Ecosystem Services 21 (2016) 2030

Contents lists available at ScienceDirect

Ecosystem Services
journal homepage: www.elsevier.com/locate/ecoser

Community-based groundwater and ecosystem restoration in

semi-arid north Rajasthan (3): Evidence from remote sensing
Thomas Davies, Mark Everard n, Michael Horswell
University of the West of England (UWE), Coldharbour Lane, Frenchay Campus, Bristol BS16 1QY, UK

art ic l e i nf o a b s t r a c t

Article history: Community-based measures recharging groundwater in semi-arid India has historically underpinned
Received 18 February 2016 rural socio-ecological resilience, though are declining through technological, policy and other changes.
Received in revised form Nevertheless, exemplars of community action are achieving catchment regeneration, including in Alwar
2 July 2016
District (Rajasthan) since the mid-1980s. This study analysed satellite remote sensing (RS) data to detect
Accepted 6 July 2016
Available online 28 July 2016
trends in groundwater and linked ecosystem services. Data from Landsat satellite missions offered a long
time series and free access, though data gaps in the LandSat archive prior to 1997 limited time series
Keywords: analysis. ISODATA (Iterative Self Organising Data Analysis Technique) was used to analyse land cover
Ecosystem services trends, detecting increasing vegetation cover but not river rejuvenation due to limited spatial and
Community-based management
spectral resolution. Analyses of NDVI (Normalised Difference Vegetation Index) and MSI (Moisture Stress
Water
Index) were used to assess change in vegetation cover, vigour and moisture stress over time. Analytical
Regeneration
Cultural outputs were equivocal, although inter-annual uctuations were observed to follow antecedent rainfall
Aquatic ecosystems as vegetation responded to rising soil moisture and groundwater. Despite these equivocal conclusions,
Remote sensing the research strongly suggests that analysis of RS data with improved resolution can provide surrogate
indicators of change in groundwater and associated ecosystem services, supporting formulation of
exible policies incorporating local action to regenerate socio-ecological systems.
& 2016 Elsevier B.V. All rights reserved.

1. Introduction also increased the vulnerability of groundwater to pollution and

increasing salinity and other forms of contamination (Postel,
Water security is and will remain a pervasive challenge in the 2015). However, the 2030 Water Resources Group (2009) predicts
face of a rising global population and a changing climate, parti- that the national supply will fall 50% below demand by 2030,
cularly across the developing world. Over 1.7 billion people glob- deepening what is already recognised as a water security crisis.
ally live in river basins where overuse of water contributes to Aside from its direct uses, water is a vector of a wide range of
depletion at a rate that exceeds natural recharge; were this trend ecosystem services upon which ecosystem and human wellbeing
to continue, two-thirds of the global population will be living in depend. Understanding groundwater trends is crucial if more
water-stressed conditions by 2025 (USAID, 2015; United Nations, sustainable management is to be achieved, particularly in arid and
2015a; Chenworth, 2008). Water is consequently an underpinning semi-arid areas where water is often a principal limiting factor to
resource for a number of the UN Sustainable Development Goals development.
(United Nations, 2015b), both in terms of indirect access to clean, In the face of declining groundwater trends across India, there
have been a number of localised successes in regeneration of
safe water and in its roles in sanitation, health, food and energy
ecosystems and closely linked socio-economic fortunes. One such
production and other human needs. Groundwater has played a
regional example has been achieved by community-based activ-
signicant role in increasing India's agricultural output and pro-
ities regenerating groundwater and the wider socio-ecological
gress with water security, supporting 62% of India's total irrigated
system in a number of adjacent catchments in Alwar District, Ra-
agricultural area and over 80% of rural and urban water supplies in
jasthan, north India (Everard, 2015, 2016). Assessing groundwater
20092010 (Central Water Commission, 2013; Kidwell, 2015).
trends and the efcacy of local measures to regenerate it over wide
Overexploitation of groundwater, particularly for irrigation, has
geographical areas is challenging. For this reason, cheaper and
resulted in substantial, widespread depletion of water tables and broader-scale methods must be found to inform more sustainable
approaches to the management of water and ecosystems, with
n
Corresponding author. detection of groundwater by direct or surrogate means a proxy for
E-mail address: [email protected] (M. Everard). a broader set of linked ecosystem services vital for human

http://dx.doi.org/10.1016/j.ecoser.2016.07.007
2212-0416/& 2016 Elsevier B.V. All rights reserved.
 T. Davies et al. / Ecosystem Services 21 (2016) 2030 21

development and ecosystem resilience. Becker (2006) notes that surface expressions of groundwater
The potential of remote sensing (RS) for groundwater mon- are usually identiable through vegetation, either as stress in
itoring has been explored extensively over the past thirty or so waterlogged soils or as vigour in water-resistant species. In arid
years (Heilman and Moore, 1982; Meijerink, 1996; Edet et al., 1998; areas in particular, where water is the main limiting factor con-
Rosenberry et al., 2000; Jin et al., 2007; Lv et al., 2013; Chinnasamy trolling plant growth such as our study area, vegetation can pro-
et al., 2015) and reviewed by Becker (2006) and Prez Hoyos et al. vide important clues regarding the occurrence of groundwater.
(2016). Becker's (2006) observation regarding the key constraint However, Becker (2006) also notes that it is often difcult to se-
on RS application in groundwater studies remains true today: parate such expressions of groundwater from seasonal or event-
current space and airborne sensors have little capacity to pene- driven surface water. The review by Prez Hoyos et al. (2016)
trate the ground surface. Consequently, RS approaches to conrms a strong coupling between rainfall, water table depth,
groundwater monitoring have focussed principally on the use of vegetation and soil. Tth (1963) conceptual model of groundwater
multispectral observations of the earth surface, relying on proxy ow has implications: if the adopted RS approach is reliant on
indicators of groundwater such as vegetation or thermal uxes. As surface expressions of groundwater or groundwater proxies,
a more recent alternative approach, the launch of the Gravity characteristic of systems with relative shallow groundwater ow,
Recovery And Climate Experiment (GRACE) satellite mission in then the shallower the groundwater ow, the smaller the surface
2002 enables detection of changes in the gravitational eld of the expressions of the ow are likely to be, and the higher the required
Earth, which is affected by the presence of large volumes of water. resolution of the RS imagery becomes.
Chinnasamy et al. (2015) used GRACE data to investigate Resolution of satellite imagery can be considered in three
groundwater storage and recharge within different agroclimatic realms: spatial resolution; spectral resolution; and temporal re-
zones in Rajasthan, India. Their methodology proved useful for solution (see Box 1). There is an inevitable tension between the
detecting large-scale trends, but found trends were highly de- resolution of the sensors and the scale of the phenomena being
pendent on antecedent moisture conditions. Jin et al. (2007) used investigated. For example, GRACE, with a ground resolution of
Normalised Difference Vegetation Index (NDVI) data at 250 m approximately 300 km is of little use in local or sub-regional
spatial resolution from the Modis (Moderate Resolution Imaging projects, while Landsat with a spatial resolution of 30 m is ap-
Spectroradiometer) satellite mission correlated with groundwater propriate for vegetation studies, however, the 60 m resolution of
depth interpolated from a series of observation wells to assess the the thermal bands is probably too coarse to resolve local surface
relationship between vegetation growth and groundwater in the expressions of groundwater such as springs. The use of any higher
arid Yinchuan Plain in China. Their ndings suggest a relationship resolution imagery sources (such as QuickBird or IKONOS) is
exists in areas with relatively shallow groundwater (NDVI values of generally constrained by cost. The implication of the above is that,
0.4 and above), suggesting that green vegetation characterised in a cost-constrained context with a local focus, the use of vege-
areas where depth to groundwater was less than 6.2 m. The tation uxes as a proxy indicator of groundwater uxes is a de-
highest NDVI values were associated with a groundwater depth of fensible, possibly inevitable, choice.
3.5 m. In the arid Ejina area, Jin et al. (2007) found that NDVI This study interpreted satellite remote sensing imagery as
peaked at groundwater depths of 3.4 m. More recently, Lv et al. evidence to test trends in groundwater and ecosystem regenera-
(2013), using LandSat TM data (30 m resolution), found that higher tion in Alwar District, Rajasthan, reported by Everard (2015). We
NDVI values in the semi-arid Hailutu River catchment were highly chose to use a time series of Landsat5 TM and Landsat8 OLI data
dependent on groundwater availability when groundwater depth because of the catchment scale of the project, the available record
was less than 10 m. These ndings suggest a stronger relationship length of data, its suitability for vegetation studies (due to its
between vegetation growth (measured using the NDVI) and a spectral resolution) and the fact that access to the data is free.
shallower water table in arid areas, than in semi-arid areas. All
these studies highlighted concerns about the inuence of local
factors and contaminating variables, such as the local species 2. The case study area
composition inuencing NDVI values, or climate (e.g. antecedent
moisture conditions), or local soil characteristics and topography. Rajasthan is India's largest state, occupying 10% of India's land

Box 1Spatial, spectral and temporal resolution of satellite imagery.

Spatial resolution determines the detail discernible, generally defined by the smallest feature that can be detected. Remotely sensed
images comprise a matrix of pixels (the smallest units of an image, normally square, representing a specific area of the image).
Spatial resolution is related to pixel density on the sensor as well as the distance between the target being imaged and the sensor
platform (satellite altitude). The spatial resolution of passive sensors, as used for example on the Landsat missions that provided
images used in this study, depends primarily on their Instantaneous Field of View (IFoV: the angular cone of visibility of the sensor).
Each recording cell on the sensor detects average brightness across all sensed features within the cell, so the relative brightness of
even small features can dominate what is detected within a particular cell. Satellite images are divided into three classifications of
spatial resolution: low resolution (301000 m2 for each pixel); medium resolution (430 m2 per pixels); and high resolution images
(0.64 m2 pixel size) (NRCAN, 2012; Satellite Imaging Corporation, 2015).
Spectral resolution describes the ability of a sensor to define fine wavelength intervals. Broad classes of sensed terrains, such as
water and vegetation, can usually be separated using very broad wavelength ranges (such as the visible and Near-IR) though
discerning more similar Earth surface types requires comparison of much finer wavelengths. Most satellite remote sensing systems
are multispectral, recording electromagnetic energy over several separate wavelength bands at different spectral resolutions.
Temporal resolution relates to the collection of imagery of the same area of Earths surface at different periods of time. The
temporal resolution of a sensor depends on a variety of factors, including the satellite/sensor capability, swath overlap (a swath is the
width of images sensed, wide swathes allow more rapid revisit and greater overlap whereas narrow swathes typically allow for higher
spatial resolution but revisit time is less frequent) and latitude. Temporal resolution is an important consideration when persistent
cloud cover obscures the view of the Earths surface, potentially obscuring short-lived phenomena (floods, oil slicks, etc.) (NRCAN,
2014).
22 T. Davies et al. / Ecosystem Services 21 (2016) 2030

area but only 1% its surface water resources. 60% of the state is restoring village-scale community management of water (also
arid, with the remaining 40% semi-arid (Jayanti, 2009). Most of reviewed by Rathore (2003), Sinha et al. (2013) and Everard
western Rajasthan falls under the Sub-tropical arid (desert) cli- (2015)). Central to the TBS approach has been restoration of a di-
mate type in the Kppen climate classication (Kppen and We- versity of locally appropriate water harvesting structures (WHSs)
gener, 1924) whilst the region east of the Aravalli range is Sub- and the community institutions necessary to initiate, build and
tropical humid (wet) experiencing dry winters. 90% of Rajasthan's maintain them as well as to govern water use on an equitable and
annual rainfall, which is often erratic and localised, occurs in the sustainable basis. This builds upon India's long tradition of water
monsoon months of July to September (Jayanti, 2009). In Ra- harvesting as an adaptation to local climatic conditions. These
jasthan, as for much of India, the strongly seasonal rainfall allied diverse WHSs go by a range of regional names, including as a
with a generally hot climate results in a heavy reliance on subset of examples: Baudis particularly in Himachal Pradesh state;
groundwater for domestic, industrial and agricultural purposes. Dongs in Assam; Virdas and jheels in the grasslands of Gujarat;
Ninety-two percent of India's extractions of groundwater is Zabo in Nagaland; Tanks or eris across South India; and Ooranis in
used for irrigation (Central Ground Water Board, 2006). Everard south Travancore (George et al., 2015). However, they are com-
(2015) reviews the substantial extent to which groundwater sup- monly referred to in Rajasthan as johadi (singular: johad). WHSs
ports India's demand for water. Following India's independence, serve to intercept or slow overland ows during monsoon rainfall,
there has been a pronounced trend towards centralisation of promoting groundwater recharge with some also storing surface
control of water management to state and national levels. India's water for year-round use. TBS initiatives, with international aid
constitution states that surface water is a government-owned re- funding matched by village investment, are located in the rural
source and only state governments are empowered to enact water Arvari (or Arwari), Sarsa and Baghani catchments, lying mostly in
management laws, and to control and regulate groundwater ex- Alwar district, north-eastern Rajasthan, the headwaters of which
ploitation. Responsibilities for management of water is divided rise from the Aravalli Hills (Fig. 1). However, work in adjacent
between a range of government institutions at national and state catchments has also restored water resources, with perennial
levels (see Table 1), some of which have competing or conicting ows returning to ve formerly seasonal rivers the Bhagani-
priorities, resulting in a fragmented approach to water manage- Teldehe, Arvari, Jahajwali, Sarsa and Ruparel benetting 250
ment that exacerbates water stress principally by disempowering villages (Jayanti, 2009). Land use in this rural region of Alwar is
community management (Singh and Singh, 2002). Major reduc- predominantly agricultural, much of it dependent upon ground-
tions in local groundwater recharge are compounded by wide- water that has historically been overexploited. Those with re-
spread energisation (diesel and electric pumps) of water extrac- sources to pump from the decreasing reserves gain preferential
tion from ever-deeper wells and tube wells, exacerbating access to the detriment of the majority of people who conse-
groundwater depletion and frequently contributing to societal quently experience decreasing resource availability and ensuing
inequities, vulnerability and poverty. poverty.
Everard (2015) reviews how these factors, exacerbated by po- Rathore (2003), Sinha et al. (2013) and Everard (2015) review
pulation growth and climatic change, have combined to remove progress enabled by TBS working in partnership with villages
water management responsibilities from local control in Rajasthan within these and adjacent catchments, rebuilding village- and
and elsewhere in India with a range of signicant adverse socio- catchment-scale institutions and reinstituting or innovating an
ecological consequences. There has been widespread abandon- extensive network of WHSs that have collectively regenerated soil
ment of locally adapted community management institutions and moisture and groundwater resources in the formerly desertied
rainwater-harvesting techniques reecting historic adaptation to landscapes of river valleys. These initiatives have combined in
local climatic conditions. Declining ecosystem vitality from greater places to reverse the cycle of socio-ecological decline, with
water stress, including the extent of forest cover, lies at the heart quantied increases in groundwater levels, food production, re-
of a cycle of linked ecological and social degradation resulting in institution of governance structures (particularly Gram Sabha:
rural hydrological poverty. This is manifest through the declining traditional village decision-making bodies with interests in de-
viability of stock and crop production, fuel wood availability, and liberation and decision-making about water management) and the
increased drudgery for women who have traditional roles in repopulation of villages. This progressive village-by-village re-
gathering water, fuel wood and fodder. Village abandonment, generation of catchments has not been uniform or entirely free of
particularly by the young seeking greater opportunities in bur- challenge, particularly from state and central government that
geoning cities, has been observed across Rajasthan as also other often regards it in strict legislative terms as illegal given the cen-
drier areas of India experiencing similar cycles of degradation of tralised control of water management. However, the TBS pro-
the linked socio-ecological system. gramme has attracted national and international recognition for
To counter this pervasive degenerative cycle, a programme led contributing to drought resilience and the reversal of pre-inter-
by the NGO Tarun Bharat Sangh (TBS) has been active since 1985, vention cycles of aridication, ecosystem degradation, human

Table 1
National and State control of water management.

Administrative level in India Institutions responsible for water management

National level Three primary institutions are involved in the management of water resources:
(1) The Ministry of Water Resources (MWR) that has responsibility for enacting policy guidelines, development programmes and reg-
ulation of national water resources;
(2) The Planning Commission that allocates nancial resources to the states and the MWR to support water resource development
programmes; and
(3) the Ministry of Agriculture that promotes irrigated agriculture
In addition to these three primary institutions, the Central Groundwater Authority was established in 1996 to control and regulate
groundwater development
State level Irrigation and Water Supply Departments. However, not all states have their own groundwater authorities, and those that do in most
cases suffer from understafng, a lack of capacity, and often they have an out-dated mandate that prioritises the surveying and de-
velopment of the resource ahead of resource management (Food and Agriculture Organisation, 2010)
 T. Davies et al. / Ecosystem Services 21 (2016) 2030 23

Fig. 1. Location of the Arvari, Sarsa and Baghani catchments in north Rajasthan (from Everard (2015)).

impoverishment and rural depopulation. et al., 2011).

Everard (2015) examined the characteristics of eleven WHSs The red, Near-infrared (Near-IR, also known as Longwave-IR)
and their associated water bodies of varying design, history, size, and Middle-IR (also referred to as Shortwave-IR) are the most
altitude, purpose and predominant uses in the Arvari, Sarsa and useful parts of the electromagnetic spectrum for vegetation ana-
Baghani catchments. He assessed their contributions to restoration lysis (see Box 2).
of ecosystem service enhancements supporting the livelihoods of
local people. Everard (2016) reported upon permanent water- 3.1. Data acquisition and preparation
bodies in the three catchments, all of which had been dry outside
of monsoon rainfall for as much as four decades prior to restora- The suitability of Landsat data for catchment-scale vegetation
tion activities, and assessed their potential to restore traditional analysis underpinning this project is established in the introduc-
medicinal uses, local spiritual and cultural meanings, and other tion. Remotely sensed images were obtained from data holdings
primarily non-utilitarian and non-market ecosystem services as- accessible using the United States Geological Survey's Earth-
sociated with catchment regeneration. Explorer website (United States Geological Survey, undated). The
Forest cover has increased in areas now managed by commu- orientation of the three catchments of central interest (the Arvari,
nities, the area under forest in Thanagazi tehsil increasing from Sarsa and Baghani) and adjacent catchments with restored upper
8.4% in 1989/90 to 14.37% by 1998/99 attributed largely to local reaches (Jahajwali and Ruparel) to this Landsat image are illu-
reforestation, forest management and grazing controls (Rathore, strated in Fig. 2.
2003). There are also increases in the extent of agricultural area Only images captured between March and May (or the nearest
enabled by improved soil moisture and accessible groundwater, available date) were used as this period is at the crossover be-
rising from 42% in 1989/90 to 54.9% by 1998/99 (Rathore, 2003), tween winter and summer and precedes the start of the monsoon
reecting a locally signicant change in land cover. Changes in period in Rajasthan (the monsoon rains fall from July to Septem-
vegetative cover in response to recharge of groundwater and other ber). It was anticipated that this would provide the strongest
interventions are therefore a surrogate both for groundwater le- evidence for vegetation change not associated with potentially
vels and a range of linked ecosystem services, including many for confounding seasonal or event-driven antecedent moisture
which the role of biodiversity is as yet uncertain but is generally conditions.
perceived as signicant (Science for Environment Policy, 2015). Landsat 5 and the more modern Landsat 8 collect images in
different wavelength bands (see Table 2). The spectral bands of
primary interest Red, Near-IR and Middle-IR are equivalent
3. Methodology across the two Landsat missions. Combining images from these
two Landsat missions into a single time series for analysis does not
The changes in the local landcover described above should be introduce signicant uncertainty (United States Geological Survey,
discernible on satellite images of the area. Research presented in 2013).
the introduction has established the potential utility of vegetation The United States Geological Survey (USGS) does not hold data
as a proxy for groundwater expression. We assessed the utility of recorded by Landsat missions for the rst 12 years of TBS-led ac-
three RS approaches to identifying the areas where vegetation has tivities. Suitable, available satellite imagery for the project com-
been reported as increasing as a result of water harvesting in the prised ten images (1997; 1998; 1999; 2000; 2009; 2010; 2011;
case study area. Established multispectral image analysis techni- 2013; 2014 and 2015).
ques were used to distinguish vegetation from other land cover The downloaded les were composited into combined multi-
types and also to assess plant vigour/condition and stress (Oritz spectral images. Subset images were created from the composite
24 T. Davies et al. / Ecosystem Services 21 (2016) 2030

Box 2Key spectral elements of RS vegetation assessment.

 Reflection of red EMR from the Earths surface is determined primarily by photosynthetic activity (Gates et al., 1965), so
measurements across this wavelength range ( 630700 nm) are useful for distinguishing between vegetation and bare soils.

 The infrared (IR) EMR spectrum is over 100 times broader than that of the visible range, and is typically divided into two basic
categories: Reflected-IR and Thermal-IR. Reflected-IR is most useful for interpretation of vegetation cover and condition due to the
characteristics of leaf structure and reflectance related to the individual roles that different layers of leaves play in leaf function:

Leaf reflectance at Longwave-IR/Near-IR wavelengths ( 0.71.0 to 5 mm wavelength) is primarily affected by the structure of the
leaf, which is in turn substantially influenced by moisture content. Many characteristics of leaf structure can contribute to the
reflection of Near-Infra Red radiation. Inside the leaf, light is scattered at cell interfaces and intercellular air spaces, due to a large
change in the refractive index (Slaton et al., 2001). Longer, more cylindrical palisade mesophyll cells propagate the visible
wavelengths deeper into the leaf interior, whereas more spherical spongy mesophyll cells tend to scatter the radiation (Vogelmann
and Martin, 1993). Near-IR is therefore useful for detecting plant age and stress. For healthy plants, reflection is high as very little
energy is absorbed from Near-IR wavelengths. However, as plants experience stress due to the limited availability of water or other
causes, the range of wavelengths absorbed at the red end of the visible spectrum has been observed to broaden with greater
absorption of longer wavelengths including into the Near-IR, a phenomenon terms red shift (National Learning Network for
Remote Sensing, 1999).
Reflectance of wavelengths of Shortwave-IR/Middle-IR ( 5 to 2540) is more sensitive to moisture, and can therefore be used to
monitor vegetation and moisture content (Geospatial Innovation Facility, 2008). This may make it possible to discern different crop
species, with differences most evident in the middle of the growing season (Goward, 1985; CCMEO, 2014; Federation of American
Scientists, undated). In conjunction with Near-IR reflectance, Middle-IR can be used in the detection of plant stress as the ratio of
reflectance of these two bands changes with leaf moisture levels (Federation of American Scientists, undated).

images, a process that simplies analysis by selecting the spectral increase in the change analysis. The change matrix used for in-
bands relevant to the project (Wallin, 2006), and extracting data terpreting changes in land cover between different dates is pre-
based on an area of interest (AOI). The dened AOI was 92 km sented in Table 3.
(eastwest) by 94 km (northsouth) covering 8648 km2, with the The NDVI (Normalised Difference Vegetation Index) was used
city of Jaipur in the bottom left corner and Alwar in the top right, to assess vegetation cover and vigour. Vegetation Indices (VI) are
highlighted in red in Fig. 2. The three principal catchments of in- mathematical transformations intended to enhance the vegetation
terest (the Arvari, Sarsa and Baghani) and the upper reaches of signal while minimising solar irradiance and soil background ef-
adjacent restored catchments (the Jahajwali and Ruparel) are
fects, designed to assess the spectral contribution of vegetation to
covered by the AOI, which also includes areas to the west and
multispectral observations (Jackson and Huete, 1991; Elvidge and
south of these catchments. However, as the ve fully or partially
Chen, 1995). Vegetation is highly reective in the Near-IR and
restored catchments cover approximately 50% of the selected
highly absorptive in the visible red parts of the spectrum. The ratio
92 km
94 km area, and TBS-led restoration has occurred locally
in adjacent catchments, it was decided that the area was suitable between these spectral bands are used as an indicator of the status
for time series analysis as it should demonstrate overall change. of vegetation, correlating with the photosynthetic activity of ve-
getation (Xie et al., 2008). As photosynthetic activity increases,
3.2. Data analysis NDVI is expected to increase. Jamali et al. (2011) found that ve-
getation growth indicators such as NDVI are highly correlated with
Researchers involved in change detection using remotely soil moisture, making NDVI a good proxy indicator for soil
sensed data have developed a broad range of methodologies for moisture and therefore groundwater in areas with shallow
identifying any changes to the environment (Mas, 1999). For this groundwater. The NDVI is calculated using the equation:
study, three established, analytical methods were used to interpret
(NearIR Red)
the images and assess changes in land cover, plant vigour and
moisture stress between 1984 and 2015. (NearIR + Red)
The ISODATA (Iterative Self Organising Data Analysis Techni- The Moisture Stress Index (MSI) measure leaf water stress. It is
que) classication and change analysis method was used to assess a reectance measurement that is sensitive to increasing leaf
landcover changes between each time step and over the whole water content, particularly as an indicator of low leaf water con-
period. ISODATA classies pixels into spectral classes using a
tent (Ceccato et al., 2001; Hunt and Rock, 1989). As the leaf water
clustering algorithm that is applied iteratively until the con-
content in vegetation increases, the strength of absorption of
vergence threshold (change in pixel allocation between iterations
Middle-IR wavelengths increases; conversely, absorption of Near-
is o95%) has been achieved (Xie et al., 2008). Meaningful in-
IR wavelengths is largely unaffected by changing water content so
formational classes are then assigned to the derived spectral
classes. In this project, these were disturbed vegetation, irrigated serves as a reference. In areas where access to groundwater is a
vegetation, bare earth and water. Once all images had been clas- key inhibitor of vegetation growth, the MSI would be expected to
sied, landcover classes from consecutive dates were compared decrease as leaf water stress decreases, which suggests MSI can
based on a matrix describing land cover change. For example, an serve as a proxy indicator for groundwater availability (Chen and
area classied as bare earth in for one observation but as vegeta- Hu, 2004; Eagleson, 1978; Harris et al., 2005). MSI is calculated
tion in the next observation would be classied as vegetation using the equation:
 T. Davies et al. / Ecosystem Services 21 (2016) 2030 25

Fig. 2. Locations of the Arvari, Sarsa and Baghani catchments and upper Jahajwali and Ruparel catchments overlaid on Landsat image PATH 147, ROW 41 and showing the
area of interest in red. (False colour composite LS5 543: greens represent vegetation and pinks and browns bare earth.) Dotted lines indicate catchment boundaries. (For
interpretation of the references to colour in this gure legend, the reader is referred to the web version of this article.)

Table 2 variability, over the period of analysis, is examined rst as these

Landsat 5 and 8 spectral bands equivalences (adapted from United States Geological results inform the interpretation of the analytical results.
Survey (2013)).

Landsat 5 Bands Landsat 8 Bands 4.1. Local climatic variability

Band Wavelength Band The weather in Rajasthan over the period of this study is very
Band 1 Blue 0.450.52 0.4520.512 Blue Band 2
similar year on year. The hottest month in each year in the time
Band 2 Green 0.520.6 0.530.59 Green Band 3
Band 3 Red 0.630.69 0.640.67 Red Band 4 series analysed was May, except for 2014 in which June was the
Band 4 Near-IR 0.770.90 0.850.88 Near-IR Band 5 hottest month (Table 4). The highest average temperature per year
Band 5 Shortwave IR 1.551.75 1.571.65 Shortwave-IR1 Band 6 ranged from a low of 37 C in May 1997, with the highest being
Band 6 Thermal IR 10.412.50 10.6011.19 Thermal-IR1 Band 10 42 C in May 2010. Many of the years had their driest period of the
11.512.51 Thermal-IR2 Band 11
Band 7 Shortwave IR 2.092.35 2.112.29 Shortwave-IR2 Band 7
year in the months just before the months chosen for the images,
generally covering a three-month period between November and
March (Weather Spark, 2015). Precipitation patterns across the
MiddleIR study period (data also from Weather Spark (2015)) are re-
NearIR presented in Fig. 3.
Most years had their driest period sometime between No-
vember and March (Weather Spark, 2015). Precipitation patterns
across the study period (Weather Spark, 2015) are represented in
4. Results Fig. 3.

Results from the three analytical methods ISODATA, NDVI and 4.2. The ISODATA classication and change analysis
MSI are presented below. The sensitivity to antecedent moisture
conditions of projects assessing groundwater using vegetation as a Table 5 presents the results from the landcover change analysis
proxy was highlighted in the introduction. Local climatic and quanties the change that occurred between image pairs.
26 T. Davies et al. / Ecosystem Services 21 (2016) 2030

Table 3
Change matrix showing the classes used to identity vegetation changes over time.

Image A (Date 1)

Bare Earth Irrigated vegetation Disturbed Vegetation Water Unknown Water/Shadow

Image B (Date 2) Bare Earth Unchanged BE Veg Loss Veg Loss Reclaimed Land Unknown Unknown
Irrigated Vegetation Veg Increase Unchanged IV Improved Veg Reclaimed and Improved Unknown Unknown
Disturbed Vegetation Veg Increase Veg Degradation Unchanged Disturbed Veg Reclaimed and Improved Unknown Unknown
Water Flooded Flooded Flooded Unchanged Water Unknown Unknown
Unknown Unknown Unknown Unknown Unknown Unknown Unknown
Water/Shadow Unknown Unknown Unknown Unknown Unknown Unknown

Table 4 183.1 km2 became degraded between 1997 and 2015, changing to
Alwar District annual weather records from Weather Spark (2015). bare earth from vegetation. 483 km2 of bare earth remained un-
changed between 1997 and 2015, though the area of land covered
Year Hottest Highest temperature Driest period
month (C) by bare earth uctuated throughout the time series with a peak of
898.7 km2 in 2009/2010 and a minimum area of 519.3 km2 re-
1997 May 37 1st January28th March corded in 2014/2015. Overall, these ISODATA results are consistent
1998 May 41 9th November31st with the vegetation increase reported previously, indicating an
December
1999 May 40 10th February8th May
increase in groundwater supporting its growth and its potential to
2000 May 40 1st January22nd March provide a range of enhanced ecosystem services.
2009 May 41 NovemberDecember Total area of irrigated vegetation remained unchanged at
2010 May 42 23rd February28th April 17.7 km2 between 1997 and 2015. However, there was signicant
2011 May 40 SeptemberDecember
year-on-year variability throughout this time series. The highest
2013 May 41 9th November31st
December amount of annual unchanged irrigated vegetation detected in
2014 June 41 28th October14th 1999/2000 (31.2 km2) represented an increase of 25.30 km2 on the
December previous year, but there was a substantial subsequent decline of
2015 May 41 30.29 km2 so that left only 0.96 km2 remaining irrigated vegeta-
tion in 2009/2010. This sharp decline could be accounted for by
the land being used for different purposes, or more likely by the
weather as 4 of the 5 months prior to the data being captured
were extremely dry (2009 experienced its driest period between
November and December whilst 2010 had its driest period be-
tween February and March). Subsequently, the area of irrigated
vegetation steadily rose to 7.1 km2 by 2014/2015. The amount of
unchanged disturbed vegetation varied throughout the 19972015
time series, but followed no pattern. Overall, 165 km2 of disturbed
vegetation remained unchanged between 1997 and 2015.
Although some water features were detectable, these were all
large water bodies such as lakes and reservoirs. It was not possible
to determine trends in extent in open surface water bodies due to
limitations imposed by the spatial resolution (30 m) and spectral
Fig. 3. Total annual precipitation in Alwar District (mm) (Weather Spark, 2015). resolution of the Landsat data within this survey, so the study
could not conrm observations recorded by Sinha et al. (2013) and
Increases in a particular change category from one date to the next Everard (2015).
are highlighted in green and decreases in area are represented in However, all of these analyses have to be interpreted with
red. caution as spectral confusion between some areas of water and
On the basis of previously reported observations of increases in shadows cast by the low-lying sun in Landsat 8 images leaves a
agricultural land and forestry by Rathore (2003) as well as those large area (42.16 km2) classied as unknown.
recorded by Sinha et al. (2013) and Everard (2015), bare earth
extent would be expected to decrease as areas used for agricultural 4.3. NDVI comparison of plant photosynthetic activity
production and tree cover increased. Vegetation increase (bare
earth transitioning to vegetative cover) did indeed rise as expected Increases in groundwater levels reported by Rathore (2003,
over the time series. Between 1997 and 2015, bare earth decreased 2005), Sinha et al. (2013) and Everard (2015) would be expected to
with 288.4 km2 becoming vegetated. The largest annual vegetation result in an increase in farmed and natural vegetation cover.
increase occurred between 2010 and 2011 when a total of Therefore, NDVI extent and value would be expected to increase
261.2 km2 became vegetated (an increase of 198.8 km2 between over the time series of this study. However, evidence presented in
years). This could be attributed to the later rains in 2010 and the Table 6 indicates that NDVI decreased by up to 20% over an area of
early rains in 2011 recorded by Weather Spark (2015). Vegetation 116.4 km2 between 1997 and 2015, with only 9.7 km2 showing an
increased again by 27.2 km2, rising to 288.4 km2 by 2011/2013, a increase between 0% and 20% in NDVI.
further increase that could again be accounted for by the increased However, inter-annual uctuation in NDVI across the time se-
amount of precipitation that fell in January and February 2013 quence was signicant. The largest single change happened be-
(117 mm fell over the two months, more than all the previous tween 2010 and 2011, when NDVI increased by 20% in 197.4 km2
years combined over the same period: Water Resources Rajasthan (NDVI values in 20% of pixels increased in comparison to the
(2015) and Weather Spark (2015)). By comparison, a total area of previous year) with no area showing a decrease of 20%. The largest
 T. Davies et al. / Ecosystem Services 21 (2016) 2030 27

Table 5
Land cover area change (km2) and total area change (km2). Red indicates a decrease in value whilst green indicates an increase.

decrease in NDVI happened between 1997/1998, where 164.9 km2

showed a 20% decrease, although 3.7 km2 did also show a 20%
increase that year. These years of inter-annual increase in NDVI
follow years after increased amounts of rainfall, as illustrated in
Fig. 4, consistent with a relationship between NDVI and pre-
cipitation observed by Jamali et al. (2011).

4.4. MSI comparison of plant water stress

It would be expected that, as groundwater levels increased, leaf

moisture would increase and therefore MSI would decrease. Data
presented in Table 7 demonstrate that MSI decreased by at least
20% over an area of 132.3 km2 between 1997 and 2015, whilst only
41.5 km2 showed an increase of 20% over the same period. Across
Fig. 4. Area changes in NDVI (km2) and total annual precipitation.
this time series, 1999/2000 experienced the smallest area of inter-
annual decrease (0.0009 km2) whilst 1998/1999 saw the largest
inter-annual decrease (295.6 km2), with areas of 0.6 km2 and 2009 and 2010, a small area (0.03 km2) experienced a 20% increase
26.4 km2 (20% decreases) in 1997/1998 and 2014/2015 respec- in MSI between these years with no decreases.
tively. There were also some incidences where MSI had increased It should also be added that the years that saw the most dra-
by at least 20%, although these covered very small areas (2.2 km2 matic decreases in MSI (1998/1999 and 2014/2015) were preceded
in 1997/1998, 0.06 km2 in 1998/1999, 1 km2 in 2000/2009, by increased precipitation (see Fig. 5). These years also experi-
0.03 km2 in 2009/2010 and 2.6 km2 in 2014/2015, with the largest enced some of the largest increases in NDVI. Since NDVI increases
increase of at least 20% between 2013/2014 of 121.9 km2). with plant growth, these increases in NDVI would be expected as
The largest inter-annual MSI decrease in 1998/1999, expected if indicator of increasing groundwater and soil moisture. This would
groundwater is recharged, could be explained by the high levels of be caused by capillary rise, meaning MSI should decrease (Harris
precipitation seen in 1998. The second-highest decrease (26.430) et al., 2005; Eagleson, 1978).
between 2014 and 2015 also followed increased precipitation.
However, although precipitation was on an increasing trend in

Table 6
Area (km2) of NDVI change at 20% threshold.

NDVI 97/98 98/99 99/00 00/09 09/10 10/11 11/13 13/14 14/15 97/15

20% Increase 3.776 80.268 4.505 6.335 0.023 197.446 0.000 0.242 170.729 9.750
20% decrease 164.869 0.710 14.846 1.173 87.636 0.000 148.990 7.410 0.049 116.454
28 T. Davies et al. / Ecosystem Services 21 (2016) 2030

Table 7
Area (km2) of MSI change at 20% threshold.

MSI change 97/98 98/99 99/00 00/09 09/10 10/11 11/13 13/14 14/15 97/15

20% increase 2.222 0.061 0.000 1.022 0.033 0.000 0.000 121.951 2.594 41.479
20% decrease 0.621 295.623 0.001 0.000 0.000 0.000 0.000 0.000 26.430 132.246

threshold than 20%, or a series of increasing thresholds, may reveal

a stronger pattern. Signicant inter-annual uctuations in NDVI
and MSI appeared to follow the amount of antecedent rainfall, at
least for the largest increases of NDVI and decreases of MSI, con-
sistent with a vegetative response to increasing soil moisture and
groundwater levels.
The results may support an inference of the impact of WHSs in
the case study area, based on vegetation uxes, but it is not un-
equivocal. However, it appears that the ISODATA-derived land-
cover change analysis does support reported revegetation over the
analytical time period. Year-on-year, rainfall-related uctuations in
NDVI and MSI may suggest a correlation between groundwater
and vegetative extent and condition, but the limitations of spatial
Fig. 5. Area changes in MSI (km2) and total annual precipitation. and spectral resolution in the available Landsat datasets meant
that it was not possible to isolate this from climatic drivers.
5. Discussion Uncertainties in the analytical outcomes are exacerbated by an
incomplete temporal dataset, as RS data for the initial 12 years of
This study has investigated whether remote sensing technology WHS activities in the case study area were not available. In some
can be used to assess the success of community-level groundwater images, a low sun angle resulted in large areas of shadows elim-
recharge initiatives in catchments in Alwar District, Rajasthan inating the ability to evaluate landcover change in these areas.
(India) by considering RS-derived trends in vegetative cover and Despite being the best available data, the Landsat spatial resolu-
health in the light of observed socio-ecological outcomes reported tion was too coarse to detect some elements of interest such as
by Rathore (2003), Sinha et al. (2013) and Everard (2015). small, shallow streams. The ISODATA algorithm as applied to
The ISODATA classication and change analysis detected an Landsat imagery failed to discern differences between some land
increase in vegetation cover between 1997 and 2015, although the cover classes.
method was unsuccessful in detecting evidence of river re- Equivocal though the conclusions drawn from analysis of lim-
juvenation due to the spatial and spectral resolution of data from ited data from Landsat missions may be, our research nevertheless
Landsat missions. The remotely sensed increase in vegetation proves the principle of an RS-based approach supporting the
cover of 254.8 km2 is generally consistent with catchment re- monitoring of indicators of likely ecosystem service changes over
generation reported by Rathore (2003), Sinha et al. (2013) and broad landscapes where other forms of consistent measurement of
Everard (2015). Between 1997 and 2015, the area of bare earth ecosystem service change may be difcult or unaffordable. It also
dropped steadily, consistent with Rathore (2003, 2005) observa- highlights the type of research necessary to provide a useful
tion that land area used for agriculture and forestry increased management support tool, based on ner-resolution satellite
between 1989 and 1998 with an anticipated increasing trend. images (the ESA Sentinel 2 mission with substantially improved
It was not possible to determine trends in the extent of surface spatial and spectral resolution may now meet some of this need)
water based on our research due to the spatial and spectral lim- and improved analytical methods. This type of RS-based analysis
itations of Landsat data. Although some water features were de- of proxy indicators may offer the best prospects of assessing long-
tectable, these were all large water bodies such as lakes and re- term trends and the efcacy of restoration measures, as a surro-
servoirs. There are two explanations for the inability detect gate for linked, water-dependent ecosystem services and their
smaller rivers in the study area. If the water level in rivers was potential to support livelihoods. It can also constitute a tool in-
very low it would not have inuenced the spectral response of the forming policy formulation, addressing extensive landscapes that
associated pixels and the rivers would be undetectable. Sediment are otherwise hard and/or expensive to monitor by other means.
in the rivers, due both to the nature of the catchment and uses This in turn can assist with identication of policy responses best
such as stock watering, may further complicate differentiating promoting sustainable water resource regeneration and manage-
surface water from bare earth. The only detected features enabling ment, providing indications of both the efcacy and the suitability
inference of the presence of narrow rivers were surrounding, ir- of measures supporting rather than undermining community-
rigated vegetation areas running along the banks. These features based institutional and physical land and water management
converge with larger rivers further downstream, outside of the practices that have a long heritage in India. Broad-scale insights
case study area, suggesting a topo-hydrological link. into trends in water and vegetation extent and condition in the
Results from the analysis of NDVI and MSI were less clear. NDVI landscape, of the kind possible through indicators derived from
would be expected to increase and MSI to decrease between 1997 remote sensing data, offer signicant potential for better-in-
and 2015. However, analyses in this study found that NDVI de- formed, more far-sighted policy and management decision-mak-
creased by at least 20% over an area of 116.4 km2 with only 9.7 km2 ing that better empowers local action within an overall policy
showing a 20% increase, though there was a general decrease in framework. These indicators can serve as a proxy for a wider suite
MSI consistent with general expectations. This may be partially of ecosystem services essential for human wellbeing and ecosys-
explained by considering that a 20% change in either index does tem resilience. Where the efcacy of restorative measures can be
not necessarily imply an increase or loss of vegetation cover, but a demonstrated by RS and other techniques, this can provide valu-
change in vigour or stress. A change analysis with a lower able input to reform India's currently fragmented water
 T. Davies et al. / Ecosystem Services 21 (2016) 2030 29

management policy environment. water management. Such overviews of water and land cover
Promotion of a transition from state-imposed regulation to- trends enabled through interpretation of remote sensing data can
wards an enabling framework that is synergistic with diverse local better inform decision-making, empowering local action attuned
culturally and environmentally appropriate solutions is important to the heterogeneity of both natural and human landscapes and
to the realisation of sustainable water management, as high-level recognising the many ecosystem service outcomes mediated by
policy has to be exible to account for substantial heterogeneity in water rather than merely treating it as a commodity.
both natural and human landscapes. This exibility is essential if
policies are to be effectively targeted, accepted and effective in
supporting the ecosystem service outcomes that people value, References
many of them mediated by water ows through landscapes, rather
than merely treating water as a commodity divorced from its 2030 Water Resources Group, 2009. Charting Our Water Future: Economic Fra-
myriad environmental functions. meworks to Inform Decision-making. 2030 Water Resources Group, McKinsey.
Becker, M.W., 2006. Potential for satellite remote sensing of ground water. Ground
Water 44 (2), 306318.
CCMEO, 2014. Tutorial: Fundamentals of Remote Sensing. Canada Centre for
6. Conclusions Mapping and Earth Observation. [Online] http://www.nrcan.gc.ca/earth-sci
ences/geomatics/satellite-imagery-air-photos/satellite-imagery-products/edu
cational-resources/9309, (accessed 6.02.16.).
Analysis of remotely sensed imagery can provide a basis for Ceccato, P., Flasse, S., Tarantola, S., Jacquemoud, S., Gregoire, J.-M., 2001. Detecting
assessment of trends in vegetation extent, vigour and stress over vegetation leaf water content using reectance in the optical domain. Remote
broad spatial scales and across time series. However, our analysis Sens. Environ. 77, 2233.
Central Ground Water Board, 2006. Dynamic Ground Water Resources of India.
of available, medium resolution remotely sensed data to corrobo- Central Ground Water Board, New Delhi.
rate reported natural and farmed landscape regeneration in Alwar Central Water Commission, 2013. Water and Related Statistics. [Online] http://
District as a result of WHSs produced equivocal results. Landcover www.cwc.nic.in/main/downloads/Water%20and%20Related%20Statistics-2013.
pdf, (accessed 6.02.16.).
change analysis generally correlated with reported observations in Chen, X., Hu, Q., 2004. Groundwater Inuences on soil moisture and surface eva-
vegetation cover and condition between 1997 and 2015, whereas poration. J. Hydrol. 297 (14), 285300.
results NDVI and MSI analysis drew less clear trends over the time Chenworth, J., 2008. Minimum water requirement for social and economic devel-
opment. Desalination 229, 254256.
series but calculated inter-annual uctuations were strongly cor- Chinnasamy, P., Maheshwari, B., Prathapar, S., 2015. Understanding groundwater
related with rainfall in the preceding year. This makes is difcult to storage changes and recharge in Rajasthan, India through remote sensing.
attribute landcover change to the WHSs implemented in the re- Water 7 (10), 55475565. http://dx.doi.org/10.3390/w7105547.
Eagleson, P.S., 1978. Climate, soil and vegetation: a simplied model of soil
gion, and conrms the difculties identied by Becker's (2006) moisture movement in the liquid phase. Water Resour. Res. 14 (5), 722730.
review. Edet, E.A., Okereke, S.C., Teme, C.S., Esu, O.E., 1998. Application of remote-sensing
Nevertheless, this research shows initial promise, highlighting data to groundwater exploration: a case study of the Cross River State, south-
eastern Nigeria. Hydrogeol. J. 6 (3), 394404. http://dx.doi.org/10.1007/
that further research based on ner-resolution satellite images and
s100400050162.
improved analytical methods may serve as a useful management Elvidge, C.D., Chen, Z., 1995. A soil adjusted vegetation index (SAVI). Remote Sens.
support tool. Issues related to resolution may be resolved by im- Environ. 25, 295309.
proved imagery. The ESA Sentinel 2 data, with spatial and spectral Everard, M., 2015. Community-based groundwater and ecosystem restoration in
semi-arid north Rajasthan (1): socio-economic progress and lessons for
resolution improvements, may be of particular interest in this groundwater-dependent areas. Ecosyst. Serv. 16, 125135.
regard. As a recently launched RS platform, it cannot provide data Everard, M., 2016. Community-based groundwater and ecosystem restoration in
for longitudinal studies such as this. However, a seasonally based semi-arid north Rajasthan (2): reviving cultural meaning and value. Ecosyst.
Serv.
intra-annual analysis, comparing the case study area with a similar Federation of American Scientists, undated. Supervised Classication. Federation of
area without WHSs would be a suitable test of the potential of American Scientists. [Online]. http://fas.org/irp/imint/docs/rst/Sect1/Sect1_17.
these new imagery products. html, (accessed 6.02.16.).
Federation of American Scientists, undated. Vegetation Applications: Agriculture,
In the absence of actual groundwater depth data, projects such Forestry and Ecology. General Principles for Recognizing Vegetation. Federation
as this unavoidably have to rely on proxy data, which will in- of American Scientists. [Online]. http://fas.org/irp/imint/docs/rst/Sect3/Sect3_
evitably be contestable. This suggests that collecting information 1.html, (accessed 6.02.16.).
Food and Agriculture Organisation, 2010. Deep Wells and Prudence: Towards
about groundwater depth in the case study area may be particu-
Pragmatic Action for Addressing Groundwater Overexploitation in India. Food
larly expedient in developing a case in support of WHSs. It will and Agriculture Organisation of the United Nations. [Online] http://www.fao.
also help make empirical links with the enhancement of a range of org/nr/water/apfarms/upload/PDF/world_bank_rep.pdf, (accessed 6.02.16.).
Gates, D.M., Keegan, H.J., Schleter, J.C., Weidner, V.R., 1965. Spectral properties of
linked ecosystem services, for which NVDI and other similar in-
plants. Appl. Opt. 4, 1120.
dices serve as indicators. George, K.V., Pillai, K.V., Mathew, J., 2015. Studies on water resource management:
Methodological improvements to the implemented analyses approaches and strategies. Int. J. Adv. Res. 3 (11), 781791.
may result in more robust evidence of the value of WHSs than we Geospatial Innovation Facility, 2008. Landsat Spectral Band Information. [On-
line] http://gif.berkeley.edu/documents/Landsat%20Band%20Information.pdf,
were able to produce. Exploring the utility of other landcover (accessed 6.02.16.).
derivation techniques, in particular, maximum likelihood su- Goward, S., 1985. shortwave infrared detection of vegetation. Adv. Space Res. 5 (5),
pervised classication techniques based on in-eld observation of 4150.
Harris, A., Bryant, R.G., Baird, A.J., 2005. Detecting near-surface moisture stress in
training areas may improve the utility of RS in this context. The Sphagnum spp. Remote Sens. Environ. 97, 371381.
particularly equivocal NDVI and MSI results may be a result of the Heilman, J.L., Moore, D.G., 1982. Evaluating depth to shallow groundwater using
20% threshold specication for change analysis. Modifying this, Heat-Capacity Mapping Mission (HCMM) data. Photogramm. Eng. Remote Sens.
48 (12), 19031906.
and relating the observed change to the initial vegetation condi- Hunt Jr., E., Rock, B., 1989. Detection of changes in leaf water content using near-
tion represented by the index value, may result in a more nuanced and middle-infrared reectances. Remote Sens. Environ. 33, 4354.
assessment or plant vigour and stress. Jackson, R.D., Huete, A.R., 1991. Interpreting vegetation indices. Prev. Vet. Med. 11,
185200.
Even with the noted limitations of the implemented methods,
Jayanti, G., 2009. 25 Years of Evolution: Restoring Life and Hope to a Barren Land.
the RS techniques implemented are potentially valuable for de- Tarun Bharat Sangh, Alwar.
tecting trends across broad landscapes, and could inform more Jin, X., Wan, L., Zhan, Y., Xue, Z., Yin, Y., 2007. A Study of the relationship between
insightful policy and management decisions that recognise the vegetation growth and groundwater in the Yinchuan Plain. Earth Sci. Front. 14
(3), 197203. http://dx.doi.org/10.1016/S1872-5791(07)60026-8.
important roles of localised, community-based water and land Lv, J., Wang, X., Zhou, Y., Qian, K., Wan, L., Eamus, D., Tao, Z., 2013. Groundwater-
management in addressing policy goals relating to sustainable dependent distribution of vegetation in Hailiutu River catchment, a semi-arid
30 T. Davies et al. / Ecosystem Services 21 (2016) 2030

region in China. Ecohydrology 6, 142149. http://dx.doi.org/10.1002/eco.1254. http://www.satimagingcorp.com/services/resources/characterization-of-sa

Jamali, S., Seaquist, J., Ardo, J., Eklundh, L., 2011. Investigating temporal relation- tellite-remote-sensing-systems/, (accessed 6.02.16.).
ships between rainfall, soil moisture and MODIS-derived NDVI and EVI for six Science for Environment Policy, 2015. Ecosystem Services and the Environment. In-
sites in Africa. In: Proceedings of the 34th International Symposium on Remote depth Report 11 produced for the Euro pean Commission, DG Environment by
Sensing of Environment. [Online] http://lup.lub.lu.se/record/4194995, (ac- the Science Communication Unit. University of the West of England, Bristol.
cessed 6.02.16.). http://ec.europa.eu/environment/integration/research/newsalert/pdf/ecosys
Kidwell, H., 2015. Recharging India's Aquifers. [Online] http://www.waterworld. tem_services_biodiversity_IR11_en.pdf, (accessed 6.02.16.).
com/articles/wwi/print/volume-25/issue-2/editorial-focus/groundwater-treat Singh, D.K., Singh, A.K., 2002a. Groundwater situation in India: problems and
ment/recharging-india-s.html, (accessed 6.02.16.). perspective. Int. J. Water Resour. Dev. 18, 563580.
Kppen, W., Wegener, A., 1924. Die Klimate der Geologischen Vorzeit. Gebr. Sinha, J., Sinha, M.K., Adapa, U.R., 2013. Flow River Rejuvenation in India: Impact
Borntraeger, Berlin, Stuttgart. of Tarun Bharat Sanghs Work. SIDA Decentralised Evaluation 2013:28. Swedish
Mas, J.F., 1999. Monitoring land-cover changes: a comparison of change detection International Development Cooperation Agency, Stockholm.
techniques. Int. J. Remote Sens. 20 (1), 139152. Slaton, M.R., Hunt Jr., R.E., Smith, W.K., 2001. Estimating near infra-red reectance
Meijerink, A.M.J., 1996. Remote sensing applications to hydrology: groundwater. from leaf structural characteristics. Am. J. Bot. 88 (2), 278284.
Hydrol. Sci. J. 41 (4), 549561. Tth, J., 1963. A theoretical analysis of groundwater ow in small drainage basins. J.
National Learning Network for Remote Sensing, 1999. Spectral Properties of Ve- Geophys. Res. 68, 47954812.
getation. [Online] http://www.nln.geos.ed.ac.uk/courses/english/ars/a2200/ United States Geological Survey, 2013. Frequently Asked Questions about the
a2200008.htm, (accessed 6.02.16.).
Landsat Missions. http://landsat.usgs.gov/best_spectral_bands_to_use.php,
NRCAN, 2012. Spatial Resolution, Pixel Size and Scale. Natural Resources Cana-
(accessed 6.02.16.).
da. http://www.nrcan.gc.ca/earth-sciences/geomatics/satellite-imagery-air-
United Nations, 2015a. Water for Life Decade. United Nations. [Online] (http://
photos/satellite-imagery-products/educational-resources/9407, (accessed
www.un.org/waterforlifedecade/water_and_sustainable_development.shtml,
6.02.16.).
(accessed 6.02.16.).
NRCAN, 2014. Temporal Resolution. Natural Resources Canada. [Online] http://
United Nations, 2015b. Resolution adopted by the General Assembly on 25 Sep-
www.nrcan.gc.ca/earth-sciences/geomatics/satellite-imagery-air-photos/sa
tember 2015: 70/1. Transforming our world: the 2030 Agenda for Sustainable
tellite-imagery-products/educational-resources/9365, (accessed 6.02.16.).
Development. United Nations General Assembly, New York. http://www.un.
Oritz, B., Shaw. J., Fulton, J., 2011. Basics of Crop Sensing. [Online] https://sites.aces.
edu/group/crops/precisionag/Publications/Basics%20of%20Crops%20Sensing% org/en/ga/search/view_doc.asp?symbol A/RES/70/1&Lang E, (accessed
20-%20Ext%20Pub%20ANR-1398.pdf, (accessed 6.02.16.). 6.02.16.).
Prez Hoyos, I.C., Krakauer, N.Y., Khanbilvardi, R., Armstrong, R.A., 2016. A review of Water Resources Rajasthan, 2015. Rainfall Data. Government of Rajasthan. [On-
advances in the identication and characterization of groundwater dependent line] http://waterresources.rajasthan.gov.in/Daily_Rainfall_Data/Rainfall_In
ecosystems using geospatial technologies. Geosciences 6 (17). http://dx.doi.org/ dex.htm, (accessed 6.02.16.).
10.3390/geosciences6020017. USAID, 2015. Water and Development Strategy. USAID. [Online] https://www.
Postel, S., 2015. India's Food Security Threatened by Groundwater Depletion. Na- usaid.gov/what-we-do/water-and-sanitation/water-and-development-strat
tional Geographic, 3rd February 2015. [Online] http://voices.nationalgeo egy, (accessed 6.02.16.).
graphic.com/2015/02/03/indias-food-security-threatened-by-groundwater-de United States Geological Survey, undated. EarthExplorer. United States Geological
pletion/, (accessed 6.02.16.). Survey. http://earthexplorer.usgs.gov/, (accessed 6.02.16.).
Rathore, M.S., 2003. Community Based Management of Groundwater Resources: A Vogelmann, T.C., Martin, G., 1993. The functional signicance of palisade tissue:
Case Study of Alwar River Basin. Institute of Development Studies, Jaipur. penetration of directional versus diffuse light. Plant Cell Environ. 16, 6572.
Rathore, M.S., 2005. Groundwater Exploration and Augmentation Efforts in Ra- Wallin, D., 2006. Creating a Subset of a Larger ERDAS Imagine File. [Online] http://
jasthan: A Review. Institute of Development Studies, Jaipur. www.wwu.edu/huxley/spatial/tut/ERDAS_subset.htm, (accessed 6.02.16.).
Rosenberry, D.O., Striegl, R.G., Hudson, D.C., 2000. Plants as indicators of focused Weather Spark, 2015. Historical Weather for India. Weather Spark. [Online] https://
ground water discharge to a northern Minnesota lake. Ground Water 38 (2), weatherspark.com/history/stations/India, (accessed 6.02.16.).
296303. Xie, Y., Sha, Z., Yu, M., 2008. Remote sensing imagery in vegetation mapping: a
Satellite Imaging Corporation, 2015. Satellite Remote Sensing Systems. [Online] review. J. Plant Ecol. 1 (1), 923.