You are on page 1of 14

Environmental Science and Pollution Research (2021) 28:38145–38158

https://doi.org/10.1007/s11356-021-13217-w

RESEARCH ARTICLE

An analytical approach for assessment of geographical variation


in ecosystem service intensity in Punjab, Pakistan
Mobeen Akhtar 1 & Yuanyuan Zhao 1,2,3 & Guanglei Gao 1,2

Received: 12 October 2020 / Accepted: 25 February 2021 / Published online: 16 March 2021
# The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature 2021

Abstract
Ecosystem services (ESs) are a diverse range of environmental properties that have significant effects on both landscape
characteristics and human well-being, an understanding of the spatial variation of which is required for effective ecosystem
management and conservation. In this study, we aimed to identify the geographical variation in ESs and examine the factors
impacting ESs in Punjab Province, Pakistan. We evaluated ecosystem service intensity (ESI) by using the benefit transfer method
and investigated the spatial heterogeneity of ESIs at the tehsil level by using an exploratory spatial data analysis (ESDA) model
and semivariance analysis. This analysis indicated that spatial distribution of ESI shows significant and optimal spatial hetero-
geneity of ESs, revealing several hot and cold spots for ESI transformation in the study area. The ESI decreased from east to west
and north to south, with a trough in the metropolitan zones, including Lahore, Sheikhupura, Faisalabad, and Multan, which was in
accordance with the land use/cover pattern. Furthermore, using geographically weighted regression (GWR) analysis, we also
identified geographical clusters within which land use/cover pressure and population growth were significantly positively
associated with ESI across the province, based on a lag distance scale. The findings of this study have functional economic
implications with respect to highlighting areas worthy of conservation and designing ecological corridors to enhance the delivery
of ESs for regional benefits.

Keywords Ecosystem service . Land use/cover . Semivariance analysis . Spatial autocorrelation . Geographically weighted
regression

Introduction deteriorating or being utilized unsustainably (MEA 2005),


and during the first half of the twenty-first century, it is pre-
Ecosystem services (ESs), the benefits that humans derive dicted that such deterioration could escalate significantly
either directly or indirectly from the natural environment, are (Vihervaara et al. 2010; Hu et al. 2015; Zhai et al. 2020). In
essential for human well-being, livelihood, and survival recent years, a growing number of ES studies have been car-
(Costanza et al. 1997; MEA 2005; De Groot et al. 2012; Xie ried out in conjunction with ecosystem degradation and trans-
et al. 2017). Approximately 60% of ESs are currently formation, based on their ability to represent and communi-
cate with human–environmental interactions (Busch et al.
Responsible Editor: Philippe Garrigues
2012; Hu et al. 2015; Huang et al. 2020). The services pro-
vided by ecosystems are neither homogeneous nor consistent
* Yuanyuan Zhao throughout landscapes, often being characterized by spatial
yuanyuan0402@bjfu.edu.cn variation (Fisher et al. 2009; Barbier 2012; Wu et al. 2013).
To date, research on ESs has tended to focus on specific eco-
1
Present address: Key Laboratory of State Forestry Administration on
system types, service categories, and geographical areas
Soil and Water Conservation, Beijing Forestry University, (MEA 2005; Syrbe and Walz 2012; Huang et al. 2019; Qi
Beijing 100083, China et al. 2020; Sun et al. 2020). However, the spatial associations
2
Yanchi Research Station, School of Soil and Water Conservation, of ESs are frequently assumed rather than being practically
Beijing Forestry University, Beijing 100083, China demonstrated, and consequently, our current understanding
3
School of Soil and Water Conservation, Beijing Forestry University, regarding the spatial variation in ESs is comparatively limited.
Beijing 100083, China This deficiency in awareness accordingly limits our ability to
38146 Environ Sci Pollut Res (2021) 28:38145–38158

undertake the appropriate conservation measures and manage- appropriate description of the spatial patterns and scales of
ment of ecosystem functioning (Troy and Wilson 2006; ESs may be obtained using a combination of these two
Turner et al. 2007; Hu et al. 2015). A consideration of both methods. Numerous studies have examined the spatial prop-
the spatial homogeneity and heterogeneity of ESs is thus es- erties of process-related landscape modules, such as those of
sential for the development of effective policies to ensure watercourses, natural habitats, and specific ecosystems (Isbell
maintenance of integrity and rational utilization of these et al. 2011; Barbier 2012; Trabucchi et al. 2014). Sometimes,
services. affected regions are frequently imprecisely designated owing
In terms of examining the spatial variation in ESs, the to contradictions associated with differences between the spa-
quantification of ES represents a vital step, and in this regard, tial scale of the method of scrutiny and the scale at which the
a group of four approaches have previously been reported, census variables are calculated and evaluated (Anselin 2001;
namely, revealed preference approaches (e.g., market price Syrbe and Walz 2012). In this regard, identifying the drivers
and travel cost), stated preference approaches (e.g., contingent of ES variability and evaluating the constant variability range
valuation and choice experiments), cost-based approaches can contribute to optimizing locations for different objectives
(e.g., avoiding cost and replacement cost), and the benefit (Pijanowski et al. 2010; Su C-h et al. 2012). In exploratory
transfer method (Costanza et al. 1997; Talberth 2015; studies, the spatial associations between ESs and their causes
Arowolo et al. 2018). Of these approaches, the benefit transfer within administrative units can thus have significant conse-
method, a secondary valuation method, adapts previously quences for appropriate policy-making and implementation.
established estimates from (original) primary valuation studies They are also of scientific value, in that spatial flow strategies
in one or more locations to other areas assumed to have related are dependent on the study of spatial correlation at the admin-
demographic, economic, and ecological characteristics istrative unit level (Plummer 2009; Serna-Chavez et al. 2014).
(Richardson et al. 2015; Yi et al. 2017). Most notable in this Punjab, the second largest province in Pakistan, is a region
regard is the pioneering study conducted by Costanza et al. characterized by a distinctive ecosystem that has developed
(1997), who used the benefit transfer method to extrapolate under the influence of a unique landscape, environment, and
the global economic value of 17 ESs provided by 16 main other natural factors that are significant to the ecological sta-
biomes. Subsequently, the estimates were updated based on bility of Pakistan (Shedayi et al. 2019). Over the past few
a larger database of more than 300 case studies worldwide (de decades, Punjab province has undergone accelerated urbani-
Groot et al. 2012; Costanza et al. 2014). For example, a rela- zation concomitant with rapid population growth and eco-
tively high value was assigned to the assessment of ecosystem nomic development (Imran et al. 2019), and these develop-
services with respect to the association between ecosystem ments have predictably led to extensive changes in land use
conditions and economic development level in China and and cover, including the conversion of cropland to built-up
Southern Asia (Xie et al. 2008, 2017). areas, and a decline in the area covered by forest (FAO
Scientists have continuously been seeking to develop pre- 2015). In the latter regard, Samie et al. (2020) found that only
cise spatial decision-making methods for managing ecosys- 3% of the total land surface area of Punjab is now forested.
tem conservation strategies (Nelson et al. 2009). In this regard, The conversion of forest to barren land or built-up areas, de-
Zawadzki et al. (2005) have illustrated how ecosystem velopment of former cropland, water body shrinkage, and
service-based activities and management can proactively pri- substantial expansion in the area of built-up land have contrib-
oritize through mapping of the biophysical and social charac- uted to declines in both environmental and ecological sustain-
teristics of ESs. However, there remains a general lack of ability. These changes have alerted the national government to
consensus regarding the spatial variation in ESs and the spatial the necessity to maintain the viability of the region’s environ-
associations between observed monitoring units and scales. mental and ecological resources. However, beyond analyzing
Certain spatial statistical techniques, particularly exploratory geographical variations in Punjab, quantitative assessments of
spatial autocorrelation (ESDA) and semivariance analysis, are the impacts of these changes on ES evaluation have seldom
currently considered beneficial in terms of evaluating spatial been attempted. Indeed, to date, there have been no assess-
characteristics (Webster and Oliver 1990; Zawadzki et al. ments of these effects at a provisional scale, and only a few
2009; Zawadzki and Fabijańczyk 2013). ESDA is a collection related studies have been conducted at the local scale (Akhtar
of spatial statistical techniques that can be applied to visualize et al. 2020). Such quantitative analyses are, nevertheless,
spatial aggregation and anomalies and reveal a framework for needed on a priority basis, particularly in those areas in which
the assessment of ESs (McMillen 2010). Semivariance analy- escalating population growth and the increasing impact of
sis is typically employed for two main purposes, namely, human activities have resulted in substantial changes in eco-
quantifying the scale of variability shown by natural resource logical sustainability.
distribution patterns and identifying the spatiotemporal scale To this end, in the present study, we used a benefit transfer
at which verified variables show maximum variability method to evaluate ESs and subsequently assessed the ecosys-
(Wallace et al. 2000; He et al. 2007). Consequently, an tem service intensity (ESI) of each administrative unit in
Environ Sci Pollut Res (2021) 28:38145–38158 38147

Punjab. Furthermore, we used ESDA and semivariance anal- agricultural land surrounded by river basins, with the Potohar
yses to examine the geographical variations in ESI. Plateau in the north, hilly landscapes in the northwest and south-
Specifically, we sought to answer the following scientific west, and sandy desert in southeast Cholistan region. The climate
questions: (1) What are the characteristics of the spatial het- is generally dry, with precipitation varying from 100 mm in the
erogeneity of ESs in Punjab? (2) What is the relationship south to 600 mm in the northwest, although sometimes reaching
between ESI and driving factors? (3) What lessons can we 1000 mm along the province’s northeastern border.
learn with respect to prioritizing areas for development and Approximately three-quarters of the mean annual precipitation
conservation and designing ecological corridors to enhance is brought by monsoon rains, which typically last from June to
the distribution of ESs for the benefit of particular areas? September, leading to varying degrees of flooding. Temperatures
can reach as low as − 2°C in winter and as high as 47°C in
summer. Punjab has the largest irrigation system in Pakistan,
Materials and methods comprising more than 3000 irrigation channels (canals), which
directly support agricultural production (e.g., vegetables, fruits,
Study area wheat, rice, cotton, and sugarcane) and indirectly contribute to
industrial production (e.g., textiles, cement, sugar mills, and ag-
This study was conducted in the province of Punjab, which is riculture machinery) (FAO 2015; Samie et al. 2017).
located along the northwest frontier of the Indian Plate in South With a total population of 110 million, Punjab is the most
Asia (Fig. 1). The province comprises nine divisions, 36 districts, populace province in Pakistan, accounting for 52.92% of the
and 137 tehsils, covering a total area of 205,344 km 2 national population (PBS PBoS 2017). Between 1951 and 1998
(representing 25.8% of the total area of Pakistan). The topogra- and 1998 and 2017, the population of the province increased by
phy of the region is mostly plains, much of which is covered by 258% and 57%, respectively, with a corresponding increase in

^
±
^
^ ^
^
^
^

^ ^

^
^

Fig. 1 A map showing the location of the study area


38148 Environ Sci Pollut Res (2021) 28:38145–38158

population density from 100 to 535 individuals per km2 be- 1 km × 1 km projection. The land use/cover dataset based on
tween 1951 and 2017 (PBS PBoS 2017). Agriculture produc- 13 categories was re-categorized into six categories (Table 1).
tion contributes 76% to the annual food grain requirements Administrative boundary data at the tehsil level were obtained
[Bureau of Statistics, Punjab (http://bos.gop.pk/)]. Although from The Urban Unit (https://www.urbanunit.gov.pk/).
agriculture is the major economy of Punjab, contributing 60% Demographic and economic data, including indirect
to the total agricultural production fueling the country’s GDP, variables for the year 2019 at the tehsil level, were acquired
the industrial sector also makes a substantial contribution. The from the Pakistan Bureau of Statistics (http://www.pbs.gov.
province makes significant financial contributions to the pk/). The land use/cover, economic, and demographic datasets
economy of the country, and despite the increasing rate of were combined with administrative vector data employing the
urbanization, most of the population are still employed in ESRI ArcGIS 10.5 platform for subsequent assessment of the
agricultural activities (Samie et al. 2020). In concert with the ecosystem services value (ESV) and spatial analysis.
ongoing rapid urbanization and economic growth, there have
been substantial changes in land use and cover, including mas-
Evaluation of ecosystem service intensity
sive deforestation and conversion of agricultural land to urban
areas, resulting in excessive exploitation of river waters, envi-
Using the ESV evaluation method proposed by Costanza et al.
ronmental quality degradation, and threats to food security
(1997), we initially calculated ESV. In this widely used meth-
(Usman et al. 2015; Samie et al. 2020). Such changes in land
od, the global biosphere is categorized into 16 types of
use and cover could have potentially severe consequences with
ecosystems and 17 types of ES. Previously, Xie et al. (2008)
respect to ESs, particularly in the more heavily populated north-
have evaluated Chinese ESVs by using the value coefficient
eastern region, southwestern mountainous areas, and southern
of Chinese terrestrial ESs, which are mainly based on the
region of the province.
global ecosystem weight factor (Xie et al. 2008). For our study
area, we allocated the values of ESs per unit area of each land
use/cover type for the corresponding weighting factors and
Data collection
standardized the values of ES to a relative scale (0–100 value
points) (Table 2).
A land use/cover dataset with a spatial resolution of approxi-
The standardized equivalent ESV weighting factor was de-
mately 30 m was obtained in 2020 from the Pakistan Space
termined using Eq. (1) as follows:
and Upper Atmosphere Research Commission (SUPARCO).
The Food and Agriculture Organization of the United Nations  
I−I min
(FAO), in consultation with SUPARCO and the Crop I norm ¼  100 ð1Þ
I max −I min
Reporting Services of Punjab, has analyzed, interpreted, and
validated land use/cover classes at a scale of 1:25,000 in where lnorm is the standardized equivalent ESV weighting fac-
SPOT’s high-resolution spatial satellite data (FAO 2015; tor per hectare for a defined land use/cover type, the score of
Samie et al. 2017). Both land use/cover data and driving factor which is between 0 and 100; I is the determined value of a
data were prepared at a standardized spatial resolution and a given land use/cover type of a specific ecosystem service; and

Table 1 Reclassification of
SUPARCO classification SUPARCO code SUPARCO land/use/cover classification New code Reclassified land
categories use/cover categories

1 Orchard 2 Cropland
2 Crop irrigated 2
3 Crop marginal and irrigated saline 2
4 Crop in flood plain 2
5 Crop rainfed 2
6 Forest 3 Forest
7 Natural vegetation in west areas 4 Grassland
8 Range land-natural Shrubs and herbs 4
9 Built-up area 1 Built-up
10 Bare areas 5 Barren land
11 Bare area with sparse natural vegetation 5
12 Wet areas 6 Water body
13 Snow and glaciers
Environ Sci Pollut Res (2021) 28:38145–38158 38149

Table 2 Standardized equivalent weighting factor of ecosystem service value per hectare for different land types

Food Raw material Gas Climate Hydrological Waste Soil formation and Biodiversity and Providing
production production regulation regulation regulation treatment conservation maintenance aesthetic value

Built-up 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
Forest land 33.00 100.00 100.00 100.00 21.79 11.58 100.00 100.00 46.85
Grassland 43.00 12.08 34.72 38.33 8.10 8.89 55.72 41.46 19.59
Cropland 100.00 13.09 16.67 23.83 4.10 9.36 36.57 22.62 3.83
Water body 53.00 11.74 11.81 50.61 100.00 100.00 10.20 76.05 100.00
Barren land 2.00 1.34 1.39 3.19 0.37 1.75 4.23 8.87 5.41

Imax and Imin correspond to the maximum and minimum identifying hot and cold spots, and proposing spatial frame-
values for all land use/cover types for particular ESs. works (Anselin 1999; Zhang et al. 2011). In the present study,
The ESs and overall performance values for each tehsil we used ESDA to identify the occurrence of spatial ESI auto-
were calculated using Eqs. (2 and 3) as follows: correlation in Punjab. The research concentrated on particular
n
features of spatial clustering, such as the “local” spatial clus-
E i ¼ ∑ wk sik ð2Þ tering and the “global” distribution of ESI patterns.
k¼1
ESI can be measured using global Moran’s Ig based on Eq.
m
(6) as follows:
wk ¼ ∑ wkj =m ð3Þ
j¼1  
∑ ∑ wij ðxi −uÞ x j −u
i j
where Ei indicates the total score of tehsil i; wk represents the Ig ¼ N ! ð6Þ
ESV of land use/cover type k, that is, the cumulative average ∑ ∑ wij ∑ðxi −uÞ2
weight of all ES types wkj divided by the sum of ES types j (0– i j
100 value points), as shown in Eq. (3); and Sik indicates the
area of land use/cover type k in tehsil i. The weight wk indi- where wij is a contiguity matrix standardized in a row, xi and xj
cates the relative value of the specific land use/cover category are the ESIs for tehsil i and j, respectively, μ indicates the
for ESs, according to Table 2 as shown in Fig. 2. The overall average level of ESI, and N indicates the overall number of
performance of a tehsil can be measured in terms of ESs by tehsils in the study area. Moran’s I g indicates spatial
multiplying the score with the corresponding tehsil for each
land use/cover and adding all services.
ESI was subsequently evaluated to assess the spatial vari-
ation in ESs (Shi et al. 2012). ESI has been used to equate the
role of ESs in one tehsil with that in another tehsil and to
evaluate the spatial heterogeneity throughout the study area.
It represents the average value of ES in each tehsil, which can
be calculated using Eqs. (4 and 5) as follows:
ESI ¼ E i =S i ð4Þ
n
S i ¼ ∑ S ik ð5Þ
k¼1

where Si represents the overall area of tehsil I and Sik indicates


the area of land use/cover type k in tehsil i.

Measurement of the geographical variation in


ecosystem services

Identification of the hot and cold spots of ESI

ESDA is a collection of strategies for defining and visualizing Fig. 2 A bar graph showing the scores for each land use/cover type in
spatial distributions, exploring spatial association patterns, terms of ecosystem service intensity
38150 Environ Sci Pollut Res (2021) 28:38145–38158

autocorrelation and ranges from − 1 to + 1 (for negative to exhibited; z(xi) is the estimated sample value for tehsil xi;
positive spatial autocorrelation), respectively, and zero indi- z(xi + h) represents the sample value (i.e., ESI) for tehsil xi +
cates a lack of spatial autocorrelation (Anselin 2003). Given h; and N(h) represents the overall number of sample data pairs
that Moran’s Ig cannot be used to determine the hot and cold split by distance h.
spots between tehsils, local spatial relationships were deter- Semivariance functions are generally defined using the fol-
mined using the local indicator of spatial association (LISA) to lowing parameters: sill (C0 + C) defines the maximum
demonstrate the importance of hot and cold spots (Yang and semivariance level, the sum of the total variance described
Wong 2013). LISA was expressed using the local Moran’s Il by the Nugget effect and the spatial structure; range (A0) is
(Eq. (7). the distance where semivariance reaches a peak, or where
xi −μ   datasets are not correlated; the Nugget effect (C0) is the value
Il ¼ 2
∑ wij x j −u ð7Þ of the position at which spatially independent variance is rep-
∑ ðxi −uÞ j
i resented by the extrapolated relationship between two param-
eters intercepting the axis of the semivariance; the spatially
Moran’s Il, “local” Moran’s Il, and “global” Moran’s Ig were dependent structural variance (C) is the systematic variation
assessed using ESRI ArcGIS 10.0. When conducting analyses, level of the data, and the ratio of the Nugget effect (C0) to the
the spatial associations between tehsils were estimated using the sill (C0 + C), which might be an important predictive mea-
inverse distance weighting method (Lukaszyk 2004). The size of surement of variation, signifies the random components in the
the neighborhood specifies how many tehsils are included in this proportion of spatial variance (Cambardella et al. 1994; Krige
method, and the observation values were taken on a weighted 1996b; Treitz and Howarth 2000; Zawadzki et al. 2005).
average within this neighborhood. In this scenario, the neighbor- Spherical, linear, Gaussian, and exponential models were
hood size was automatically set using a default radius value, used to match the goodness of the experimental semivariance.
which is a Euclidean distance that ensures at least one neighbor- The model parameters were determined using GS+ software.
hood for each measurement. A cluster map was subsequently Semivariance models were cross-validated, with cross-
generated that incorporated information on the occurrence of validation root mean square error (RMSECV) and coefficient
local spatial trends. Generally, in terms of ESI, the map is pre- of determination (r2) being used to evaluate the goodness of
sented using spatial symbology representing five types of tehsil: the models (Davis and Sampson 2002).
(1) high–high (hotspots, positive spatial autocorrelation) indi-
cates tehsils with a high level of ESI correspondence to tehsils Examination of the relationship between ESV and driving
with high ESI rates; (2) low–low (cold spots, positive spatial factors
autocorrelation) indicates tehsils with a lower level of ESI, which
corresponds to tehsils with low ESI rates; (3) low–high (negative Generally, land use/cover changes and population growth are
spatial autocorrelation) indicates tehsils with a low ESI level that key factors affecting ESs (MEA 2005; Vihervaara et al. 2010).
corresponds to tehsils with high ESI rates; (4) high–low (nega- In the present study, we used population density coefficients
tive spatial autocorrelation) indicates tehsils with a high level of to reflect population growth (Fig. 3). The proportions of six
ESI that corresponds to tehsils with low ESI rates; and (5) not significant types of land use/cover (cropland, forestland,
significant, indicating tehsils with no spatial autocorrelation. grassland, water bodies, built-up land, and unused land) were
used to characterize the land use/cover pressures. Values ob-
tained for these variables are listed in Table 3.
Spatial heterogeneity of ESI The ordinary least squares (OLS) method was initially im-
plemented to determine ESI-related factors. The OLS method
Semivariance, a fundamental geostatistical method, was used is based on the assumption that anywhere within a specified
to calculate the spatial consistency of adjacent measurements geographic range, the association between independent and
(Zawadzki et al. 2005). The exploratory semivariance value dependent variables remains the same. Thus, using this ap-
for a variable is extracted from the calculation of half a mean proach, it is not possible to define the spatial non-stationarity
square difference for each data pair split by a particular lag of the ESI distribution. Consequently, the OLS approach pa-
distance of h (Krige 1996a; Chen and Feng 2013). rameter estimates are probably one-sided and invalid. We then
The standard semivariance is determined using Eq. (8) as used a geographically weighted regression model (GWR) that
follows: measures the occurrence of spatial interaction between vari-
1 ables at different locations (Fotheringham et al. 2002). The
γ ð hÞ ¼ ½zðxi Þ−zðxi þ hÞ2 ð8Þ regression model for the OLS is defined using Eq. (9) as
2N ðhÞ
follows:
where γ(h) is the estimated semivariance value at distance
yi ¼ β0 þ ∑β j xi þ ei ð9Þ
interval h, representing the degree of autocorrelation
Environ Sci Pollut Res (2021) 28:38145–38158 38151

Table 3 Descriptive statistics of


population density and land use/ Variables Mean Standard deviation
cover variables
ESI score (ha) 20.68 6.39
Indirect variables Population density (PD) (individuals/ha) 8.30 11.19
Direct Variables Proportion of cropland (%) 74.52 30.03
Proportion of forest land (%) 0.56 1.74
Proportion of grassland (%) 0.39 0.96
Proportion of water bodies (%) 0.77 0.76
Proportion of built-up land (%) 1.72 4.72
Proportion of unused land (%) 22.04 29.82

where yi represents the ESI value for the ith tehsil and xi is the The determined GWR coefficients are defined using Eq.
determining factor of the ith tehsil. β0 is constant, βj is the (11) as follows:
estimating determinant coefficient, and ei is the stochastic er-  −1 0
ror term. The GWR model is an updated version of Eq. (9), as bðui ; vi Þ ¼ X 0 wðui ; vi ÞX
β X wðui ; vi ÞY ð11Þ
determined using Eq. (10) as follows:
yi ðui ; vi Þ ¼ β0 ðui ; vi Þ þ ∑β j ðui ; vi Þxi þ ei ðui ; vi Þ ð10Þ where w(ui,vi) is a diagonal spatial weight matrix that is
specific for any observation (i.e., tehsil). The model is
where (ui,vi) describes an observation’s spatial position or based on the spatial weight matrix elements, which are
geographical coordinates (i.e., tehsils), with v i and u i dependent on the spatial position, and the bandwidth is
representing the latitudes and longitudes of the ith tehsil’s used to define the spatial location to analyze the non-
centroids, respectively. The GWR model performs well within stationary features (Poudyal et al. 2012; Yang and Wong
the observations grouped near a given tehsil and thus provides 2013). In the present study, both the GWR and OLS
a separate estimate of the parameters for each tehsil. models were run using GS+ software. We used a
Furthermore, to measure parameters individually for each ob- Gaussian function for GWR models to allocate a spatial
servation, GWR assigns a higher weight to observations (i.e., weight matrix and a cross-validation method to determine
tehsils) closer to the ith tehsil than to those farther away the optimum bandwidth required to reduce the Akaike in-
(Fotheringham et al. 2002). formation criterion (AICc).

Fig. 3 Spatial distribution of the population density and land use/cover of Punjab
38152 Environ Sci Pollut Res (2021) 28:38145–38158

Results Lahore, the capital of the province (17.91), whereas areas with
a low ESI level included Jampur (13.84 score/ha), followed by
Spatial distribution of ESI Multan (12.83 score/ha), Darya Khan (11.60), and Rojhan
(10.68). In contrast, ESI values for the northwestern, south-
We found that in Punjab, the ESI showed spatial heterogene- western, and southeastern parts of the region were found to be
ity, characterized by a gradual reduction from east to west extremely low, with an ESI level of (8.13) for Chaubara,
(Fig. 4). Notably, ESI levels in the extensive northeastern followed by Chakwal (6.70), Noorpur (5.35), Gujranwala
and central regions were extremely high for tehsil Hazro (5.36), and Kallar Kahar (3.47).
(26.97 Â score/ha), followed by Layyah (26.00), Pakpattan
(25.56), Vehari (25.31), and Faisalabad (24.68), with high ESI hot and cold spots
ESI levels being detected for Jehlum (23.00), followed by
Mianwali (22.40), Rawalpindi (21.67), Isakhel (20.43), and The statistical value of Moran’s Ig was 0.419, which was sta-
Chishtian (19.68). Tehsils in the northwestern and central re- tistically significant at the 0.05 level, showing a tendency
gions were found to be characterized by moderate ESI levels toward neighboring tehsils with low or high ESI clustering.
for Rahim Yar Khan (19.21), followed by metropolitan A total of 46 tehsils were found to be characterized by positive

Hazro
Kahuta
±
Chakwal
Kallar Kahar Jehlum
Mianwali

Gujranwala

Darya KhanNoorpur
Lahore
Faisalabad

Taunsa Chaubara
Layyah

D G Khan Kot Addu


Pak Pattan
Multan Burewala

Chishtian
Jampur

Ecosystem service intensity (score/ha)


Rojhan Yazman
Extremely low (3.47 - 8.37)
Rahim Yar Khan
Low (8.37 - 14.51)
Moderate (14.5 - 19.68)
High (19.68 - 23.70)
0 50 100 200 Km Extremely high (23.70 - 26.97)

Fig. 4 Spatial distribution of ESI in Punjab, Pakistan, in 2019


Environ Sci Pollut Res (2021) 28:38145–38158 38153

spatial autocorrelation (i.e., high–high clusters), and were random) factors influencing the level of ESI in the study
mainly located in the northeastern and central parts of the area at certain small scales, whereas structural elements
study area. A total of 18 tehsils, which were primarily located in the study area led to spatial heterogeneity in the ESI.
in the northwestern region, along with small tehsils in the In contrast, the values obtained based on spherical,
western region (Fig. 5), were identified as having positive exponential, and Gaussian models were 0.39, 0.10, and 0.14,
spatial autocorrelation (i.e., low–low clusters). In contrast, respectively, indicating that there were non-structural
negative spatial autocorrelation (i.e., low–high clusters or (random) factors influencing the level of ESI at the tehsil
high–low outliers) were found to be a feature of highly popu- level at certain small scales. The ESI fitting model parameters
lated areas (i.e., municipalities) within the study area. are listed in Table 4. The r2 and RMSECV values obtained
tended to indicate that the spherical model provided a closer
Spatial heterogeneity of ESI approximation to the hypothesis of the four models.
Throughout the assessed regions, spatial heterogeneity was
On the basis of linear model analysis, we obtained a C0/(C + primarily characterized by randomness and autocorrelation
C0) value of 0.79, indicating that there are structural (non- components.

Legend
Not Significant
High-High Cluster
High-Low Outlier
Low-High Outlier
0 50 100 200 Km
Low-Low Cluster
Fig. 5 A cluster map showing the spatial association of ecosystem service intensity
38154 Environ Sci Pollut Res (2021) 28:38145–38158

Table 4 Parameters of spatial


heterogeneity of ESI for different Model C0 C C0 + C A0 C0 / C0 + C r2 Residual SS RMSECV
models
Spherical 45.58 71.60 117.50 1.29 0.39 0.51 609 0.00
Linear 96.60 25.47 122.05 2.61 0.79 0.25 945 0.00
Exponent 12.40 105.80 118.20 1.15 0.10 0.45 697 0.00
Gauss 16.70 99.90 116.60 0.84 0.14 0.45 698 0.00

The driving forces of ESI heterogeneity and dependency at all scales are omnipresent and
provide an essential foundation for ecological hypotheses. For
The findings of GWR and OLS regressions for ESI are summa- decades, the association between patterns and scales has been
rized in Table 5. The adjusted R-squared value indicated the acknowledged in both ecology and geography (Meentemeyer
good fit of both models, whereas the residual and AICc values 1989; Levin 1992; Chave 2013). The dilemma of scale tends
tended to suggest the better performance of the GWR model, to be associated with different outcomes determined from the
thereby providing evidence for the greater applicability of the same data, or the fact that patterns may be better described at
locally weighted regression model. The individual and combined certain scales (e.g., degree of the datasets) (Wu 1999; Zawadzki
effects of driving forces on ESI were analyzed based on GWR at et al. 2005). Two approaches have been described for elucidat-
the tehsil level, as shown in Fig. 6. The interaction effect of ing spatial variability at a landscape scale (i.e., autocorrelation
forestland on ESI was found to be the most pronounced, follow- and semivariance analyses), with a view to identifying the char-
ed by that of built-up, unused land, and cropland. Additionally, acteristic scales of landscape patterns (Bellehumeur et al. 1997).
population density showed a strong interaction effect with ESI Ecological characteristics are also derived from physical features
level. Tehsil-level regression coefficients evaluated using GWR associated with the inherent nature of spatial sampling (Anselin
revealed a non-significant relationship between the proportion of 2001). Following this approach, our main objective in the pres-
forestland and ESI (Fig. 6a). Figure 6b–d show that for all tehsils, ent study was to identify an analytical approach that would
ESI was significantly associated with the proportions of crop- facilitate the measurement of statistical parameters and spatial
land, built-up land, and unused land, whereas Fig. 6c and d variance characteristics when an ESI study is conducted based
provide evidence of an increasing tendency for cropland and on sampling units of differing size. However, our analysis indi-
unused land to influence the ESI of the study region from east cated that most of the statistical parameters and characteristics
to west. In contrast, built-up area was found to be characterized showed unpredictable characteristics in response to differences
by a more significant inverse trend in ESI (Fig. 6c). With respect in sample size. Although prediction error can occur owing to a
to direct factors, we established that 136 tehsils were affected by deficiency in the statistical validity of the sample size, these
forest land, with built-up, crop, and unused land affecting 131, associations [C0 / (C0 + C)] have demonstrated that unsystem-
126, and 82 tehsils, respectively. Contrary to expectations, we atic variations in the present study were less significant (Burnett
found that for all the tehsils in the study area, grassland, water and Blaschke 2003). Using spherical, linear, exponential, and
bodies, and population density coefficients had no significant Gaussian models, we obtained C0 / (C0 + C) values of 0.39,
effects on ESI. 0.79, 0.10, and 0.14, respectively. Each of these models indicat-
ed a random pattern influencing the ESI level at small scales.
According to the r2 values obtained, the spherical model, with a
Discussion value of 0.51, provided the best fit. Corresponding values for the
linear, exponential, and Gaussian models were 0.25, 0.45, and
The problem of scale relating to geographical 0.45, respectively. Given that one of the goals of the present
variation research was to examine the nature of the drivers of the spatial
tendency, we selected a contrast-sampling configuration (a
The pattern is rooted in spatial heterogeneity, which in turn tehsil-level administrative unit) to fit the spatial scale for socio-
stems from spatial dependence variabilities (Wu 2004). Spatial economic assessments.

Table 5 Parameter estimates based on GWR and OLS regression

Model Constant PD (individuals/ha) PFL (%) PBL (%) PGL (%) PCL (%) Adjusted R-squared AICc

GWR − 2.54 − 2.53 − 0.18 − 1.69 0.22 0.99 281.97


OLS − 2.53 − 2.52 − 0.15 − 1.69 0.20 0.99 296.11
Environ Sci Pollut Res (2021) 28:38145–38158 38155

± ±

± ±

Fig. 6 Geographically weighted regression clusters of forestland, cropland, and built-up and unused land

Implications for the flows of ESs Balvanera 2009; Wunder 2007). By describing spatial hetero-
geneity and dependency, we can characterize the distribution
As shown in Fig. 3, we detected substantial differences among of ESs more precisely and specifically than is achievable
areas in terms of ESI. These observation might indicate that based on the conventional use of equity indices and
specific ESs are “provided” by biophysical or anthropogenic choropleth mapping (Yang and Wong 2013). Consequently,
processes from source areas, with flows to those receiving the findings of this study have certain policy implications with
benefits (Serna-Chavez et al. 2014). In this regard, the spatial respect to prioritizing conservation or development zones and
distribution of services needs to be examined with a view the construction of ecological corridors to enhance access to
toward the effective design of ES flow (Guariguata and those areas receiving benefits. Studying spatial variations
38156 Environ Sci Pollut Res (2021) 28:38145–38158

enables us to take into account threshold distances via spatial The approach used in the present study represented an ad-
interpolation or precise computation. This provides a verifica- vance in the use of simple regression models, and by intro-
tion of previous findings relating to the application of geo- ducing a model with spatially varying parameters, we were
graphical variation analysis (Yang and Wong 2013), which able to effectively detect the variance between variables and
has indicated that it is possible to gain more accurate details ESI at a local level. The findings of our GWR model analysis
regarding the distribution of artifacts via such analysis. These indicated that associations differ on a spatial scale, thereby
map features illustrate the prospects for better distribution to indicating that the magnitudes of positive/negative associa-
areas receiving the benefit of ESs, for example, through the tions may differ accordingly. However, in the study site, the
development of recreational and natural reserve areas and transition from east to west across the study site is character-
linking these to those areas receiving benefits. Given the ac- ized by positive changes in a number of gradients, which are
knowledged variability among ESs, studies focusing on dif- closely associated with economic factors, urban growth, and
ferent services are required to promote the study of spatial emerging developments. In this regard, identified regions in
service flows. which certain factors have detrimental effects with respect to
ESI could potentially be approached by the government to
enforce mitigation measures at the tehsil level, given that these
Proximate forces of ESs tehsils are expected to resolve both social and biophysical
issues at the administrative unit level (Poudyal et al. 2012).
The design of an effective ecosystem service management
scheme necessitates a recognition of the dynamic relationships
between socioecological constraints and ESs (Kremen 2005).
In the present study, we examined the intensity of ESs in Conclusions
individual tehsils depending primarily on local land use/
cover patterns, that is, the proportions of forestland, cropland, This study was conducted to determine the spatial variation and
built-up land, and unused land. Our findings are consistent driving factors of ESI in Punjab. The results indicated signifi-
with those presented in an earlier review, indicating that the cant variation in ESI among tehsils, with a decreasing trend
capacity of natural processes to meet human needs via ESs is from east to west, north to south, and from metropolitan to more
directly influenced by land use and land cover (Vitousek et al. rural areas. Our analysis also revealed that built-up area and
1997). A finding of the present study that was unpredicted is unused land showed a negative trend, whereas cropland
that population density and the proportion of water bodies showed a positive trend. Furthermore, the findings of
have non-significant effects on ESI. In contrast, we observed semivariance analysis indicated a non-structural patterning of
that for each tehsil in the study area, the proportion of cropland ESI at the tehsil level. Within the study area, we obtained more
showed a significantly positive association with ESI, whereas detailed information regarding trends in ESI spatial associations
the proportion of forestland showed a significantly negative by using the lag distance range as the neighborhood size rather
association. Similarly, for all tehsils, the proportion of unused than using the default value. Results obtained using a GWR
land was found to be negatively associated with the level of approach were found to be more rational and appropriate com-
ESI. Conversely, the proportion of built-up showed both neg- pared with those obtained based on an OLS model when con-
ative and positive associations, which supports the hypothesis sidering features of the spatial non-stationarity distribution of
that population is the primary and significant underlying cause ESI. GWR model analysis indicated that for each tehsil, a
of global changes in land use and cover (Lambin et al. 2001). change in land use/cover was the principal factor contributing
Changes in ESs typically reflect human reactions to economic to a variation in ESI. Consequently, the findings of this study
opportunities, as interpreted by economic issues (Helfenstein have certain policy-related implications with respect to the pri-
and Kienast 2014; Li et al. 2010). Although the effect of oritization of sustainability or development areas, as well as the
population density on ESI was comparatively limited com- formation of ecological corridors to enhance distribution to
pared with direct local influences (e.g., tehsil), a growing pop- those areas receiving ES benefits. Given the discrepancies be-
ulation necessitates an expansion of the area land use/cover tween ESs, studies focusing on different service are required to
associated with human development (e.g., housing settle- gain a more in-depth insight into spatial service flows.
ments, utilities, industries, and commercial malls), eventually
affecting the extent and quality of ecological land (e.g., crop-
land and forestland) (Liu et al. 2005; Li et al. 2010). Acknowledgements We would like to thank the Editor and anonymous
reviewers for their constructive and valuable remarks, which have en-
Industrialization has also contributed to an increasing loss of
abled us to substantially enhance the quality of the paper.
ecological land (Deng et al. 2008), and consequently, popula-
tion density might represent a key driver in the ongoing dete- Author contribution MA, YZ, and GG conceived and designed the
rioration of local ESs. study; MA collected the data; MA and YZ analyzed the data; MA wrote
Environ Sci Pollut Res (2021) 28:38145–38158 38157

the paper, and YZ and GG provided useful suggestions and modified the Davis JC, Sampson RJ (2002) Statistics and data analysis in geology. 3rd
draft. All authors have read and approved the final manuscript. edition. John Wiley & Sons, New York (57–61 pp.)
De Groot R, Brander L, Van Der Ploeg S, Costanza R, Bernard F, Braat L
Funding Our research was supported by the Fundamental Research et al (2012) Global estimates of the value of ecosystems and their
Funds for the Central Universities (No. 2015ZCQ-SB-02). services in monetary units. Ecosyst Serv 1:50–61
Deng X, Huang J, Rozelle S, Uchida E (2008) Growth, population and
industrialization, and urban land expansion of China. J Urban Econ
Data Availability The datasets used and/or analyzed during the current
63:96–115
study are available from the corresponding author upon reasonable
request. FAO (2015) The land cover atlas of Pakistan: The Punjab Province.
Available online: http://www.fao.org/geospatial/resources/detail/
en/c/1024589/e
Declarations Fisher B, Turner RK, Morling P (2009) Defining and classifying ecosys-
tem services for decision making. Ecol Econ 68:643–653
Ethics approval and consent to participate Not applicable. Fotheringham AS, Brunsdon C, Charlton M (2002) Geographically
weighted regression: the analysis of spatially varying relationships:
Consent for publication Not applicable. John Wiley & Sons, New York (27–64 pp.)
Guariguata MR, Balvanera P (2009) Tropical forest service flows:
Improving our understanding of the biophysical dimension of eco-
Competing interests The authors declare that they have no competing
system services. For Ecol Manag 258:1825–1829
interests.
He Z, Zhao W, Chang X (2007) The modifiable areal unit problem of
spatial heterogeneity of plant community in the transitional zone
between oasis and desert using semivariance analysis. Landsc Ecol
References 22:95–104
Helfenstein J, Kienast F (2014) Ecosystem service state and trends at the
Akhtar M, Zhao Y, Gao G, Gulzar Q, Hussain A, Samie A (2020) regional to national level: a rapid assessment. Ecol Indic 36:11–18
Assessment of ecosystem services value in response to prevailing Hu X, Hong W, Qiu R, Hong T, Chen C, Wu C (2015) Geographic
and future land use/cover changes in Lahore, Pakistan. Regional variations of ecosystem service intensity in Fuzhou City, China.
Sustain 1:37–47 Sci Total Environ 512:215–226
Anselin L (1999) Interactive techniques and exploratory spatial data anal- Huang Q, Zhao X, He C, Yin D, Meng S (2019) Impacts of urban ex-
ysis. In: Longley PA, Godchild MF, Maguire DJ (Eds.), pansion on wetland ecosystem services in the context of hosting the
Geographical Information Systems. John Wiley & Sons, New Winter Olympics: a scenario simulation in the Guanting reservoir
York, pp. 253–266 basin, China. Reg Environ Chang 19:2365–2379
Anselin L (2001) Spatial effects in econometric practice in environmental Huang Q, Yin D, He C, Yan J, Liu Z, Meng S, Ren Q, Zhao R, Inostroza
and resource economics. Am J Agric Econ 83:705–710 L (2020) Linking ecosystem services and subjective well-being in
Anselin L (2003) GeoDa 0.9 user’s guide. Spatial Analysis Laboratory rapidly urbanizing watersheds: insights from a multilevel linear
(SAL). Department of Agricultural and Consumer Economics, model. Ecosyst Serv 43:101106
University of Illinois, Urbana-Champaign Imran M, Sumra K, Abbas N, Majeed I (2019) Spatial distribution and
Arowolo AO, Deng X, Olatunji OA, Obayelu AE (2018) Assessing opportunity mapping: applicability of evidence-based policy impli-
changes in the value of ecosystem services in response to land- cations in Punjab using remote sensing and global products. Sustain
use/land-cover dynamics in Nigeria. Sci Total Environ 636:597–609 Cities Soc 50:101652
Barbier EB (2012) A spatial model of coastal ecosystem services. Ecol Isbell F, Calcagno V, Hector A, Connolly J, Harpole WS, Reich PB,
Econ 78:70–79 Scherer-Lorenzen M, Schmid B, Tilman D, van Ruijven J,
Bellehumeur C, Legendre P, Marcotte D (1997) Variance and spatial Weigelt A, Wilsey BJ, Zavaleta ES, Loreau M (2011) High plant
scales in a tropical rain forest: changing the size of sampling units. diversity is needed to maintain ecosystem services. Nature 477:199–
Plant Ecol 130:89–98 202
Burnett C, Blaschke T (2003) A multi-scale segmentation/object relation- Kremen C (2005) Managing ecosystem services: what do we need to
ship modelling methodology for landscape analysis. Ecol Model know about their ecology? Ecol Lett 8:468–479
168:233–249 Krige DG (1996a) A study of gold and uranium distribution patterns in
Busch M, La Notte A, Laporte V, Erhard M (2012) Potentials of quanti- the Klerksdorp gold field. Geoexploration 4(1):43–53
tative and qualitative approaches to assessing ecosystem services. Krige DG (1996b) Two-dimensional weightedmoving average trend sur-
Ecol Indic 21:89–103 face for ore evaluation. J South Afr Inst Min Metall 67:13–79
Cambardella CA, Moorman TB, Parkin TB, Karlen DL, Novak JM, Lambin EF, Turner BL, Geist HJ, Agbola SB, Angelsen A, Bruce JW,
Turco RF, Konopka AE (1994) Field-scale variability of soil prop- Coomes OT, Dirzo R, Fischer G, Folke C, George PS, Homewood
erties in central Iowa soils. Soil Sci Soc Am J 58(5):1501–1511 K, Imbernon J, Leemans R, Li X, Moran EF, Mortimore M,
Chave J (2013) The problem of pattern and scale in ecology: what have Ramakrishnan PS, Richards JF, Skånes H, Steffen W, Stone GD,
we learned in 20 years? Ecol Lett 16:4–16 Svedin U, Veldkamp TA, Vogel C, Xu J (2001) The causes of land-
Chen L, Feng Q (2013) Geostatistical analysis of temporal and spatial use and land-cover change: moving beyond the myths. Glob
variations in groundwater levels and quality in the Minqin oasis, Environ Chang 11:261–269
Northwest China. Environ Earth Sci 70:1367–1378 Levin SA (1992) The problem of pattern and scale in ecology: the Robert
Costanza R, d'Arge R, De Groot R, Farber S, Grasso M, Hannon B et al H. MacArthur award lecture. Ecology 73:1943–1967
(1997) The value of the world's ecosystem services and natural Li M, Mao L, Zhou C, Vogelmann JE, Zhu Z (2010) Comparing forest
capital. Nature 387:253 fragmentation and its drivers in China and the USA with Globcover
Costanza R, de Groot R, Sutton P, Van der Ploeg S, Anderson SJ, v2. 2. J Environ Manag 91:2572–2580
Kubiszewski I et al (2014) Changes in the global value of ecosystem Liu J, Liu M, Tian H, Zhuang D, Zhang Z, Zhang W et al (2005) Spatial
services. Glob Environ Chang 26:152–158 and temporal patterns of China's cropland during 1990–2000: an
38158 Environ Sci Pollut Res (2021) 28:38145–38158

analysis based on Landsat TM data. Remote Sens Environ 98:442– Trabucchi M, O’Farrell PJ, Notivol E, Comín FA (2014) Mapping eco-
456 logical processes and ecosystem services for prioritizing restoration
Lukaszyk S (2004) A new concept of probability metric and its applica- efforts in a semi-arid Mediterranean river basin. Environ Manag 53:
tions in approximation of scattered data sets. Comput Mech 33:299– 1132–1145
304 Treitz P, Howarth P (2000) High spatial resolution remote sensing data
McMillen DP (2010) Issues in spatial data analysis. J Reg Sci 50:119– for forest ecosystem classification: an examination of spatial scale.
141 Remote Sens Environ 72(3):268–289
MEA (2005) Ecosystems and human well-being. Vol 5. Island press, Troy A, Wilson MA (2006) Mapping ecosystem services: practical chal-
Washington, DC lenges and opportunities in linking GIS and value transfer. Ecol
Meentemeyer V (1989) Geographical perspectives of space, time, and Econ 60:435–449
scale. Landsc Ecol 3:163–173 Turner WR, Brandon K, Brooks TM, Costanza R, Da Fonseca GA,
Nelson E, Mendoza G, Regetz J, Polasky S, Tallis H, Cameron D, Chan Portela R (2007) Global conservation of biodiversity and ecosystem
KMA, Daily GC, Goldstein J, Kareiva PM, Lonsdorf E, Naidoo R, services. BioScience 57:868–873
Ricketts TH, Shaw MR (2009) Modeling multiple ecosystem ser- Usman M, Liedl R, Shahid MA, Abbas A (2015) Land use/land cover
vices, biodiversity conservation, commodity production, and classification and its change detection using multi-temporal MODIS
tradeoffs at landscape scales. Front Ecol Environ 7:4–11 NDVI data. J Geogr Sci 25:1479–1506
PBS (2018) Pakistan bureau of statistics (online). http://www.pbs.gov.pk/ Vihervaara P, Rönkä M, Walls M (2010) Trends in ecosystem service
Accessed 21.03.16 research: early steps and current drivers. Ambio 39:314–324
Pijanowski BC, Iverson LR, Drew CA, Bulley HN, Rhemtulla JM,
Vitousek PM, Mooney HA, Lubchenco J, Melillo JM (1997) Human
Wimberly MC et al (2010) Addressing the interplay of poverty
domination of Earth's ecosystems. Science 277:494–499
and the ecology of landscapes: a Grand Challenge Topic for land-
Wallace CS, Watts JM, Yool SR (2000) Characterizing the spatial struc-
scape ecologists? Landsc Ecol 25:5–16
ture of vegetation communities in the Mojave Desert using
Plummer ML (2009) Assessing benefit transfer for the valuation of eco-
geostatistical techniques. Comput Geosci 26:397–410
system services. Front Ecol Environ 7:38–45
Poudyal NC, Johnson-Gaither C, Goodrick S, Bowker J, Gan J (2012) Webster R, Olivier MA (1990) Statistical methods in soil and land re-
Locating spatial variation in the association between wildland fire source survey. Oxford University Press New York (316 pp.)
risk and social vulnerability across six southern states. Environ Wu J (1999) Hierarchy and scaling: extrapolating information along a
Manag 49:623–635 scaling ladder. Can J Remote Sens 25:367–380
Qi Y, Lian X, Wang H, Zhang J, Yang R (2020) Dynamic mechanism Wu J (2004) Effects of changing scale on landscape pattern analysis:
between human activities and ecosystem services: a case study of scaling relations. Landsc Ecol 19:125–138
Qinghai lake watershed, China. Ecol Indicators 117:106528 Wu J, Feng Z, Gao Y, Peng J (2013) Hotspot and relationship identifica-
Richardson L, Loomis J, Kroeger T, Casey F (2015) The role of benefit tion in multiple landscape services: a case study on an area with
transfer in ecosystem service valuation. Ecol Econ 115:51–58 intensive human activities. Ecol Indic 29:529–537
Samie A, Deng X, Jia S, Chen D (2017) Scenario-based simulation on Wunder S (2007) The efficiency of payments for environmental services
dynamics of land-use-land-cover change in Punjab Province, in tropical conservation. Conserv Biol 21:48–58
Pakistan. Sustainability 9:1285 Xie G, Zhen L, Lu C-X, Xiao Y, Chen C (2008) Expert knowledge based
Samie A, Abbas A, Azeem MM, Hamid S, Iqbal MA, Hasan SS et al valuation method of ecosystem services in China. J Nat Resour 23:
(2020) Examining the impacts of future land use/land cover changes 911–919
on climate in Punjab province, Pakistan: implications for environ- Xie G, Zhang C, Zhen L, Zhang L (2017) Dynamic changes in the value
mental sustainability and economic growth. Environ Sci Pollut Res: of China’s ecosystem services. Ecosyst Serv 26:146–154
1–19 Yang Y, Wong KKF (2013) Spatial distribution of tourist flows to
Serna-Chavez H, Schulp C, Van Bodegom P, Bouten W, Verburg P, China's cities. Tour Geogr 15:338–363
Davidson M (2014) A quantitative framework for assessing spatial Yi H, Güneralp B, Filippi AM, Kreuter UP, Güneralp İ (2017) Impacts of
flows of ecosystem services. Ecol Indic 39:24–33 land change on ecosystem services in the San Antonio River Basin,
Shedayi A, Xu M, Gonalez-Redin J, Hagist S, Aslam S, Khan N (2019) Texas, from 1984 to 2010. Ecol Econ 135:125–135
Spatiotemporal assessment and valuation of provisioning ecosystem Zawadzki J, Fabijańczyk P (2013) Geostatistical evaluation of lead and
services of Pakistan. Appl Ecol Environ Res 17:6735–6759 zinc concentration in soils of an old mining area with complex land
Shi Y, Wang R, Huang J, Yang W (2012) An analysis of the spatial and management. Int J Environ Sci Technol 10:729–742
temporal changes in Chinese terrestrial ecosystem service functions.
Zawadzki J, Cieszewski CJ, Zasada M, Lowe RC (2005) Applying
Chin Sci Bull 57:2120–2131
geostatistics for investigations of forest ecosystems using remote
Su C-h FB-J, He C-S, Lü Y-H (2012) Variation of ecosystem services and
sensing imagery. Silva Fennica 39:599
human activities: a case study in the Yanhe Watershed of China.
Zawadzki J, Magiera T, Fabijańczyk P (2009) Geostatistical evaluation of
Acta Oecol 44:46–57
magnetic indicators of forest soil contamination with heavy metals.
Sun X, Tang H, Yang P, Hu G, Liu Z, Wu J (2020) Spatiotemporal
Stud Geophys Geod 53:133–149
patterns and drivers of ecosystem service supply and demand across
the conterminous United States: a multiscale analysis. Sci Total Zhai T, Wang J, Jin Z, Qi Y, Fang Y, Liu J (2020) Did improvements of
Environ 703:135005 ecosystem services supply-demand imbalance change environmen-
Syrbe R-U, Walz U (2012) Spatial indicators for the assessment of eco- tal spatial injustices? Ecol Indic 111:106068
system services: providing, benefiting and connecting areas and Zhang Y, Xu J-H, Zhuang P-J (2011) The spatial relationship of tourist
landscape metrics. Ecol Indic 21:80–88 distribution in Chinese cities. Tour Geogr 13:75–90
Talberth J (2015) Valuing ecosystem services in the lower mekong basin:
country report for Cambodia. Washington DC: USAID Mekong Publisher’s note Springer Nature remains neutral with regard to jurisdic-
ARCC tional claims in published maps and institutional affiliations.

You might also like