Assessment of vegetation variation and its response to ENSO and IOD in the semi-arid ecosystem of Western India

Vegetation cover over the Earth's surface changes rapidly due to several parameters including climatic factors like changes in precipitation and temperature. Various patterns of vegetation on the Earth's surface and the annual dynamics of the photosynthetic activity of vegetation cause variations in the phenology (Myneni et al. 1997; Claussen et al. 2003). Overall, it may not be possible for the vegetation to be uniformly distributed throughout as it is constrained by the availability of water and nutrients. Apart from many factors which cause variations in vegetation like Land Use Land Cover (LULC), topography, etc., climatic factors including water and moisture transport in the environment contribute to the variation of vegetation (Piao et al. 2019). The distinct climatic patterns like dry and wet climates in the semi-arid region are mainly due to the differences in large-scale ocean – atmospheric circulation and different climate feedbacks (Liang et al. 2022). Ummenhofer et al. (2011) have investigated the impacts of El Niño-Southern Oscillation (ENSO) and the Indian Ocean Dipole (IOD) on Australian rainfall and soil moisture at different time scales. The study finds that there are significant differences between the dominant drivers of drought at interannual and decadal time scales. On interannual time scales, both ENSO and IOD alter the Southeastern Australian soil moisture, with the driest conditions over the southeast and more broadly over large parts of Australia occurring during an El Niño and positive IOD event. The study performed by Liguori et al. (2022) also shows that the positive phases of ENSO and IOD often co-occur making it difficult to attribute observed rainfall anomalies in Australia to either of them individually. The study also concludes that understanding the independent contributions of ENSO and IOD to rainfall variability is important for improving seasonal rainfall forecasts.

This indicates that climatic patterns like ENSO and IOD have distinct impacts on rainfall and soil moisture which in turn controls vegetation. Vegetation studies conducted using satellite data generated by NDVI have been helpful for land management, planning, long-term sustainability and assessing the natural resources and agricultural production (Nemani et al. 2003). Normalized difference vegetation index (NDVI) data have been used to study the interconnections between climate and landscape ecosystems, to monitor the effects of natural as well as anthropogenic disasters such as floods, drought, fire and desertification (Myneni et al. 1997; Seelan et al. 2003).

The most critical environmental and climate issues facing the world today are climate extremes (Stocker 2014), such as droughts and floods (nowadays flash floods), which ultimately change the composition, cover, structure and functions of surface vegetation. The impacts of climate extremes events such as ENSO and IOD may also be appropriately investigated by studying the vegetation dynamics. El Niño (warm phase) and La Niña (cold phase) are the two phases of the ENSO, while positive IOD and negative IOD are the two phases of the IOD. Regardless of the significant correlation between ENSO and the IOD (Allan et al. 2001), it is still uncertain if the IOD is independent of ENSO (Feng & Meyers 2003; Cai et al. 2012). Future studies on these aspects will help understand the certainties and uncertainties between the ENSO and IOD events of climate change.

Numerous research studies have examined the monsoon rainfall variability and its connection to ENSO and IOD (Ghosh et al. 2009; Propastin et al. 2010; Bothale & Katpatal 2016; Rishma & Katpatal 2016; Shukla & Huang 2016; Hrudya et al. 2021). It is essential to research how ENSO and IOD affect vegetation because ENSO and IOD have an impact on the variability of monsoon precipitation, which is a significant driver of vegetation variation (Krishnaswamy et al. 2015). In addition, the global and regional hydrological balance affects vegetation due to the spatial and temporal variation in precipitation.

Few studies have studied the impact of IOD and ENSO on the country-level geographical area over Madagascar (Ingram & Dawson 2005) and South Africa (Anyamba et al. 2018). According to Ingram and Dawson's study of Madagascar, the moderate but significant correlation observed here does indicate that El Niño has a considerable impact on the greenness level of vegetation on the island. The analysis of the strongest El Niño event followed by a weak La Niña event was conducted by Anyamba et al. in 2018. Time series correlation analysis was carried out between Niño 3.4, the Dipole Mode Index (DMI), precipitation and NDVI. The study concluded that the ENSO and IOD contributed to the effect on NDVI and agricultural production.

However, very few studies have been carried out on the subsequent impact of ENSO on vegetation variation. The present study aims to evaluate how sensitive the vegetation is to ENSO and IOD events at watershed/regional levels rather than at the country level/global levels. First of all, the ENSO (El Niño and La Niña) and IOD (positive and negative IOD) events are identified based on Oceanic Niño Index (ONI) and DMI, respectively. Then the MODIS NDVI data were processed for the period of selected ENSO and IOD years. Finally, the spatial variation of the NDVI has been analyzed using the inverse distance weightage (IDW) interpolation using geospatial techniques. According to the spatial variation analysis results, El Niño events have a more negative effect on the vegetation than La Niña events when ENSO events are taken into account.

The NDVI anomalies are calculated for monthly NDVI data and are processed using Geographic Information System (GIS) tools to generate NDVI anomaly maps. Based on the extreme anomaly values, sensitive vegetation pixels are identified. There are 71 micro watersheds in the sub-basin. The global spatial autocorrelation and LISA has been performed to analyze the anomalies in the watershed. In local spatial autocorrelation (LISA) analysis, cluster maps and significant maps are generated in the Geo-Da tool to identify the most affected micro-watersheds within the sub-basin. The novelty of this study is it introduces the ENSO and IOD-sensitive vegetation areas within the watershed using vegetation means to difference anomalies. The study utilizes LISA to identify clusters of anomalous pixels showing areas sensitive to ENSO and IOD events. To identify the seasonal effects on NDVI, ENSO-sensitive and IOD-sensitive vegetation areas within the watershed have been identified for January February March (JFM), April May June (AMJ), July August September (JAS) and October November December (OND) months. The results of this study are important as the study area consists of a rainfall shadow zone which is a drought-prone area in western India with diversified hydrological characteristics.

The paper is organized as follows: an overview of the data used in this study is presented in Section 2; the method used to identify the spatial variation, sensitive pixels and spatial autocorrelation is described in Section 3; the results and discussion are presented in Section 4, and conclusions are summarized in Section 5.

Brief information about the study area used for the research and an overview of the data used in this study are presented in this section. The Upper Bhima Sub-basin lies in the semi-arid ecosystem of Western India. This sub-basin is an example of a unique combination of diverse agro-climatic zones. Sahyadri Mountain Range divides the study area into the Western Ghat zone, which receives the highest rainfall, and the Water Scarcity zone, which is a shadow zone of precipitation and receives the lowest rainfall in the sub-basin. This section also provides information about the MODIS NDVI data and the information on ENSO and IOD events used for the analysis.

The hydrological characteristics of the Upper Bhima River Sub-basin are highly diverse, caused by the interaction between the monsoon and the Western Ghat mountain range. The Western Ghat zone is covered with thick forest and receives heavy rainfall, reaching a maximum of 4,500 mm/year. Rainfall decreases rapidly towards the eastern slopes and plateau areas, where it is less than 500 mm/year. It again increases towards the east; hence, the central part of the Upper Bhima receives the lowest rainfall (Garg et al. 2012).

The mathematical expression for NDVI is given in Equation (1). NDVI is the index which is a good indicator of growth stages and vegetation parameters, such as leaf area, green biomass yield, vegetation coverage rate and photosynthetically active vegetation, and so on (Piao et al. 2003; Stow et al. 2003; Morgan et al. 2020).

NDVI quantifies the vegetation by using the reflection in the near-infrared region which vegetation strongly reflects and the red region which vegetation absorbs. NDVI is an index of the vegetation greenness, i.e., the density and health of vegetation of each pixel in a satellite image. NDVI is often used worldwide to monitor drought, monitor and predict agricultural production, assist in predicting hazardous fire zones and map desert encroachment (Arjasakusuma et al. 2018).

R is reflected value of red band.

ENSO (El Niño-Southern Oscillation) is the oscillations between wind and sea surface temperature (SST) that occurs in the equatorial eastern Pacific, which is represented by the transition of El Niño and La Niña in the ocean and the Southern Oscillation in the atmosphere (McPhaden et al. 2006). The ENSO significantly impacts global climatic parameters and the hydrological cycle.

The IOD is defined by the difference in SST between two areas or poles; a western pole in the Arabian Sea (Western Indian Ocean) and an eastern pole in the eastern Indian Ocean south of Indonesia; hence it is called as a dipole. The IOD affects the climate of Australia and other surrounding countries of the Indian Ocean and is a significant contributor to rainfall variability in this region (Cai et al. 2011; Min et al. 2013). As IOD affects the Indian Ocean surrounding countries and rainfall variability in those countries, it is essential to study its impact on the change in NDVI pattern.

The Bureau of Meteorology, Australia has decided on three positive IOD years and three negative IOD years during the study period. NDVI analysis has been done in accordance with those IOD years. The list of IOD years considered for the analysis is given in Table 2.

In this section, the methodology used for the study is briefly explained. The methodology includes the NDVI data processing and its representation in ArcGIS software, calculation of monthly NDVI anomalies for the study period, and spatial autocorrelation (global spatial autocorrelation and LISA) of monthly NDVI anomalies for 71 micro watersheds in Geo-Da software.

The downloaded MOD13C2 NDVI data product has been processed in ArcGIS software. First, the raster data in the form of pixels are converted to vector data (points) to create spatial variation maps. Then, the point-converted NDVI data was used to generate spatial variation maps of NDVI. The IDW interpolation technique has been used to perform the interpolation analysis. The IDW interpolation technique is suitable for the sample points that are equidistant and in a grid format. So, as points are converted from the raster, all the points are equidistant and gridded. Then the generated NDVI variation maps are reclassified into three classes; less than 0.3, 0.3–0.5 and more than 0.5 based on NDVI value. Finally, the area covered by all three classes of NDVI has been calculated and analyzed.

where,

Anomaly_NDVI is the monthly NDVI anomaly for January 2001;

X is the actual value of the average monthly NDVI January 2001 and

is the long-term average monthly NDVI for January (an average NDVI value of January month from 2001 to 2022)

After the monthly NDVI anomalies have been calculated, the pixels sensitive to ENSO and IOD events are identified based on the range of NDVI anomaly values. The threshold value considered in this study is +0.05 for positive anomalous pixels and −0.05 for negative anomalous pixels, i.e., all the pixels having monthly NDVI anomaly value greater than +0.05 are positive sensitive pixels and all the pixels having monthly NDVI anomaly value less than −0.05 are negative sensitive pixels. The threshold values were selected based on data distribution of anomalous pixels. Sample histograms of the normal distribution of anomalous pixels are shown in Figure S1. The number of anomalous pixels will vary if the threshold values change. More data points will probably be classified as positive or sensitive anomalous pixels when the threshold value is lowered, and vice versa. The NDVI anomaly values are analyzed for the El Niño, La Niña, Positive IOD and Negative IOD years, and the specific event-sensitive pixels are identified based on monthly NDVI anomaly values.

This study uses spatial autocorrelation analysis to identify the growth and spatial features of the monthly NDVI anomaly over the Upper Bhima Sub-basin. Spatial autocorrelation indicates the interconnection between a specific attribute value at a regional level and the same attribute value in a neighboring geographical location. It is a metric of the degree of value aggregation in the spatial domain. This aggregation property is measured using Moran's I, divided into global spatial autocorrelation and LISA. The global spatial autocorrelation has been applied to the monthly NDVI anomalies, but the results of all the events show clustered patterns with zero or negligible p values. So, the LISA has been applied for the analysis.

The LISA coefficient is used to determine whether there is spatial clustering of NDVI anomaly. The LISA coefficient greater than 0 indicates that there is a positive spatial correlation between the local spatial unit and the nearby spatial unit, which is represented by ‘high–high’ or ‘low–low’; the LISA index less than 0 indicates ‘low–high’ or ‘high–low’. The performance of the aggregation is negatively correlated.

The spatial distribution of NDVI during El Niño and La Niña events have been generated for JAS, OND, JFM and AMJ months and are shown in Figures 5 and 6, respectively. The total area covered by the sparse, medium dense and highly dense vegetation during El Niño and La Niña events is tabulated in Tables 3 and 4, respectively.

It has been observed that in AMJ months during both the ENSO phenomenon, the spatial variation shows a similar pattern and a majority of the area is covered by sparse vegetation (NDVI < 0.3). The percentage area covered by sparse vegetation in AMJ months of El Niño events are 84.45 (2002–2003), 78.95 (2004–2005), 70.07 (2009–2010) and 80.76 (2015–2016) and for La Niña events are 76.54 (2007–2008), 67.01 (2010–2011), 72 (2011–2012) and 37.22 (2020–2021). This illustrates that ENSO has no impact on vegetation during AMJ.

The micro-watersheds from the western and southern regions of the sub-basin are most vulnerable to El Niño events as this region shows sparse vegetation during El Niño events. The central, eastern and southern region of the sub-basin is already in the rainfall shadow zone.

The impact of ENSO on NDVI is clearly observed during JAS and OND months as shown in Figures 5 and 6. The percentage area covered by dense vegetation (NDVI > 0.5) during La Niña years in JAS and OND is more as compared to El Niño years. The percentage area covered by dense vegetation during El Niño events in JAS months are 19.94 (2002–2003), 24.34 (2004–2005), 40.53 (2009–2010) and 24.21 (2015–2016) and in OND months are 8.11 (2002–2003), 11.66 (2004–2005), 35.53 (2009–2010) and 26.71 (2015–2016). While, percentage area covered by dense vegetation during La Niña years in JAS months are 55.75 (2007–2008), 78.55 (2010–2011), 42.06 (2011–2012) and 95.29 (2020–2021) in OND months are 17.11 (2007–2008), 61.91 (2010–2011), 25.21 (2011–2012) and 67.55 (2020–2021). This variation in the percentage of area covered by highly dense vegetation shows how ENSO has a significant influence on the vegetation during JAS and OND months. The impacts of ENSO on the spatial variation of NDVI are also observed during JFM months. The area covered by sparse vegetation in JFM months during El Niño years is more than that of La Niña years. The results related to the impact of El Niño and La Niña year on vegetation match with the study conducted in Haryana state in India by Chauhan et al. (2022). The study analyzed the vegetation for monsoon only and concluded that vegetation in monsoon during La Niña is higher than in El Niño.

The western region within the study area has a dense vegetation cover due to geographical factors and the Sahyadri mountain range. According to the fluctuations in temperature and precipitation, El Niña is a warm phase, whereas La Niña is the cold phase of ENSO and the El Niño event has less rainfall in India than the La Niña event (Wang et al. 2014).

The results show a distinct dipole pattern that is reversed between El Niño and La Niña and shows that vegetation conditions are highly connected with ENSO climatic phenomenon. In the Upper Bhima sub-basin, ENSO events are controlled due to the oscillations between wind and SST controlling the amount of monsoon rainfall in the sub-basin. This results in less rainfall during El Niño events and excessive rainfall during La Niña events, which is evident from the regional vegetation growth pattern.

Sustained changes in the difference between sea surface temperatures of the tropical western and eastern Indian Oceans are known as the Indian Ocean Dipole or IOD. According to the fluctuations in SST, the IOD event is considered a positive (negative) IOD when the Western Indian Ocean is warmer (cooler) than the Eastern Indian Ocean. The rise in temperature in a particular geographical region causes a rise in convection which ultimately causes an increase in the rainfall in that region. As a result, the countries surrounding the Indian Ocean are affected by the IOD events. There are many studies and research on the impact of IOD events on African countries surrounding the Eastern Indian Ocean (Anyamba et al. 2018), concluding that IOD has significant impacts such as the incidence of severe floods due to an increase in convection in that area. As India is geographically located between the Eastern and Western Indian Oceans, it is very complicated to decide whether IOD events affect India's monsoon cycle.

The spatial distribution of vegetation derived NDVI values in positive and negative IOD events is shown in Figures 7 and 8, respectively. To analyze the regional variations and change in the pattern of vegetation during positive and negative IOD events, the area covered by the three vegetation classes (sparse vegetation, medium vegetation and highly dense vegetation) in the Upper Bhima Sub-basin, is shown in Table 5. During IOD events also, the NDVI variation has been analyzed for JFM, AMJ, JAS and OND months to understand the seasonal variation.

It has been observed that in AMJ months during both the positive and negative IOD events, the spatial variation shows a similar pattern and the majority of the area is covered by sparse vegetation (NDVI < 0.3). The percentage area covered by sparse vegetation in AMJ months during positive IOD events are 59.34 (2006), 69.85 (2012) and 53.50 (2015) and of negative IOD events are 51.01 (2010), 55.97 (2014) and 74.68 (2016). This similarity in the percentage area covered by sparse vegetation in AMJ months during both positive and negative IOD events illustrates that IOD has much less impact on the vegetation in AMJ months.

The impact of IOD on NDVI is clearly observed during JAS and OND months as shown in Figures 7 and 8. The percentage area covered by dense vegetation (NDVI > 0.5) during negative IOD events in JAS and OND is more as compared to positive IOD events. The percentage area covered by dense vegetation during positive IOD events in JAS months are 47.03 (2006), 26.31 (2012) and 24.21 (2015) and in OND months are 38.68 (2006), 26.24 (2012) and 26.71 (2015). While, the percentage of area covered by dense vegetation during negative IOD events in JAS months are 78.55 (2010), 38.69 (2014) and 63.40 (2016) and in OND months are 61.91 (2010), 33.03 (2014) and 24.74 (2016). Also, the sparse vegetation is not even present in JAS and OND months during negative IOD events and is present up to 20% in JFM. This difference in the area covered by dense vegetation illustrates that a positive IOD event has a major adverse impact on the vegetation during JAS and OND months than a negative IOD event.

The results show a similar dipole pattern as seen in ENSO events. The vegetation patterns in positive and negative IOD occurrences indicate that it is connected to the IOD climate phenomenon. Positive and negative IOD events caused by changes in SST in the Indian Oceans regulate rainfall in the Upper Bhima sub-basin, impacting the regional vegetation pattern.

The pixels sensitive to ENSO and IOD events are determined based on the range of NDVI anomaly values. Based on the NDVI anomaly range, the threshold value considered in this study is +0.05 for positive anomalous pixels and −0.05 for negative anomalous pixels. Positive anomaly pixels have a monthly NDVI anomaly value larger than +0.05, whereas negative anomaly pixels have a monthly NDVI anomaly value less than −0.05.

In the El Niño and La Niña events for JFM and AMJ, the area covered by negative sensitive pixels is much less than positive anomaly pixels. All positive anomaly pixels in El Niño and La Niña events during the JFM and AMJ seasons are dispersed throughout the study area and are not concentrated in any specific region. Also, during JAS and OND seasons, i.e., during the monsoon and post-monsoon period, El Niño and La Niña events show the opposite results. JAS and OND seasons have maximum negative sensitive pixels in El Niño events and maximum positive anomaly pixels in La Niña events. Hence, it may be concluded that El Niño has an adverse effect on the monsoon rainfall, which can be identified from the presence of a large number of negative anomaly pixels during the post-monsoon period (JAS and OND). The study conducted by Lü et al. (2012) analyzed the El Niño and La Niña sensitive vegetation and concluded that the area of El Niño sensitive vegetation was much larger than the area of La Niña sensitive vegetation, although the study was conducted for specific El Niño and La Niña events without considering the seasonality.

In case of IOD events, positive anomaly pixels are almost absent in the JFM season of positive IOD events and are very few in JFM and AMJ in negative IOD events. Also, there are very few negative anomaly pixels in all seasons of positive IOD events. However, during JAS, i.e., during monsoon in positive IOD event, negative anomaly pixels are concentrated in the southern region of the study area, which is a drought-prone area due to the rainfall shadow zone and all positive anomaly pixels are distributed throughout the study area. This concludes that negative IOD events have a negative impact during JFM and AMJ seasons, and the rainfall shadow zone is also affected adversely during JAS of positive IOD events.

The concentration and significance of monthly NDVI anomaly within a specific region can be clearly expressed through LISA analysis.

In accordance with Table 6, the spatial clusters (high–high and low–low) in the total study area are higher than that of spatial outliers for all the ENSO and IOD events. However, El Niño and La Niña cluster maps reveal contrasting results. In El Niño events, the Western Ghat region and surrounding region show a cluster of high anomaly values and a cluster of low anomaly values in the study area's eastern and northeastern parts throughout the year except JFM. In JFM, the cluster of high and low anomaly values is present in an irregular pattern, as shown in Figure 10.

The Moran index for all the events lies within an 0–1 interval indicating that the monthly NDVI anomaly of 71 micro watersheds has significant spatial autocorrelation. Local Moran's I of 71 watersheds in the study area are given in Table 7. Moran's I autocorrelation values of JFM, AMJ, JAS and OND in El Niño events are 0.387, 0.626, 0.779 and 0.809, respectively, and in La Niña events, 0.449, 0.707, 0.701 and 0.562, respectively.

Whereas in La Niña events (Figure 11), high anomaly values are clustered in the eastern and central part of the study area, and low anomaly values are clustered in Western Ghat and the surrounding region for the whole year except OND. Here, in the La Niña event during the post-monsoon period, i.e., in OND, both high and low clusters are concentrated in the eastern part of the study area, and the low–low cluster is located precisely at the rainfall shadow zone. This indicates that, even if precipitation is higher during a La Niña year, the rainfall shadow zone does not receive adequate water.

In positive IOD events, significantly fewer micro watersheds show significant clusters generated after LISA analysis compared to the ENSO events (Figure 12). This indicates that a positive IOD event does not affect the monthly NDVI anomaly. However, the central region of the study area in JAS and OND during positive IOD events shows a cluster of high NDVI anomaly values, and negative anomaly pixels are clustered in the southern part, which is a drought-prone area during the post-monsoon period (OND).

In negative IOD events, JFM and AMJ seasons have high anomaly values clustered in the western region and low anomaly pixels are clustered in the central and southern regions (Figure 13). During the post-monsoon period, very few micro watersheds have high and low clusters. As far as the IOD events are considered, the negative IOD event has more effect on monthly vegetation anomaly clusters than the positive IOD event.

The outcome of this study will be helpful to key stakeholders such as farmers, the Water Resource Department in the state of Maharashtra, the Central Water Commission of India and policymakers. The study provides extensive information on the response of vegetation to climate change events that can be used for the implementation and planning of farming-related activities to resolve the problems faced by farmers during extreme events like floods and droughts. The study also gives information about the impact of ENSO and IOD events on vegetation cover and the location of vegetation clusters in the study area, i.e., it can also be used for assessment and planning of agricultural activities. For example, seasonal variation can be mapped in relation to ENSO and IOD events in the agricultural region (Figure 9). Vulnerable micro watersheds will be given more priority for future planning of water supply and demand management (Figures 10–13).

This study presents a complete assessment of vegetation sensitivity to ENSO and IOD events over the Upper Bhima Sub-basin, Western Maharashtra, India, from 2000 to 2022. MODIS NDVI data were used to represent vegetation, and the ENSO and IOD events were defined. El Niño and La Niña events of ENSO are identified using the Southern Oscillation Index (SOI), and positive and negative IOD is identified using DMI. This study identifies ENSO and IOD-sensitive vegetation regions for selected ENSO and IOD events from 2000 to 2022 monthly NDVI anomalies. The following conclusions can be drawn based on this work:

• 1.

  Seasonality affects the vegetation variation during all the events due to the rainfall and temperature variation which are predominant parameters causing vegetation variation at the regional level.

• 2.

  The effect of ENSO on NDVI is noticeable and observed during JAS and OND months as percentage area covered by dense vegetation during El Niño event is less than 30% and that of during the La Niña event is more than 50%. Therefore, when ENSO events are considered, most of the vegetation in the watershed is adversely affected during El Niño which is the warm phase of the ENSO event while higher NDVI values are observed during the La Niña events as expected. Hence, a dipole pattern is seen during ENSO events.

• 3.

  The vegetation patterns in positive and negative IOD events indicate that vegetation is connected to the IOD climate phenomenon. Positive and negative IOD events caused by changes in SST in the Indian Oceans regulate rainfall in the Upper Bhima sub-basin, impacting the regional vegetation development pattern. The percentage area covered by dense vegetation is more in negative IOD events than positive IOD events during JAS and OND, i. e. during monsoon and post-monsoon periods. Hence it can be concluded that a positive IOD event has an adverse effect on vegetation growth compared to a negative IOD event. The lowest vegetation has been seen in the drought-prone areas which is the rainfall shadow zone in the watershed.

• 4.

  The identification of positive and negative sensitive anomaly pixels shows that El Niño has a negative impact on the monsoon rainfall which can be identified from the presence of a large number of negative sensitive pixels in post-monsoon period (JAS and OND) during El Niño events and a large number of negative sensitive pixels are recognized in post-monsoon period (JAS and OND) during La Niña events.

• 5.

  The negative IOD events have a negative impact during JFM and AMJ seasons, and the rainfall shadow zone is also affected negatively during JAS of positive IOD events.

• 6.

  The cluster map and outliers of the monthly NDVI anomalies are identified using LISA analysis in Geo-da. The results of the LISA analysis also show that the El Niño and La Niña events affect the vegetation in the watershed. During El Niño events, the micro watersheds in the eastern region show lower anomaly clusters and the western region shows high anomaly clusters. The micro watersheds in the central and eastern regions during La Niña events shows high anomaly clusters and the western region shows low anomaly clusters. This shows that the central and eastern region which is the draught prone areas is predominantly affected due to the ENSO phases.

• 7.

  According to the LISA analysis, the rainfall shadow zone in the study area has a cluster of negative sensitive pixels even in the monsoon and post-monsoon periods except for negative IOD events.

The limitation of the study is that it focuses only on the response of vegetation to ENSO and IOD events and does not consider other meteorological and local influencing factors that may affect vegetation dynamics. As a future scope of the study, the analysis conducted in the present study can also be performed by using Enhanced Vegetation Index. The analysis can also be performed by dividing the study area into multiple LULC classes. However, the study recommends a novel method that analyzes the impact of ENSO and IOD events to identify vulnerable regions which helps in the effective planning and management of water in agricultural basins. The methodology proposed in this study has been applied to the upper Bhima sub-basin, although it can be applied to any other basin worldwide.

We thank the anonymous editors and reviewers for suggestions on the manuscript. This research received no specific grant from any funding agency in the public, commercial or not-for-profit sectors.

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.