the zone.

Among the tools developed in recent decades, remote sensing has shown great promise for improving the calculation and monitoring of drought events. For this, vegetation indices and data on drought, humidity, and surface temperature derived from satellite images are used by Sahaar and Niemann (2020); Ghulam et al. 2007). Following (Tadesse et al. 2012), reported significant correlations between the normalized difference vegetation index (NDVI), obtained from MODIS images, calculated every 16 days, and drought. Tadesse et al. (2005) showed that when using NDVI, despite being an effective indicator of humidity and vegetation conditions, there is a lag between the occurrence of the drought event and the change in its values, which indicates that this index is not appropriate to monitor the drought conditions of crops, in real-time conditions (Nihoul 2005)

5.2 Case Study

Central America is one of the world’s regions most exposed to the risks of natural hazards and climate variability making climate hazards the most prevalent, represented mainly by frequent droughts and severe spontaneous floods, affecting agricultural production, with greater intensity in degraded areas (Eckstein et al. 2017).

Belize, Guatemala, Honduras, Panama, El Salvador, constitutes the dry forest ecosystem corridor of Central America (Fig. 5.1). Caused by the geographic location, Central America have more probability to receive strong heat waves, also, influence of global climatic phenomena and hurricane seasons prevent a progressive recovery of its hydric condition in terms of food and environmental security (van der Zee Arias et al. 2012).

Fig. 5.1
A map of Central America depicts the dry corridor states.

Study area, Central America Dry Corridor

Full size image

The hydrometeorological records shows that Central America had experienced an average temperature change, closing of 0.05 °C yearly, it means 0.5–1 °C of historical anomalies by temperature (Fig. 5.2) in last 40 years. Otherwise, rainfall rates have tended to decrease in 12 mm/year during the same period (Fig. 5.3).

Fig. 5.2
A map depicts the directions of trajectory areas based on the V C I dataset in Central America with arrows.

Surface temperature anomalies and total precipitation by 1950–2020

Full size image
Fig. 5.3
A line graph depicts the total precipitation anomaly from 1950 to 2020. The line decrease and increase with a sharp rise. It was at its peak in 1950 and 2010 and at its trough in 1958 and 1994.

Surface temperature anomalies and total precipitation by 1950–2020

Full size image

According to the Intergovernmental Panel on Climate Change (IPCC), Central America is categorized as the Latin America region, with most likely to be influenced by climate change impacts and effects (Frieler et al. 2017; Anandhi et al. 2008). Which is supported by increased flooding and prolonged droughts among other adverse natural hazards (hurricanes, earthquakes).

Water supply is linked to intra-annual precipitation fluctuations as well as geographical variability. During El Niño events, annual rainfall can decrease between 38 and 42%, with extended warm periods and more extended dry spells.

High temperature seasons and the dry climate conditions have catastrophic consequences on agricultural development yield and affecting food security (Bae et al. 1981; FAO 2019). Added to the tropical storm season has damaging effects in terms of critical water levels, pollutions preventing the water system near resilience (FAO 2019).

The hazards generated by flooding or droughts mainly, have increased in recent years. Growing food insecurity and water supply availability, increasing climatic trends, and socioeconomic stresses in the region have caused the displacement of people from their homes and communities in 1.6 millions in the last decade (Buttafuoco and Caloiero 2014).

As an example, Fig. 5.4 and 5.5 allows to understand the ecological, socio-ecological, economic, and cultural impacts of land use change on Central America, trends in land use and subsoil use change have been identified, allowing us to establish few grid cells changes mainly in forest and wetlands to woody savannas about 20% reduction of forest structures to pasture (Anderson et al. 2019; Maldonado et al. 2016). Reinforced by agricultural expansion in the region around 46% (van der Zee Arias et al. 2012). This to establish potential land cover that may be affected by drought and may cause multi-country unsustainability.

Fig. 5.4
3 maps with different color shading depict the land cover migration in 2001, 2010, and 2020. A shade key is provided at the bottom of the maps.

Land-cover migration and trends over 2000 to 2020

Full size image
Fig. 5.5
A bar graph with 11 different colored bars depicts the land cover changes from 2000 to 2020. A color key is located below the graph in one line.

Land-cover migration and trends over 2000 to 2020

Full size image

5.3 Methodology

5.3.1 Data Acquisition

Past drought events required the use of comprehensive and detailed satellite data to assess and monitor past drought events. Both hydrometeorological variables and earth observation data were used (Sexton et al. 2013). As shown in Table 5.1. Where data sets of the best available temporal and spatial resolution are used.

Table 5.1 Climatological and remote sensing data source description
Full size table

5.3.2 Drought Calculation

Drought calculation needs and indicator that represents a quantitative drought condition according with the magnitude of specific variable. Drought indicator can be used to define drought (dry condition) with any variable like meteorological drought: based on precipitation condition or agricultural drought: based on water balance [P-ETP] (Van Loon 2015).

However, drought indices are not universally accepted to characterize the drought conditions (Okal et al. 2020) because depends of the driver factor and regional context. In this way, the rainfall and vegetation health are significant factors that could show the influence of drought frequency or severity for instance.

5.3.3 Drought Vegetation Monitoring Indexes

The analysis of surface drought conditions (vegetation response, soil moisture) based on remote sensing information involves transforming the data or bands and applying a standardized transformation as a vegetation index. MODIS datasets were used to estimate vegetation indices for the period 2000–2020. Among these indices, the most common and sensitive to vegetative dynamics are NDVI and VCI.

5.3.4 Normalized Difference Vegetation Index NDVI

The normalized difference vegetation index NDVI, is an index that measure of the phenological vegetation condition. It was computed from Terra platform with 250 m spatial resolution from MODIS server using the relation of Infrared (NIR) and Red band (RED) as its shown in (5.1).

$$ {\text{NDVI}} = \frac{{{\text{Band}}_{{{\text{NIR}}}} - {\text{Band}}_{{{\text{RED}}}} }}{{{\text{Band}}_{{{\text{NIR}}}} + {\text{Band}}_{{{\text{RED}}}} }} $$
(5.1)

This product was corrected using GIS techniques to correct geometrical overlapping and radiometric fluctuations and cloudy pixels were removed (Kogan 1997). Then was categorized to represent drought condition scale as (Rokhmatullah et al. 2018), that represents whether photosynthesis characteristics is above or below of average condition (Table 5.2).

Table 5.2 NDVI drought category based on (Rokhmatullah et al. 2018)
Full size table

5.3.5 Vegetation Condition Index VCI

Vegetation condition index VCI, is calculated from NDVI product using (5.2). VCI is a pixel-based index by considering of mean and maximum multi-annual variability (Kogan 1997).

$$ {\text{VCI}} = \frac{{{\text{VCI}}_{ijk} - {\text{VCI}}_{{i,\,{\text{min}}}} }}{{{\text{VCI}}_{{i,\,{\text{max}}}} - {\text{VCI}}_{{i,\,{\text{min}}}} }} \times 100 $$
(5.2)

where VCIijk is the VCI value for the pixel i during month j for year k, VCIijk is the monthly, VCI value for pixel i in month j for year k. VCI, VCIi, min and VCIi, max are the multiyear minimum and maximum of VCI, respectively, for pixel i (Kogan 1997).

Then was categorized to represent drought condition scale regarding with the vegetation growing season as 30% below, as its shown below on Table 5.3 (Rokhmatullah et al. 2018).

Table 5.3 VCI drought category based on (Rokhmatullah et al. 2018)
Full size table

5.3.6 Climatological Drought Indexes

Drought indices are associated numerically with the value of hydroclimatic variables that may reflect changes in their hydrological regime or quantity, such as meteorological, agricultural, and hydrological variables, representing factors that contribute to or counteract the occurrence and propagation of droughts (Habibi et al. 2020). These methodologies, mainly involve a simplification process, in which an anomaly or a normalization of the values of these driving factors is considered to influence the magnitude of other physical variables highly related to water deficits, the most used of which generalize a standardization of the driving source of water such as precipitation (SPI) and water balance (SPEI) (Depsky and Pons 2020; Belayneh et al. 2014).

5.3.6.1 Standardized Precipitation Index SPI

The SPI computation is based on the long-term precipitation records cumulated over a selected time scale. According with Husak, 2007. This methodology allows to fit this kind of data using gamma probabilistic distribution (5.3) and then standardized by normal distribution to obtain deviations from each precipitation record (Blain 2011).

$$ g\left( x \right) = \frac{1}{{\beta^{\alpha } \Gamma \left( \alpha \right)}}x^{\alpha - 1} e^{ - x/\beta } $$
(5.3)

where, α, β are scale and shape parameters that could be obtained using L-moments method. According to the SPI, drought starts when the SPI value is equal or less than −1.0 and ends when the value becomes zero or high (Table 5.4).

Table 5.4 SPI drought category based on (Rokhmatullah et al. 2018)
Full size table

5.3.6.2 Standardized Precipitation Evapotranspiration Index SPEI

Standardized Precipitation Evapotranspiration Index SPEI is calculated as water balance stablishing the difference between precipitation and potential evapotranspiration PET for the desired month as (5.4), then its normalized by log-logistic probabilistic distribution (5.5) to define the drought indicator (Diaz et al. 2020b), describing the water balance impacts according to the drought severity. The Thornwaite approach is widely applied to compute the PET (Bae et al. 1981).

$$ {\text{WB}}i = {\text{P}}i - {\text{PET}}i $$
(5.4)
$$ f\left( x \right) = \frac{\beta }{\alpha }\left( {\frac{x - \gamma }{\alpha }} \right)^{\beta - 1} \left[ {1 - \left( {\frac{x - \gamma }{\alpha }} \right)^{\beta } } \right]^{ - 2} $$
(5.5)

where, (α, β ,γ) are scale, shape and origin parameters that could be obtained using L-moments method. Then as SPI, drought scale is the same (Table 5.5).

Table 5.5 SPEI drought category based on (Rokhmatullah et al. 2018)
Full size table

5.3.7 Spatiotemporal Monitoring

The spatiotemporal approach is based on the methodology proposed by Corzo and Vitali, 2018. This methodology allows to perform an articulated analysis between the spatial and temporal changes of the observed drought events within the region, after the extraction of the drought extensions (areas), and its subsequent estimation of its location and temporal marks that allow evaluating its propagation and mobility in space and time, which allows addressing the problem related to its dynamics. (Diaz et al. 2020b). However, this methodology has only been applied under the use of hydroclimatic drought indices such SPI or SPEI, so the proposed methodology proposes three stages, initially correlative analysis for the estimation of related accumulation periods (SPI-n, SPEI-n, VI-n) followed by the analysis of non-contiguous drought areas (NCDA). Finally, applying continuous drought areas (CDA) approach and tracking.

5.3.7.1 Correlation Analysis

To implement the spatiotemporal tracking methodology, it is necessary to identify the relationship between drought areas, magnitude and frequency, which are measurable between climatological and vegetation-based indices. This will allow generating a better understanding of the applicability of satellite indices in spatiotemporal drought monitoring (Jiao et al. 2016). For this purpose, Pearson’s correlation test was used, between the monthly time series of SPI-n, SPEI-n, (where n, represents the lag-time variable) and Vegetation Indices (VI) such as VCI, and NDVI for the period 2000–2020. Considering the time series from January to December (annual cycle), and the growing season (April to September), thus establishing the periods with the highest degree of autocorrelation in different periods of accumulation and significance (Kogan 1997).

5.3.7.2 Non-Contiguous Drought Area Analysis (NCDA)

After the drought indices computation and their spatial representation verification. Threshold analysis is defined. This threshold allows to define a boundary at which the drought will be defined for its integrated analysis (in space and time) for time series and spatial grid.

That means, the threshold values below the water normal condition −1.0, was established to extract the drought true zones in the period 1981–2020 for SPI-n and SPEI-n and below 0.25 and 30% for NDVI and VCI, respectively, as shown in Fig. 5.6:

Fig. 5.6
3 maps of Central America labeled 1, 2, and 3. 1 depicts the drought index, 2 represents threshold of drought categories and 3 denotes the drought areas.

Drought index thresholding and binarizing

Full size image

where, I indicates normal values of the drought index, II, establishes thresholder the drought categories and III represents the drought areas in a cluster.

5.3.7.3 Contiguous Drought Area Analysis (CDA)

Contiguous Drought Area, is based on spatial clustering of drought adjacent cells. These drought cells are identified at each time step (Corzo 2019). If the drought indicator is lower or equal than the threshold, it is binarized between 0 and 1, where a value of 1 indicates that the cell is in drought. Otherwise, 0 is used, indicating no drought (Diaz et al. 2020b).

Image analysis approach was done using a Graphos Technique. Connected Labeling Component technique that consists of generating subgroups of patterns in an image (clusters) based on its cell value to connect it (Fig. 5.7) (Herrera-Estrada et al. 2017).

Fig. 5.7
2 tables with 10 columns and 8 rows depict the extraction process of C D A units. One is initially image, where some cells are darkened, and other is clustering, where the dark cells are given a number.

Extracting CDA units using connected components technique

Full size image

5.3.7.4 Spatiotemporal Propagation–Tracking Events

Once the contiguous areas are estimated, the largest cluster is extracted at each time step and geospatial properties such as centroid coordinates (x, y), relative area/perimeter are calculated (Diaz et al. 2020b). Then, for each largest event over time, the centroid obtained is analyzed by calculation of overlapping percentage and Euclidean distance to define the trajectory for each largest drought in time (Diaz et al. 2020a) (Fig. 5.8).

Fig. 5.8
3 rectangular panels with 1 or 2 circles at the center are arranged parallelly. Lower rectangle has 1 circle, other 2 have two circles in each. These circles are connected by 6 lines.

Outlining of tracking approach based on (Blain 2011)

Full size image

where t represents each time step (2000–2020), Ci represents the clusters, and the black spots CDA units.

5.4 Results and Discussion

5.4.1 Climatological Drought Index

Based on ERA5 data, the standardized SPI-n and SPEI-n indices were calculated for accumulations of (n = 1, 3, 6 months) as shown in Fig. 5.9, shows the drought and non-drought values (red and blue, respectively), and their attenuation when considered as oscillations in the more attenuated time series, thus allowing to identify temporally how an event develops as a function of its magnitude in response to the water deficit.

Fig. 5.9
8 graphs in 2 columns depict S P E I and S P I for 1, 3, 6, 9, and 12 months lagged time series. Left column has S P E I and right column has S P I. Each graph depicts a frequent rise and fall.

SPEI (left), SPI (right) for (1, 3, 6, 9, 12 months lagged) time series

Full size image

As its shown in the Fig. 5.9, values for drought index SPI and SPEI below −1.0, reflects a moderate to extreme drought pattern, increasing in frequency over the last decade, with an average of 45–70 periods of drought-associated water deficits (Van Loon 2015; Sutanto et al. 2020; Hydrologic remote sensing: capacity building for sustainability and resilience 2007).

Temporal characteristics of interannual changes may reflect the characteristics of the long-term change in drought. SPI and SPEI variations, in other words, share similar patterns at various time scales, but the lag in the accumulation (lag-time 1, 3, 6) allows the temporal attenuation of the drought magnitude to be appreciated (lower peaks that represents extreme drought magnitude), causing its trend to increase in relative area terms (Entekhabi et al. 2010; Zhang et al. Mar. 2022; Ebrahimi et al. 2010).

On the other hand, the SPI and SPEI monthly variation characteristics clearly reflect the change in the degree of dryness and humidity in each month, especially in the last 20 years (Jiao et al. 2016; Poornima and Pushpalatha 2019; Al-Shujairy et al. 2019). However, when increasing the periods of analysis and looking in contrast at the same accumulations (SPI-6, SPEI-6), drought was more frequent in the months of March to June and October to January (Fig. 5.10), generating extensions in terms of area and deficit more frequent in meteorological drought (SPI) and more prolonged for agricultural drought (SPEI).

Fig. 5.10
Two graphs with various color shades depict comparison between S P E I and S P I for 6 months lagged. Here x axis plots year and y axis plots month.

Comparison SPEI (left -a-), SPI (right -b-) for 6 months lagged (x-axis years and y-axis month)

Full size image

5.4.2 Drought Vegetation Monitoring Indexes

NDVI trends in the vegetative cycle is analyzed, considering the phenological growth as illustrated in Fig. 5.11, which allows inferring the vegetative activity as a function of temperature and water stress cycles (Colliander et al. 2017; Hafni et al. 2022). For this purpose, the monthly NDVI series corresponding to the maximum mean NDVI was selected.

Fig. 5.11
A line graph with month on the x axis and N D V I on the y axis depicts the phenological growth. Initially, the line is straight until mid April, then it begins to rise and after mid June.

Phenological growing season scheme, based on (Peters et al. 2002)

Full size image

For the dry corridor of Central America, a bimodal phenological pattern is observed for the periods January–May and July–October, which are directly related to rainfall patterns (Fig. 5.12). In this way, the vegetation onset depends on the meteorological conditions in the area by year, but not necessarily depends on the vegetation type or crop productivity. (Jong et al. 2011).

Fig. 5.12
A line graph with month on the x axis and monthly median N D V I on the y axis depicts the N D V I monthly trends from 2000 to 2020. Line for each year forms a web in the graph.

NDVI monthly trends

Full size image

However, the derivation of the VCI required a NDVI trend normalization, which is implemented by the LOWESS smoothing technique (locally weighted running line smoother) proposed by (Kogan 1997; Cai et al. 2017).

$$ x = f\left( x \right) + \varepsilon_{1} x_{1} + \varepsilon_{2} x_{3} + \varepsilon_{n} x_{n} $$
(5.6)
$$ w\left( x \right) = \{ \left( {1 - \left| x \right|^{3} } \right)^{3 } :\left| x \right|\left\langle {1 0 :\left| x \right|} \right\rangle 1 $$
(5.7)

Which, if NDVI trends (5.6) with noise ε should be adjust to a normal distribution, then each data value (x) is replaced by a combination of adjacent values of the normal series in a window w(x), using a second order polynomial least squares fit (Kogan 1997). Generating a new time series normalized, (Hilda 2017; Prasetyo et al. 2019) allowing a reconstruction of the time series in which the leaf area and photosynthetic activity is easily appreciable (Murakami et al. 2016; Soudani et al. 2012; Gong 2022). This methodology is tested with 3 filters or smoothing window, in which the degree of smoothing is estimated by least squares minimization, thus allowing to reduce by attenuations sudden changes in NDVI as its shown in Fig. 5.13 (Cai et al. 2017).

Fig. 5.13
A line graph and 3 scatterplots depict the N D V I comparison of applying LOESS smoothing.

NDVI comparison applying LOWESS smoothing

Full size image

5.4.3 Spatiotemporal Approach

After obtaining the drought indexes based on hydro climatological data like SPI, SPEI, and remote sensing-based like NDVI and VCI, a correlative analysis was carried out to establish those time accumulations that were related to each other to implement the spatiotemporal analysis. This correlation was generated from the binarization of each index given the threshold set out in the methodology (NCDA).

However, each drought scale proposed, in terms of magnitude varying, their spatial relationship may not fit because each index takes into account different aggregate factors, e.g., VCI standardizes NDVI in phenological response and SPI 3 standardizes precipitation as a function of three-monthly accumulation, which involves highly variable significances (−0.42 to 0.86) as shown in Figs. 5.14 and 5.15.

Fig. 5.14
A line graph depicts the N C D A correlation between S P I and V C I from 2000 to 2020. Here, two lines with frequent rise and fall overlapping.

NCDA approach correlation for SPI (−1 to inf) and VCI (0.2−1)

Full size image
Fig. 5.15
A line graph depicts the N C D A correlation between S P E I 3 and V C I from 2000 to 2020. Here, two lines depict frequent rise and fall and there is a gap between the lines.

NCDA approach correlation for SPEI 3 (−1 to inf) and VCI (0.2- inf)

Full size image

Therefore, equivalent drought ranges had to be established for their analysis, which implied for each category, moderate, severe, and extreme drought, to find those optimal ranges that had greater significance and representativeness in terms of the area under drought condition (NCDA).

Thus, a relationship range is found between the percentages of the area in representative drought and their spatial autocorrelation (Ghulam et al. 2007; Jiao et al. 2016; Peters et al. 2002). As shown in Table 5.6, the net correlation between climatic and remotely sensed indices varies between (0.39 and 0.89), being the combination SPI-3 (moderate drought)– VCI (extreme) and SPEI 6 (extreme)– VCI (extreme) the most significant one. These results allow validating a priori the viability of NDVI and VCI as indices for spatiotemporal drought monitoring.

Table 5.6 Drought indices general correlation
Full size table

Following the maximum significance criterion (Pearson coefficient > 0.6), the non-contiguous drought area percentages (NCDA) are obtained, as shown in Fig. 5.16, which comparatively compares the NCDA results for the SPEI-6-VCI combination, where there is a ratio of 2.4:1 in terms of the magnitude of the area. Locating in a temporal way, months between October to January and May to August with greater areas of drought and annually related to El Niño Southern Oscillation periods in the last two decades, like 2019, 2015, 2014, 2012, 2004, and 2003 (Sheffield et al. 2009).

Fig. 5.16
38 bar graphs depict N C D A percentage of area for S P E I and V C I.

NCDA percentage of area for SPEI 6 (−1 to inf) and VCI (0.2–1)

Full size image

Binary grids are then generated, and the connected component algorithm is used to find the largest clusters or contiguous drought units (CDA) that are most representative for subsequent monitoring as its shown in Fig. 5.17.

Fig. 5.17
12 maps of Central America depict the V C I index of C D A clustering from January to December.

CDA clustering for VCI index

Full size image

5.4.4 Drought Tracking

Considering, these grids that represent the largest clusters (Fig. 5.18), the characteristics of centroid location and area are extracted to evaluate whether an event t is related to an event t + 1. (Fig. 5.19) where shows the most representative centroids related to drought area extensions. However, as it is a dynamic phenomenon, the drought that can be visualized in a hydro climatological index is not so perceptible based on vegetation, as the phenological growth cycle only indicates for a certain crop or cover which orientation will have the water deficit. Overlap metrics were employed, excluding events that do not exceed at least 30%.

Fig. 5.18
A table with 9 columns and 7 rows C D A cluster characteristics dataset.

CDA clúster characteristics dataset

Full size image
Fig. 5.19
A map depicts the main centers of the highest drought areas in Central America with different sized and colored dots. In the upper right corner is an index key.

Main centroids of highest drought events

Full size image

As shown below in Figs. 5.20 and 5.21, the main trajectories reported according to SPI-SPEI and VCI are presented.

Fig. 5.20
A map depicts the main trajectory areas based on the S P I dash S P E I dataset in Central America with different sized and colored dots. The dots are connected by arrows.

Main trajectories based on spatiotemporal approach driven with SPI-SPEI dataset

Full size image
Fig. 5.21
A map depicts the directions of trajectory areas based on the V C I dataset in Central America with arrows.

Main trajectories directions based on spatiotemporal approach driven with VCI dataset

Full size image

For SPI and SPEI, estimated trajectories emphasize that the average extend is between 8 and 13 km (Sánchez Hernández 2022). The longest drought trajectories mainly have origin in Guatemala and El Salvador, where are related to wind patterns, low atmospheric moisture, and high pressure/temperature (Cook et al. 2018; Dominguez and Magaña 2018; Aguilar et al. 2005). Orographically moderate thermal gradient phenomena in the western of the Central America dry corridor countries, allow decreasing the severity of drought and extension (Sánchez Hernández 2022).

However, the trajectories determined for the indices derived from remote information are mainly characterized by being short compared to the previous ones (average distance less than 3 km), which, although they do not allow the evaluation of cluster migration or large drought events over long distances, they do give an overview of the average direction in which they manifest themselves and their orientation allows the adoption of crop protection technologies or agricultural practices.

As can be seen in Fig. 5.21, the orientation is in a northeast position, and when contrasted with Fig. 5.20, there is a high relationship concerning regional hydro climatological aspects.

5.5 Conclusions

Central American dry corridor, drought impacts and water scarcity is increasing. Spatiotemporal analysis and validation of remote sensing information is a game-changer in this type of study requirements, becoming efficiently applicable in sectors with low data availability or early warming tool for regional monitoring (Gong 2022; Jiang et al. 2017; Zhang et al. 2019).

This study evaluated the drought dynamics using a reproducible drought vegetation index VCI using remote sensing data to 2000–2020, allow identifying the stressed vegetation spatial distribution, associated with drought conditions with cutting-edge technologies.

Some preliminary results allow using remote sensing data as NDVI as drought monitoring data source. However, remote sensing data still have some limitations that should be considered in future studies. Radiometric interferences, satellites position can affect the geometry of source, scale, for instance.

Regarding drought areas, the approximation and adaptation of the spatiotemporal analysis methodology allowed us to observe a greater concentration of drought events in regions that are directly affected by global oscillations such as ENSO, hurricanes (geographical limits with Pacific Ocean).

El Salvador, Nicaragua, Costa Rica and Guatemala for example, in the northwest, show the highest density of events per phenological cycle (January–May) and (July–October). Additionally, due to aspects such as resolution and scale of the index, it is established that the most important migrations of water stress are highly related to the mentioned density, which allows the generation of strategies or opportunities for sustainable agriculture and improvement of drainage systems.

Central America dry corridor countries, over 20 years was found that the relationship between the SPI water deficit on short-term scales together with the VCI values increases proportionally, however, SPEI shows a stronger relationship because of the physical comparison of two observable or measurable agricultural conditions (vegetation condition, evapotranspiration) observed for SPEI 6.

For the years 2000, 2004, 2015, and 2019, longer drought durations and extensions were observed, which lead to extreme events and are highly related to the anthropic migrations of land use reported by FAO (2019); To reduce El Niño’s impact on Central America’s Dry Corridor, build resilience and invest in sustainable agriculture 2022; World Food Programme 2002).

5.6 Recommendations

According to the findings of this chapter, it was validated that the VCI has greater potential for agricultural drought monitoring than NDVI, and that NDVI is highly correlated with water deficit SPI in 3 months lags and SPEI in 6 months lags. However, some interferences may be associated with the scale magnitudes of the products, which is suggested to verify in detail which geo-correlation and de-scaling method allows geostatistics to have more consistent products, as well as methodologies such as Whitaker (Jiao et al. 2016) for time series smoothing.

More research is needed in case we want to apply this methodology at the crop level, which can strengthen the analysis by establishing productivity in terms of stress associated with drought, thus generating more detailed results and greater efficiency in terms of drought to the cross-border productive sector.

Finally, it is recommended the use of matching algorithms based on geostatistics to improve the performance of both satellite-based indices and their trajectory tracking, such as GWR and Multivariate Kriging, mainly (Nejadrekabi et al. 2022; Dutra et al. 2021; Baniya et al. 2019).