Content uploaded by Isaac Sarfo
Author content
All content in this area was uploaded by Isaac Sarfo on Nov 23, 2024
Content may be subject to copyright.
Acta Sci. Pol.
Formatio Circumiectus 23 (3) 2024, 27–56
DOI: http://dx.doi.org/10.15576/ASP.FC/190971www.acta.urk.edu.pl ISSN 1644-0765
ORIGINAL PAPER Accepted: 9.07.2024
e-mail: jjqiao@henu.edu.cn
© Copyright by Wydawnictwo Uniwersytetu Rolniczego wKrakowie, Kraków 2024
ENVIRONMENTAL PROCESSES
CAUSAL EFFECTS AND PREDICTION OF LAND USE SYSTEMS IN
RURAL LANDSCAPES: EVIDENCE FROM HENAN PROVINCE
Isaac Sarfo1, 2 0000-0002-6914-5764, Jiajun Qiao1 0000-0003-3494-2197,
Emmanuel Yeboah3 0000-0003-3838-6837, Abraham Okrah4, 5 0009-0006-8049-7310,
Charafa El Rhadiouini3, Benjamin Kwapong Osibo6, Anita Boah7,
Dhekra Ben Amara1 0009-0000-2467-8754
1 College of Geography and Environmental Science, Henan University, Kaifeng city, Henan Province, China
2 Organization of African Academic Doctors (OAAD), Off Kamiti Road, P.O. Box 25305000100, Nairobi, Kenya
3
School of Remote Sensing and Geomatics Engineering, Nanjing University of Information Science and Technology, 210044
Nanjing, Jiangsu, China
4
Collaborative Innovation Center on Forecast and Evaluation of Meteorological Disaster, Nanjing University of Information
Science and Technology, 210044 Nanjing, Jiangsu, China
5
Department of Meteorology and Climate Science, Kwame Nkrumah University of Science and Technology,
00233, Kumasi, Ghana
6 School of Computer and Software, Nanjing University of Information Science and Technology, 210044 Nanjing, Jiangsu, China
7 Department of Public Health and Allied Sciences, Catholic University, Fiapre, Ghana
ABSTRACT
Aim of the study
In rural and agricultural development, land plays a crucial role in driving productivity. To understand the
impact of specic causes or combinations of causes on outcomes, it is essential to identify and establish
clear causal relationships. Our study investigates the causal eects of dierent land use systems against
Land Surface Temperature (LST) in Henan Province. We further make land use predictions based on current
trends. Understanding these dynamics is essential for enhancing agricultural informatization, environmental
management, and climate-smart choices of local districts, counties and villages across China’s agriculturally
important regions and beyond.
Material and methods
The study utilized integrated remote sensing data, techniques and a causality approach to investigate land
use systems (LUS) and LST in Henan Province. We further used Modules for Land Use Change Evaluation
(MOLUSCE) and Cellular Automata-Articial Neural Network (CA-ANN) to predict LUS for the near fu-
ture (2023–2053).
Results and conclusions
Results revealed that built-up areas (+500%), forests (+50.88%) and water bodies (+83.56%) have expanded
massively during the past 40 years. In contrast, cultivated (–20.81%) and barren areas (–60.53%) declined
steadily. The temporal causal inference analysis demonstrated a strong convergence between built-up areas
and land surface temperature (LST), which substantiates built-up areas’ profound impact on LST intensity.
The spatial causal inference analysis shows moderate to robust positive indirect cross-mapping relationships
between built-up areas (ρ = 0.63) and bare land (ρ = 0.32) against LST. Land use predictions (2023–2053)
Sarfo, I., Qiao, J., Yeboah, E., Okrah, A. , El Rhadiouini, Ch., Osibo, B.K., Boah, A., Ben Amara, D. (2024). Causal effects and prediction of land
use systems in rural landscapes.. . Acta Sci. Pol., Formatio Circumiectus, 23 (3), 27–56. DOI: http://dx.doi.org/10.15576/ASP.FC/190971
28 www.acta.urk.edu.pl
show a reduction in areas covered by forests and water bodies, and a reversed trend in cultivated lands. These
are particularly important when formulating targeted policy-directives needed to regulate unsustainable land-
use processes and undesirable economic trade-os.
Keywords: causal analysis, Geodetector, land use and land cover, land surface temperature (LST), China
INTRODUCTION
In the face of multiple climate stressors, land use sys-
tems development (LUSD) plays an essential role in
addressing concerns related to land degradation, ru-
ral-urban resilience, and the optimal land resource
use (Wang et al., 2021). Changes in land use systems
continue to be the most fundamental and signicant
landscape feature that demonstrates human activities’
ever-changing impact on the environment. Land sys-
tems are the interactions and responses of dierent key
players, institutions, cultural traditions, and competing
interests (Turner, 2020). According to Magliocca et al.
(2023), one important aspect of land system science
(LSS) is the understanding of the driving processes of
land use change (LUC). However, the intricate interplay
between various elements of dierent sizes frequently
results in complex causality pathways, making it dif-
cult to identify and evaluate causal eects and pro-
cesses. This phenomenon is critical to understanding
local, regional, and global environmental shifts (Reay,
2020). Throughout history, the transition from undis-
turbed to cultivated areas has been the primary driver
of global land use and land cover change (LULCC)
(FAO, 2020). In their study titled “Worldwide research
trends on sustainable land-use in agriculture,” Az-
nar-Sánchez et al. (2019) found that 42% of the glob-
al population relies on agriculture for their livelihood.
Many developing countries’ economies rely heavily on
this industry. Previous studies have shown that when
the human population grows at a high rate, so does the
demand to convert and use land for agriculture and
other purposes. In recent years, there has been a grow-
ing focus on studying the sustainable use of land in
agriculturally-oriented regions. These regions play an
essential role in ensuring food security and economic
development, not just in China but also in other parts of
the world (Hinz et al., 2020; Xi et al., 2023).
Land use predictions, according to Xu et al. (2022),
inform the decisions of city planners, provincial, dis-
trict, and other administrative bodies, among other
interested parties, to optimize current eorts aimed
at averting undesirable consequences or building re-
silience against unforeseeable events. In Central Chi-
na, particularly in Henan province, which contributes
immensely to the nation’s food basket and economy,
a comprehensive understanding of the driving mech-
anisms through this causality study will provide the
much-needed technical basis that supports the sus-
tainable utilization of land and the management of
the province’s rural (i.e., agricultural areas) and urban
settings. Notable solutions among the commonly used
predictive land use models in human, regional, and
economic geographies include the CA-Markov model
(Xu et al., 2022), the GEOMOD model (Sakieh and
Salmanmahiny, 2016), the MLP – NN model (Gir-
ma et al., 2021), the CLUE-S model (Huang et al.,
2019), and the CA-ANN model (Değermenci, 2023).
Nevertheless, these standardized models are ecient
in modelling trends over a given period. However,
each model has some limitations linked to spatial lev-
els, duration/extent of coverage, number of images,
as well as ease/convenience and accuracy of model-
ling. To this end, this study employs a Geodetector
and a CA-ANN to investigate and predict the driving
mechanisms behind Henan province’s LULCC and
LST, respectively. It is worth noting that the core and
potential drivers, thus, spatial and non-spatial factors
of LULCC emanate from dierent sources. The latter
is mainly driven by policy-driven and planning initia-
tives, which we attempt to explore extensively in this
study. Similarly, we focus on the applicability of cau-
sality and simulation of future land use systems in an
agriculturally-driven region such as the Henan prov-
ince, which remains underdeveloped considering the
region’s relevance to the People Republic of China’s
sustenance and socio-economic development. Identi-
fying and establishing causal relationships is vital to
interpreting how a cause or a combination of factors
inuences an outcome. Understanding these dynamics
Sarfo, I., Qiao, J., Yeboah, E., Okrah, A. , El Rhadiouini, Ch., Osibo, B.K., Boah, A., Ben Amara, D. (2024). Causal effects and prediction of land
use systems in rural landscapes.. . Acta Sci. Pol., Formatio Circumiectus, 23 (3), 27–56. DOI: http://dx.doi.org/10.15576/ASP.FC/190971
29
www.acta.urk.edu.pl
is essential for enhancing agricultural informatization,
environmental management, and climate-smart choic-
es of local districts, counties and villages across Chi-
na’s agriculturally important regions and beyond. To
this end, this study sought to:
1. Examine the historical shifts in some land use
systems (i.e., based on remotely-sensed indi-
ces such as Enhanced Vegetation Index (EVI),
Norma lized Dierence Built-Up Index (NDBI),
Normalized Dierence Water Index (NDWI), and
Normalized Dierence Bareness Index (NDBaI))
and Land Surface Temperature (LST).
2. Investigate the causal eects of dierent land use
systems against LST in Henan Province.
3. Make land-use prediction in the near future (2023–
2033) and mid-century (2043–2053).
METHODOLOGY
Study setting
Henan, a province located in the Central part of China,
covers an area of 167,000 km2 at Latitude 34.765 oN and
Longitude 113.753o E. The region is landlocked, and
halved by the Yellow River (also known as Huang He):
one-sixth north and ve-sixths south of the major riv-
er. The total population of the province, according to
HPBS (2018), oscillates around 109.06 million.
Data acquisition, processing and image classication
This study utilized ve Landsat images with a resolu-
tion of 30 meters, which were obtained from the Unit-
ed States Geological Survey (USGS) website (http://
earthexplorer.usgs.gov/), for the period of 1983–2023.
Fig. 1. Geographical location of Henan Province (source: Authors’ own elaboration)
Sarfo, I., Qiao, J., Yeboah, E., Okrah, A. , El Rhadiouini, Ch., Osibo, B.K., Boah, A., Ben Amara, D. (2024). Causal effects and prediction of land
use systems in rural landscapes.. . Acta Sci. Pol., Formatio Circumiectus, 23 (3), 27–56. DOI: http://dx.doi.org/10.15576/ASP.FC/190971
30 www.acta.urk.edu.pl
The image pre-processing, processing, and map com-
position were conducted using ArcGIS 10.8, ENVI 5.0,
and ENVI 5.3. Additional image pre-processing proce-
dures (Fig. 2) included image calibration, layer stack-
ing, and supervised classication. The types of imagery
used for analysis were obtained from the LANDSAT
4 TM, LANDSAT 5 TM, LANDSAT 7 ETM+ data-
sets (using spectral bands 7, 4, and 2 together), and the
LANDSAT 8 OLI/TIRS dataset (using spectral bands
7, 5, and 3 together). The satellite data corresponds to
paths/rows 123, 124/035, 036, and 037.
Change Detection Analysis
Change detection analysis was run to ascertain the reg-
ularity of land use systems and its driving mechanisms
in Henan Province, using Eqns. 1–3:
Change in LULC
LULC LULC
LULC
Current yea
rP
ast year
Past year
...
(1)
%Change in LULC
LULC LULC
LULC
Current yea
rP
ast year
Past year
1000% ...
(2)
Rate of change in LULC per year
LULC LULC
LUL
Current yearPast year
CC Nyears
Past year
100% ...
(3)
where:
LULCCurrent year – the nal year under study (2023)
within the context of the present
study,
LULCPast year – the initial year being studied (1983),
N – the dierence in the understudied year span, thus,
40 years.
Fig. 2. Workow for image preprocessing and post-classication designed for the study (source: Authors’ own elaboration)
Input Data
Causal Analysis
Temperature Analysis
Remotely-sensed
Indices
Reference Data
Driving variables
Study Area
Landsat Imagery for 1983,
1993, 2003, 2013 & 2023
*Summer seasonal data was
utilized for temperature analysis
Image pre-
processing and
enhancement
Supervised
Classification with
Maximum Likelihood
Classification
Algorithm (MCLA)
Change Detection
Analysis
Accuracy Assessment
*Temporal Causal Inference
(CCM)
*Spatial causal inference
(GCCM)
*Conversions ( C), calibrations,
o
radiometric corrections,
gap-filling, emissivity and
correlation analysis
EVI, NDWI, NDBaI, NDVI
and NDBI
Henan Province, China
Google Earth Engine
& Google Earth Pro
Land Use Metrics
EVALUATION
Sarfo, I., Qiao, J., Yeboah, E., Okrah, A. , El Rhadiouini, Ch., Osibo, B.K., Boah, A., Ben Amara, D. (2024). Causal effects and prediction of land
use systems in rural landscapes.. . Acta Sci. Pol., Formatio Circumiectus, 23 (3), 27–56. DOI: http://dx.doi.org/10.15576/ASP.FC/190971
31
www.acta.urk.edu.pl
The rate of change in LULC per year was computed
to ascertain the gains/expansion and losses/reduction
in areas covered by the given land use classes (refer
to Table A.1) over the past 40 years. This provides de-
tailed information on how changes occurred annually
to substantiate major events, transitions and underlying
mechanisms that drove such changes. The expressions
used in computing landscape metrics are captured in
the supplementary material (See appendix).
Temperature analysis
Radiometric correction (radiance) was applied to rec-
tify atmospheric eects and enhance clarity. Gap-ll-
ing was performed for images that may have had
stripes. Distortions in images were removed in the cal-
ibration process to enhance image quality. We tailored
the conversion of DN to spectral radiance according to
the procedures of Coll et al. (2010), as retrieved from
USGS Landsat User handbook. The ETM + DN val-
ues range between 0 and 255. Equation (4) was used to
determine the radiance for the study domain:
LLMAX LMIN
QCALMAXQCALMIN DN QCALMIN
LMIN
λλλ
λ
=−
−×− +
+
()
()
()
... (4)
where:
Lλ – cell value, simplied as radiance in
w
Ms
rm()
2⋅⋅µ,
LMAXλ – sensor spectral radiance scaled to (QCAL-
MAX) in
w
Ms
rm()
2⋅⋅µ,
LMINλ – sensor spectral radiance scaled to (QCAL-
MIN) in w
Ms
rm()
2⋅⋅
µ,
(QCALMAX) – maximum quantized calibrated pixel
value that corresponds to LMAXλ [DN],
(QCALMIN) – minimum quantized calibrated pixel
value corresponding to LMINλ [DN],
QCAL – quantized calibrated pixel value [DN].
LMIN and LMAX – spectral radiances for each band
at DN 1 and 255 for Landsat 7
ETM+ 1 and 65535 for Landsat
8 OLI/TIRS,
λ – the wavelength.
Conversion of Spectral Radiance (Lλ) to Kelvin
with emissivity value (Eqns. 5–6) is conducted as fol-
lows:
T
K
KE
L
=⋅+
2
11ln
...
λ
(5)
BT K
KL
=+
[]
2
11ln (/)...
λ
(6)
Table A.2 presents k1 and k2 becoming coecients,
determined by eective wavelength of a satellite sen-
sor based on these constants. Removal of atmospheric
distortions from the thermal infrared data was per-
formed using ENVI 5.0 software for the correction of
thermal band 10 (Table A.3). Values generated were
converted from Kelvin (K) (TB) to degree Celsius (°C)
using the expression (eqn.7). It is worth noting that
only summer seasonal data was used in the analysis.
TT
CB
=−
273 15. ... (7)
Indirect causality analysis: Temporal and spatial
causal relationship between LULCC and LST
In this study, we focus on Henan Province to explore
the temporal and spatial indirect causal relationship
between LULCC variables and land surface tempera-
ture (LST). In the dynamic urbanization context of
Henan, characterized by extensive urban expansion
and industrial growth, we sought to understand how
specic LULCC variables inuence the development
and intensication of LST. Through the analysis of sat-
ellite imagery, temperature data, and Henan-specic
land use data spanning recent decades, Geographical
Convergent Cross Mapping (GCCM) and Convergent
Cross Mapping (CCM) causal analysis are employed
to reveal spatial patterns and temporal trends. Using
causal models (Gao et al., 2023), we sought to es-
tablish a causal link between LULCC variables and
LST in Henan, considering unique contextual factors.
The ndings provide insights into the implications
for urban planning and environmental management
in Henan Province, contributing to a broader under-
standing of LULCC and LST dynamics in rapidly ur-
banizing regions. Future research recommendations
could focus on the evolving urban landscape in Henan,
Sarfo, I., Qiao, J., Yeboah, E., Okrah, A. , El Rhadiouini, Ch., Osibo, B.K., Boah, A., Ben Amara, D. (2024). Causal effects and prediction of land
use systems in rural landscapes.. . Acta Sci. Pol., Formatio Circumiectus, 23 (3), 27–56. DOI: http://dx.doi.org/10.15576/ASP.FC/190971
32 www.acta.urk.edu.pl
emphasizing the importance of sustainable urban de-
velopment in this context. Expressions and variables
used in computing temporal and spatial causal infer-
ences are detailed in the supplementary material (see
appendix).
Land use prediction and validation
Forecasts for 2033, 2043, and 2053 (Fig. 3) were cre-
ated using the Modules for Land Use Change Evalua-
tion (MOLUSCE) in QGIS software version 2.18.24.
This plugin utilizes Cellular Automata and Articial
Neural Network (CA-ANN) techniques and simula-
tions to make predictions for Henan Province. The
analysis primarily involves Evaluating Correlation
(EC), area changes, Transition Potential Computation
(TPC) modeling, and validation based on four itera-
tions. To make these predictions, a Digital Elevation
Model (DEM) and a road raster georeferenced image
of Henan Province were used as reference data. The
key variables used as references for future projections
include the built environment (such as the likelihood
of change, density of developed lands, crop land, and
transportation), socio-economy (including population
density, number of households, urban population den-
sity, urbanization, and industrialization), and natural
environment (such as climatic variables - temperature,
precipitation, and moisture, as well as ecological and
topographical variables).
Accuracy Assessment
In examining the accuracy of each study period, ground
truth sample points were obtained using ENVI 5.0 and
ArcGIS 10.8 software. These points were overlaid on
Google Earth Pro for verication. Hundred sample
points were generated from each class in the classi-
ed images for accuracy assessment (Fig. 4). We em-
ployed Congalton (1991) confusion matrices, cited in
Sarfo et al. (2022) (equation 8) to validate the spatial
results obtained for this study. This standardized ma-
trix (https://pages.cms.huberlin.de/EOL/geo_rs/S10_
Fig. 3. Evaluation of land use predictions procedures (source: Authors’ own elaboration)
Sarfo, I., Qiao, J., Yeboah, E., Okrah, A. , El Rhadiouini, Ch., Osibo, B.K., Boah, A., Ben Amara, D. (2024). Causal effects and prediction of land
use systems in rural landscapes.. . Acta Sci. Pol., Formatio Circumiectus, 23 (3), 27–56. DOI: http://dx.doi.org/10.15576/ASP.FC/190971
33
www.acta.urk.edu.pl
Accuracy_assessment.html#Confusion_matrix) com-
bines and improves upon the user and producer accu-
racy assessments applied by several scholars and re-
search institutes across the globe to ensure validity and
accuracy in images generated.
Accuracy AssessmentAAASP TSP
()[( /) ...]
=×
100
(8)
where:
ASP – number of sample points that accurately fall
on each required feature (ASP = 450),
TSP – number of total sample points generated (TSP
= 500),
AA = Accuracy Assessment [(450/500) × 100 = 90%].
Therefore, the present study had 90% accuracy over
the study period considering the samples collected.
RESULTS
Spatial distribution of Henan Province’s land use
systems
The given distributions (Tables 3, 4 and 5) indicate land
cover conversion over the past 40 years for each land
cover type. Fig. 5 shows that built-up areas (+500.46%),
forests (+50.88%) and water bodies (+83.56%) were
the key land cover types, which expanded massively
between 1983 and 2023. By contrast, a reduction can
be observed for farmlands/shrubs (–20.81%) and bare
land (–60.53%) over the same period.
Fig. 4. Accuracy assessment of satellite imagery over the given study period (source: Authors’ own elaboration)
Sarfo, I., Qiao, J., Yeboah, E., Okrah, A. , El Rhadiouini, Ch., Osibo, B.K., Boah, A., Ben Amara, D. (2024). Causal effects and prediction of land
use systems in rural landscapes.. . Acta Sci. Pol., Formatio Circumiectus, 23 (3), 27–56. DOI: http://dx.doi.org/10.15576/ASP.FC/190971
34 www.acta.urk.edu.pl
Table 3. Area coverage for each class (km2) in Henan Province (1983–2023) (source: Authors’ own elaboration)
Class/Period 1983 1993 2003 2013 2023
Farmlands/ shrubs 12 6032 113 186 123 175 104 785 99 808
Bare land 5 419 8 377 4 588 3 038 2 139
Built-up areas 1 723 1 799 3 905 8 269 10 432
Forests 22 858 27 245 20 865 33 729 34 488
Water bodies 10 968 16 393 14 467 17 179 20 133
***Total area coverage (km2) (Absolute) = 167.000
Fig. 5. Henan Province’s land use cover changes between 1983 and 2023 (source: Authors’ own elaboration)
Sarfo, I., Qiao, J., Yeboah, E., Okrah, A. , El Rhadiouini, Ch., Osibo, B.K., Boah, A., Ben Amara, D. (2024). Causal effects and prediction of land
use systems in rural landscapes.. . Acta Sci. Pol., Formatio Circumiectus, 23 (3), 27–56. DOI: http://dx.doi.org/10.15576/ASP.FC/190971
35
www.acta.urk.edu.pl
Table 4. Temporal variations of land cover changes (LCC) (%) for Henan Province (1983–2023) (source: Authors’ own
elaboration)
Class/Period 1983–1993 1993–2003 2003–2013 2013–2023 1983–2023
Farmlands/shrubs –10.19 +8.83 –14.93 –4.75 –20.81
Bare land +54.59 –45.23 –33.78 –29.59 –60.53
Built-up areas +4.41 +117.07 +111.75 +26.16 +500.46
Forests +19.19 –23.42 +61.65 +2.26 +50.88
Water bodies +49.46 –11.75 +18.75 +17.19 +83.56
Table 5. Rate and magnitude of change (sq.km) of LCC in Henan Province (source: Authors’ own elaboration)
1983–2023
Class/Period 1983 2023 Magnitude of ∆
(km2)Rate of ∆/Yr (%) Magnitude of ∆
(km2)/Yr
Farmlands/shrubs 126 032 99 808 –26 224 –0.5 –655.6
Bare land 5 419 2 139 –3 280 –1.5 –82
Built-up areas 1 723 10 432 +8 709 +12.5 +217.7
Forests 22 858 34 488 +11 630 +1.3 +290.75
Water bodies 10 968 20 133 +9 165 +2.1 +229.1
Land Surface Temperature (LST) evaluation
A fundamental drift in LST can be observed
throughout the study period. Seasonal droughts in
China have long prevailed over China, particularly
over the past few decades. For emphasis, temperature
variations reported in this study for Henan Province
utilized seasonal summer data due to the data acqui-
sition date that presents an appropriate mean maxi-
mum and minimum temperatures for further analysis.
Fig. 6 shows mean maximum (33.7ºC) and minimum
(20.9ºC) temperatures during summer over the last 40
years. Overall, the annual average temperature for the
study domain is 15.6ºC.
Evaluation of remotely-sensed indices for Henan
Province
Enhanced Vegetative Index (EVI)
Fig. 7 presents spatiotemporal dynamics of vegeta-
tion health and density across Henan Province be-
tween 1983 and 2023. Observations from the spa-
tial analysis identify areas with dark green patches
as high EVI hotspots, whilst light green, yellowish
and dark brown zones mark areas with moderate
to low EVI spots. A close observation shows how
most EVI hotspot zones within the far western, east-
ern, and southernmost parts of Henan Province like
Xinyang, Yongcheng, Zhoukou, Shuizhai, etc. have
several water bodies, mainly rivers, in their catch-
ments. Additionally, it is worth noting that most parts
of western Henan, with dark green areas are char-
acterized by elevation or high altitudes with dense
Sarfo, I., Qiao, J., Yeboah, E., Okrah, A. , El Rhadiouini, Ch., Osibo, B.K., Boah, A., Ben Amara, D. (2024). Causal effects and prediction of land
use systems in rural landscapes.. . Acta Sci. Pol., Formatio Circumiectus, 23 (3), 27–56. DOI: http://dx.doi.org/10.15576/ASP.FC/190971
36 www.acta.urk.edu.pl
vegetation coverage. Other areas displayed stable
or uctuating EVI trends concurrently, suggesting
a complex interplay of various factors such as land
use changes driven by economic impulses that move
beyond population growth, such as urbanization and
industrialization, among other rural revitalization
strategies, as well as climate-induced and conserva-
tion parameters. A close observation shows a mix of
EVI patterns, coinciding with rapid urban sprawl and
infrastructural development during the 1993–2003
period. The 2003–2013 and 2013–2023 regimes re-
ected complex vegetation transitions in response to
urbanization, industrialization and ecological civili-
zation initiatives.
Normalized Difference Built-Up Index (NDBI)
Assessment of NDBI (Fig. 8) in the area revealed sig-
nicant uctuations of built environment/settlements
and urbanization trends. The generated NDBI values
(Fig. 8) demonstrated spatial variability across the
province, reecting the varying degrees of urban ex-
pansion and LUC. NDBI slowed down between 1983
and 1993 due to prolonged dryness, droughts and
famine that occurred between the 1980s and 1990s
across the globe. This propelled a shift in other land
use forms like farmlands/shrubs and migration in
subsequent windows; thus, in the 1993–2003 period.
This corresponds to rural development in subsequent
periods, and spatial trends observed in Figs. S1–S2.
A considerable increase in NDBI for 1993–2003 and
Fig. 6. LST variations for Henan Province during summer (1983–2023) (source: Authors’ own elaboration)
Sarfo, I., Qiao, J., Yeboah, E., Okrah, A. , El Rhadiouini, Ch., Osibo, B.K., Boah, A., Ben Amara, D. (2024). Causal effects and prediction of land
use systems in rural landscapes.. . Acta Sci. Pol., Formatio Circumiectus, 23 (3), 27–56. DOI: http://dx.doi.org/10.15576/ASP.FC/190971
37
www.acta.urk.edu.pl
subsequent decades arm evidence of urbanization,
economic growth and infrastructure expansion, driv-
en by an array of policy-driven factors. Dark red
zones mark areas with high urban concentration (i.e.,
cities and urban areas) like Zhengzhou, Shangqiu,
Luoyang, Nanyang, Xinyang, Xinxiang, Hebi, Jia-
zuo, Anyang, etc. Areas with yellow, cyan, green and
magenta color patches mark zones with moderate
built environment to areas with low built-up concen-
tration (i.e., sub-urban, rural and areas covered by
water bodies).
Normalized Difference Water Index (NDWI)
Fig. 9 illustrates that areas covered by water bodies
have undergone expansion. Dark blue areas exhibit
areas with high NDWI, whereas green to light yel-
low/green zones indicate areas with moderate or low
NDWI. Continuous and recurrent changes in NDWI
can be observed across Henan Province. NDWI is
high around Huanglongsi, Kaifeng, Luoyang, Jiyu-
an, Zhumadian, Nanyang and its environs, due to the
presence of major rivers as illustrated in the study
area map (Fig. 1). A steady increase was observed
between 1983 and 2003 due to extreme climatic
conditions experienced around the globe. However,
NDWI plummeted between 2003 and 2023 due to
the unprecedented heavy rains and historical ood-
ing event that occurred in Zhengzhou (i.e., the pro-
vincial capital) and across Henan province between
July 17 and 31 2021. This historical event had dire
Fig. 7. EVI variations in Henan Province (1983–2023) (source: Authors’ own elaboration)
Sarfo, I., Qiao, J., Yeboah, E., Okrah, A. , El Rhadiouini, Ch., Osibo, B.K., Boah, A., Ben Amara, D. (2024). Causal effects and prediction of land
use systems in rural landscapes.. . Acta Sci. Pol., Formatio Circumiectus, 23 (3), 27–56. DOI: http://dx.doi.org/10.15576/ASP.FC/190971
38 www.acta.urk.edu.pl
consequences by inundating hundreds of villages,
claiming lives and driving several economic losses
worth billions. Early studies (Chen et al., 2022; Hsu
et al., 2023) had reported human driven climate-in-
duced factors amplied the rain that fell. Expansion
in Henan Province’s water bodies can be attributed
to climate-induced stressors such as temperature rise,
resulting in the expansion of water bodies and expan-
sion of irrigational projects, aimed to enhance indus-
trialization, granary and agricultural productivity in
the region. Again, regrowth in forested areas serves
as cover for water bodies, hence, somewhat inuenc-
ing these expansions.
Normalized Difference Bareness Index (NDBaI)
Fig. 10 provides insights into changes in barren areas
or impervious surfaces within the province in Central
China. NDBaI (Fig. 10) exhibited spatial heteroge-
neity across the province, reecting varying degrees
of bareness/imperviousness in dierent areas. A gen-
eral decline in barren areas based on spatial analysis
presented in Fig. 5 indicates bare land has been re-
placed by the natural vegetation through ecological
restoration/greener landscapes, and by built-up areas
between 1983 and 2023. The continuous variations
in NDBaI are in tandem with urbanization trends
(Figs. S1–S2). Fig. S1 depicts areas with white and
dark brown patches, indicating high NDBaI. Simi-
larly, areas with high altitudes in the western part of
Fig. 8. NDBI variations in Henan Province (1983–2023) (source: Authors’ own elaboration)
Sarfo, I., Qiao, J., Yeboah, E., Okrah, A. , El Rhadiouini, Ch., Osibo, B.K., Boah, A., Ben Amara, D. (2024). Causal effects and prediction of land
use systems in rural landscapes.. . Acta Sci. Pol., Formatio Circumiectus, 23 (3), 27–56. DOI: http://dx.doi.org/10.15576/ASP.FC/190971
39
www.acta.urk.edu.pl
Henan Province can be observed, particularly those
marked by dark brown spots. Barren areas are being
cleared for farming or construction purposes around
water bodies, in and around cities such as Nanyang,
Huazhou, Xinyang, Jiyuan, Pingdingshan, and their
surrounding areas. On the other hand, domains with
low barren surfaces are indicated by green and light
brown spots.
Temporal causal inference
The patterns depicted in Figure 11 reveal a strong in-
direct convergence (Fig. 11a–b) between built-up ar-
eas and LST, underscoring built-up areas’ profound
impact on LST intensity/patterns. A moderate indirect
convergence is evident (Fig. 11c–d) between bare
land and LST, signifying the moderate inuence of
bare land on LST. While an indirect convergence is
observed in terms of the impact of water bodies on
LST, the extent of their signicant inuence on LST
intensity in Henan Province remains unclear, as illus-
trated in Fig. 11e. Notably, there is substantial indirect
convergence (Fig. 11f) between forest and LST, high-
lighting the dominant eect of forests on LST. Fur-
thermore, a strong indirect convergence (Fig. 11g) is
observed between farmlands/shrubs, and LST, indi-
cating that farmlands/shrubs play a regulatory role in
inuencing LST.
Fig. 9. NDWI variations in Henan Province (1983–2023) (source: Authors’ own elaboration)
Sarfo, I., Qiao, J., Yeboah, E., Okrah, A. , El Rhadiouini, Ch., Osibo, B.K., Boah, A., Ben Amara, D. (2024). Causal effects and prediction of land
use systems in rural landscapes.. . Acta Sci. Pol., Formatio Circumiectus, 23 (3), 27–56. DOI: http://dx.doi.org/10.15576/ASP.FC/190971
40 www.acta.urk.edu.pl
Spatial causal inference
In Fig. 12 (a, b), a positive cross-mapping skill val-
ue (ρ = 0.63) indicates a robust positive indirect
cross-mapping relationship between built-up areas
and LST. This implies that the presence or attributes
of built-up areas can be utilized to predict or elucidate
the patterns of LST to a considerable extent. Further-
more, a positive value (ρ = 0.32) in Fig. 12 (a, c) sig-
nies a moderate positive indirect cross-mapping re-
lationship between bare land and LST. This suggests
a moderate positive association between the charac-
teristics of bare land and the LST phenomenon.
Nevertheless, a negative cross-mapping skill val-
ue (ρ = –0.36) indicates a moderate negative indirect
cross-mapping relationship in Fig. 12 (a, d) between
areas covered by water bodies and LST. This sug-
gests that areas covered by water bodies might have a
cooling eect on LST, contributing to lower LST val-
ues. Additionally, a negative value of (ρ = –0.12) in
Fig. 12 (a, e) indicates a negative indirect cross-map-
ping relationship between forest and LST. This im-
plies that areas covered by forests might have a cool-
ing eect on LST, contributing to lower LST values.
It is worth noting that dierent types of vegetation
have varying degree or cooling eects on LST. Fur-
thermore, a negative value of (ρ = –0.27) in Fig. 12
(a, f) suggests a relatively moderate negative indi-
rect cross-mapping relationship between farmlands/
shrubs and LST. This implies that LST patterns are
inversely related to the characteristics of farmlands,
and farmlands might contribute to lowering LST in-
tensity in the area.
Fig. 10. NDBaI variations in Henan Province (1983–2023) (source: Authors’ own elaboration)
Sarfo, I., Qiao, J., Yeboah, E., Okrah, A. , El Rhadiouini, Ch., Osibo, B.K., Boah, A., Ben Amara, D. (2024). Causal effects and prediction of land
use systems in rural landscapes.. . Acta Sci. Pol., Formatio Circumiectus, 23 (3), 27–56. DOI: http://dx.doi.org/10.15576/ASP.FC/190971
41
www.acta.urk.edu.pl
Fig. 11. Temporal causal inference of LULCC variable and LST: Built Up -LST causality and xmap (a, b); Bare land-LST
causality and xmap (c, d); Water bodies-LST causality (e); forest-LST causality (f) and Farmlands and Shrubs-LST causality
(g) (source: Authors’ own elaboration)
a) b)
c) d)
e) f) g)
Fig. 12. Spatial causal inference of LULCC variable and LST: Built Up-LST causality and xmap (a, b); Bare land-LST causality and xmap (c, d); Water bodies-LST causality (e, f); forest-LST
causality (g, h) and Farmlands and shrubs-LST causality (i, j) (source: Authors’ own elaboration)
Sarfo, I., Qiao, J., Yeboah, E., Okrah, A. , El Rhadiouini, Ch., Osibo, B.K., Boah, A., Ben Amara, D. (2024). Causal effects and prediction of land
use systems in rural landscapes.. . Acta Sci. Pol., Formatio Circumiectus, 23 (3), 27–56. DOI: http://dx.doi.org/10.15576/ASP.FC/190971
43
www.acta.urk.edu.pl
Land-use predictions for Henan Province (2033–
2053)
The simulations (Fig. 13 and Table 6) show built-up
and farmlands/shrubs will expand at a rate of 117.3%
(with 3.9% increment annually) and 10.7% (with 0.4%
annual expansion), respectively, over the next three
decades. Conversely, forests, barren and areas cov-
ered by water bodies will undergo a steady decline by
61.6% (at a 2.1% reduction rate each year), 39.7% (at
a 1.3 decreasing rate annually) and 4.2% (at a 0.1%
reduction rate annually), respectively, over the same
period.
Table 6. Area coverage, temporal variations, rate and magnitude of change for each class in Henan Province (2023–2053)
(source: Authors’ own elaboration)
Area Coverage for each class (sq. km) (2023–2053)
Class/Period 2033 2043 2053
Farmlands/shrubs 106,398 113,337 110,500
Bare land 1,712 1,471 1,290
Built–up areas 13,226 18,829 22,671
Forests 24,837 15,837 13,252
Water bodies 20,827 17,526 19,287
Temporal variations
Class/Period 2023–2033 2033–2043 2043–2053 2023–2053
Farmlands/shrubs +6.6 +6.5 –2.5 +10.7
Bare land –19.9 –14.1 –12.3 –39.7
Built–up areas +26.8 +42.4 +20.4 +117.3
Forests –27.9 –36.2 –16.3 –61.6
Water bodies +3.5 –15.9 –10.1 –4.2
Rate and magnitude of change
2023–2053
Class/Period 2023 2053 Magnitude of ∆
(km2)
Magnitude of ∆
(km2)/yr Rate of ∆/yr (%)
Farmlands/shrubs 99808 110,500 +10,692 +356.4 +0.4
Bare land 2139 1,290 –849 –28.3 –1.3
Built–up areas 10432 22,671 +12239 +407.9 +3.9
Forests 34488 13,252 –21236 –707.9 –2.1
Water bodies 20133 19,287 –846 –28.2 –0.1
***Total area coverage (km2) (Absolute) = 167,000
Sarfo, I., Qiao, J., Yeboah, E., Okrah, A. , El Rhadiouini, Ch., Osibo, B.K., Boah, A., Ben Amara, D. (2024). Causal effects and prediction of land
use systems in rural landscapes.. . Acta Sci. Pol., Formatio Circumiectus, 23 (3), 27–56. DOI: http://dx.doi.org/10.15576/ASP.FC/190971
44 www.acta.urk.edu.pl
DISCUSSION
Spatial distribution of land use systems in Henan
Province
Scientic studies have extensively explored land use
systems and climate variability within the context of
sustainable development goals (SDGs) such as 11, 13,
and 15. However, there is a relative scarcity of stud-
ies that thoroughly examine the interplay between
remotely-sensed indices, policy-driven options, and
causal analysis over long periods at the provincial
or regional level. This is particularly true for Henan
Province in Central China, which is characterized by
ecological livability, rich cultural heritage, rural civ-
ilization, industrial prosperity, and immense agricul-
tural productivity. Between 1983 and 2023, land use
transitions in Henan Province, as shown in Table 4,
reveal a gradual decline in farmlands/shrubs at an an-
nual rate of 0.5%. Similarly, barren areas have expe-
rienced a decline of 1.5% over the past four decades.
On the other hand, built-up areas, forests, and water
bodies have expanded signicantly during the same
period, with annual growth rates of 12.5%, 1.3%, and
2.1%, respectively. While farmlands/shrubs were the
dominant land use class in 1983, followed by forests
and water bodies, there has been a fundamental shift in
recent years. The expansion of built-up areas has been
particularly exponential, despite being the least dom-
Fig. 13. Land-use predictions for Henan Province over the next three decades (2033–2053) (source: Authors’ own elab-
oration)
Sarfo, I., Qiao, J., Yeboah, E., Okrah, A. , El Rhadiouini, Ch., Osibo, B.K., Boah, A., Ben Amara, D. (2024). Causal effects and prediction of land
use systems in rural landscapes.. . Acta Sci. Pol., Formatio Circumiectus, 23 (3), 27–56. DOI: http://dx.doi.org/10.15576/ASP.FC/190971
45
www.acta.urk.edu.pl
inant class in 1983. In the following sections, we will
comprehensively evaluate land metrics and remote-
ly-sensed indices, as well as the driving mechanisms
behind these changes. We will also review relevant lit-
erature to support our ndings.
Driving mechanisms of LULCC and LST based on
causality analysis and relevant literature
Based on the results presented in sections 3.1 and 3.5,
it is clear that Henan Province’s land use and land cov-
er is inuenced by a combination of factors, including
socio-cultural, political, economic, and environmental
factors. While there was some convergence between
the studied variables of LULCC and LST based on
the results of the temporal causal inference analysis, it
was found that built-up areas have the most signicant
impact on the intensity of LST in Henan Province. The
inuence of other variables such as forests, farmlands/
shrubs, and water bodies on LST intensity in the study
area remains unclear, especially when compared to the
impact of built-up areas, which amplify LST. It is im-
portant to note that the type of vegetation plays a criti-
cal role in either triggering cooling eects or minimiz-
ing the eects of LST. This nding is consistent with
the research conducted by Jin et al. (2020), who inves-
tigated the eect of vegetation variation on surface air
temperature. Similarly, Sarfo et al. (2022) found that
unlike forests, farmlands/shrubs have a minimal inu-
ence on LST, which aligns with the ndings of this
study. The analysis of spatial causal inferences (Fig. 12
a–c) shows a moderate to strong positive cross-map-
ping relationship between bare land, built-up areas,
and LST. This suggests that the characteristics of these
land features can be used to understand surface tem-
perature patterns to a large extent. On the other hand,
weak to moderate inverse cross-mapping values were
observed for water bodies, forests, farmlands/shrubs,
and LST. Contextually, these features have some reg-
ulatory impact on LST patterns/intensity (Fig. 12 e–j),
but it is minimal. Interestingly, a closer examination of
Fig. 12 (i) reveals that in some cases, farmlands/shrubs
contribute to LST, possibly due to the presence of
built-up areas or settlements around cultivated lands.
Based on the existing literature and the results of
land metrics and causality analysis, the changes in
Henan Province’s land use systems can be largely
attributed to population growth/distribution and pol-
icy-driven options related to the socio-economic and
ecological development of the area. Biophysical pa-
rameters such as climate stressors (e.g., ooding) and
terrain/topographical factors have also played a role
in inuencing these unprecedented changes. Between
1983 and 1993, initiatives were undertaken as part of
the People’s Republic of China’s 40-year economic
reform and opening-up policy. For example, directives
such as the “preferential regional development strat-
egy in the 1980s,” “the 1992 Hinterland City open-
ing-up policy,” and the “Launching of urban greening
regulation” led to the conversion of cultivated lands
into built-up areas, settlements, and ecological/green
landscapes (Mu et al., 2016). This trade-o of cultivat-
ed lands for infrastructure and economic development,
particularly in the western, eastern, and central parts of
China compared to the northern and southern regions,
was aimed at bridging regional development gaps.
This, in turn, resulted in changes in microclimatic
conditions and increased settlements due to migration
and high fertility/growth rates. The period from 1993
to 2003 witnessed a rapid rate of urbanization and
industrialization. This was largely driven by increas-
ing support and planning policies from the national/
central government. State-driven policies, such as the
1994 Basic Agricultural Land Protection Regulations,
the 1998 Requisition-Compensation Balance of Ara-
ble Land Policy, the 1999 “Grain-to-Green” policy, the
National Landscape Garden City Policy, and the 2003
Coordinated Regional Development, led to a recon-
guration of the urban-rural landscape based on the
Central Government’s “ve-coordinated strategies.”
These policies resulted in diverse development strat-
egies and infrastructure development, with a focus on
rural road construction/improvement in line with the
saying “If you want to get rich, build roads rst.” As
a result, barren areas, forests, and some water bodies
were transformed into built-up areas and irrigational
farms (Camille, 2020).
In the post-2000 era, specically from 2003 to
2013, various regional policies were implemented in
the province. These policies include the “2006 Requi-
sition-Compensation Balance of Arable Land Policy”,
“2012 Urban-Rural Integration and Ecological Civi-
lization Initiative”, “Central Plain Urban Agglomer-
ation (CPUA) Development Planning (2006–2010)”,
“Zhengzhou-Kaifeng New Urban Area Develop-
Sarfo, I., Qiao, J., Yeboah, E., Okrah, A. , El Rhadiouini, Ch., Osibo, B.K., Boah, A., Ben Amara, D. (2024). Causal effects and prediction of land
use systems in rural landscapes.. . Acta Sci. Pol., Formatio Circumiectus, 23 (3), 27–56. DOI: http://dx.doi.org/10.15576/ASP.FC/190971
46 www.acta.urk.edu.pl
ment”, and “Forest Eco-City Planning”. These poli-
cies had a positive impact on the social stratication,
economic welfare, and overall Gross Domestic Prod-
uct (GDP) of both rural and urban residents. Addition-
ally, early studies conducted by Mu et al. (2016), Ca-
mille (2020), and Li et al. (2022) revealed that these
initiatives signicantly inuenced and reshaped the
ecological landscape and security patterns of the prov-
ince. As a result of these strategies, settlements were
transformed, spontaneous migration occurred, and
land and other resources were redistributed, all in the
context of rural development and land consolidation
eorts. From 2013 to 2023, the focus shifted to the
“2017 rural revitalization initiatives”, aimed at eradi-
cating poverty among rural residents living below the
absolute or extreme poverty line. During this period,
emphasis was also placed on the “Rising of Central
China”, the connectivity of suburban areas to major
cities, and the planning of the “Inter-City Railway
Network in the CPUA”. The Central Plain Economic
Region (CPER), as described by Chen et al. (2022)
and Hsu et al. (2023), played a vital role in the overall
development of the province.
Future of Henan Province based on land use pre-
dictions (2023–2053)
The continuous expansion of the built environment
and the changing trajectory of farmland and shrub
cover types indicate a need for a comprehensive and
actionable roadmap to ensure sustainable growth in
all aspects of the province. The central government
aims to promote equitable economic growth through
urbanization, industrialization, and rural revitaliza-
tion. However, without regulation, the current eco-
nomic and policy-driven initiatives that prioritize
land development over other gains could have devas-
tating consequences for farmland and shrub-covered
areas in the province, as shown in Figure 13 simu-
lations. These ndings support the claims made by
Zhao et al. (2021) in their study on “China’s future
food demand and its implications for trade and the
environment.” Spatial analysis highlights the urgent
need to regulate urbanization (Wang et al., 2021)
and other economic trade-os involving land and re-
sources for infrastructure development, as this will
have long-term impacts on agricultural productivi-
ty. Additionally, Wang et al. (2021) have projected
a steady increase in built-up areas by 2050, which
in turn leads to a decline in pristine environments.
These ndings align with the current study’s predic-
tions of expansion in built-up areas and the reduction
in forests, water bodies, and bare land, as farmland
and shrub cover increase.
CONCLUSIONS
We conducted a study in Henan Province using an
indirect causality approach to explore the eects of
dierent land use systems on land surface tempera-
ture (LST). This region, which is predominantly ru-
ral, plays a crucial role in China’s socio-economic and
environmental development. Additionally, we predict-
ed future land use patterns for the next three decades
based on current trends. The key ndings are summa-
rized as follows:
1. Henan Province is currently undergoing signi-
cant changes in its land use systems. The areas oc-
cupied by farmlands and shrubs are experiencing
a decline of 20.81%. These unprecedented chan-
ges are mainly driven by factors such as popula-
tion growth and distribution, industrialization, and
policy-driven options. They are also closely lin-
ked to the Central Government’s economic reform
and opening-up policy, which has been in eect
for the past 40 years.
2. Currently, there is a noticeable regrowth or green-
ing happening, which can be attributed to spatial
trends and policy directives like “The ecological
civilization initiative” implemented by the Cen-
tral Government, as well as regional and local
authorities.
3. The analysis of temporal causal inference, speci-
cally the causal analysis, revealed a robust associ-
ation between built-up areas and surface tempera-
ture. This nding provides substantial evidence to
support the inuential role of built-up areas on the
intensity of land surface temperature.
4. Results from the spatial causal inference anal-
ysis shows moderate to robust positive indirect
cross-mapping relationships between built-up
areas (ρ = 0.63) and bare land (ρ = 0.32) against
LST. This implies that the characteristics of built-
up and barren areas can be utilized to predict or
elucidate LST patterns to a considerable extent
Sarfo, I., Qiao, J., Yeboah, E., Okrah, A. , El Rhadiouini, Ch., Osibo, B.K., Boah, A., Ben Amara, D. (2024). Causal effects and prediction of land
use systems in rural landscapes.. . Acta Sci. Pol., Formatio Circumiectus, 23 (3), 27–56. DOI: http://dx.doi.org/10.15576/ASP.FC/190971
47
www.acta.urk.edu.pl
in Henan Province. Contrarily, weak to moder-
ate negative indirect cross-mapping relationships
were observed for forests (ρ = –0.12), farmlands/
shrubs (ρ = –0.27) and water bodies (ρ = –0.36).
This implies they somewhat play regulatory
roles in reducing LST intensity in some areas.
Interestingly, it is worth noting that dierent types
of vegetation have varying degrees of inuence on
LST patterns/intensity.
5. Land use predictions over the next three deca-
des show continuous expansion in built environ-
ment (+117.3%), a reversed trend in farmlands/
shrubs (+10.7%) currently under decline. Forests
(–61.6%), bare land (–39.7%) and areas covered
by water bodies (–4.2%) are expected to reduce.
To this end, understanding these dynamics is essen-
tial for safeguarding and managing rural landscapes,
and enhancing Henan Province’s environmental and
agricultural informatization. Causality and relevant
literature analyses demonstrate that shifts in Henan’s
land use systems moves beyond identifying popula-
tion growth and distribution, and urbanization as the
sole driving mechanisms of Henan’s land-use/envi-
ronmental changes. Instead, a combination of factors
is responsible for these shifts. Furthermore, ndings
will facilitate the understanding of forecasting future
land use patterns, based on current trends to optimize
resource utilization and the need to address institu-
tional, economic and environmental impacts emanat-
ing from these undesired changes. In future studies,
researchers can utilize emerging digital technologies,
such as innovative machine and deep learning ap-
proaches, to develop early warning systems. These
systems could further be used to comprehensively
investigate the spatiotemporal dynamics of land use
systems in both high and emerging economic nations.
Additionally, further studies could assess the dynam-
ics of land use among rural residents, particularly in
the context of rural revitalization and sustainability
concerns.
DECLARATION OF COMPETING INTEREST
The authors declare that they have no known compet-
ing nancial interests or personal relationships that
could have appeared to inuence the work reported in
this paper.
DATA AVAILABILITY
The data that backs up the study’s conclusions is ac-
cessible with the appropriate link provided in the
methodology section.
CODE AVAILABILITY
Codes used for this study will be made available upon
reasonable request.
REFERENCES
Ahmed, B., Zhu, X., Rahman, S., Choi, K. (2019).
Simulating land cover changes and their impacts
on land surface temperature in Dhaka, Bangladesh.
Remote Sensing, 5 (11), 5969–5998. DOI: 10.3390/
rs5115969
Avdan, U., Jovanovska, G. (2016). Algorithm for auto-
mated mapping of land surface temperature using
LANDSAT 8 satellite data. Journal of Sensors, 1–8.
DOI: 10.1155/2016/1480307
Al, A., Rahman, S., Faisal, A. (2020). Modelling future
land use land cover changes and their impacts on
land surface temperatures in Rajshahi, Bangladesh.
Remote Sensing Applications: Society and Environ-
ment, 18, 100314. DOI: 10.1016/j.rsase.2020.100314
Asgarian, A., Amiri, B.J., Sakieh, Y. (2014). Assessing
the eect of green cover spatial patterns on urban
land surface temperature using landscape metrics ap-
proach. Urban Ecosyst, 18, 209–222. DOI: 10.1007/
s11252-014-0387-7
Aznar-Sánchez, J.A., Piquer-Rodríguez, M., Velasco-
-Muñoz, J.F., Manzano-Agugliaro, F. (2019). World-
wide research trends on sustainable land use in agri-
culture. Land Use Policy, 87, 104069, 1–15. DOI:
10.1016/j.landusepol.2019.104069
Camille, B. (2020). Poverty alleviation in China: The rise of
state-sponsored corporate paternalism. China Perspect.,
3, 47–56. DOI: 10.4000/chinaperspectives.10456
Coll, C., Galve, J.M., Sanchez, J.M., Caselles, V. (2010).
Validation of landsat-7/ETM+ thermal-band calibration
and atmospheric correction with ground-based measure-
ments. IEEE T Geosci Remote, 48 (1), 547–555. DOI:
10.1109/TGRS.2009.2024934
Chen, Z., Kong, F., Zhang, M. (2022). A case study of the
“7–20” extreme rainfall and ooding event in Zhengzhou,
Henan Province, China from the perspective of fragmen-
tation. Water, 14(19), 2970. DOI: 10.3390/w14192970
Sarfo, I., Qiao, J., Yeboah, E., Okrah, A. , El Rhadiouini, Ch., Osibo, B.K., Boah, A., Ben Amara, D. (2024). Causal effects and prediction of land
use systems in rural landscapes.. . Acta Sci. Pol., Formatio Circumiectus, 23 (3), 27–56. DOI: http://dx.doi.org/10.15576/ASP.FC/190971
48 www.acta.urk.edu.pl
Cao, X., Liu, Y., Li, T., et al. (2019). Analysis of spatial pat-
tern evolution and inuencing factors of regional land
use eciency in China based on ESDA-GWR. Sci Rep,
9, 520. DOI: 10.1038/s41598-018-36368-2
Congalton, R.G. (1991). A review of assessing the accuracy
of classications of remotely sensed data. Remote Sens.
of Environ., 37(1), 35–46.
Değermenci, A.S. (2023). Spatio-temporal change analysis
and prediction of land use and land cover changes us-
ing CA-ANN model. Environ Monit Assess, 195, 1229.
DOI: 10.1007/s10661-023-11848-9
FAO (Food and Agricultural Organization) (2020). Land
use in agriculture by the numbers. https://www.fao.org/
sustainability/news/detail/en/c/1274219/ (accessed: Jan-
uary 9, 2024).
Gao, B., Yang, J., Chen, Z., et al. (2023). Causal inference
from cross-sectional earth system data with geographi-
cal convergent cross mapping. Nat Commun., 14, 5875.
DOI: 10.1038/s41467-023-41619-6
Girma, R., Fürst, C., Moges, A. (2021). Land use land
cover change modeling by integrating articial neural
network with cellular Automata-Markov chain mod-
el in Gidabo river basin, main Ethiopian rift. Envi-
ronmental Challenges, 6, 100419. DOI: 10.1016/j.envc.
2021.100419
Godfray, H.C.J., Garnett, T. (2014). Food security and sus-
tainable intensication. Phil. Trans. R. Soc. B., 369
(1639). DOI: 10.1098/rstb.2012.0273
Hagan, D.F.T., Wang, G., San Liang, X., Dolman, H.A.J.
(2019). A time-varying causality formalism based on
the Liang–Kleeman information ow for analyzing di-
rected interactions in nonstationary climate systems.
J. Clim., 32(21), 7521–7537. DOI: 10.1175/JCLI-D-
-18-0881.1
HPBS (Henan Province Bureau of Statistics) (2018). Sta-
tistical Yearbook of Henan Province. China Statis-
tics Press, China. (in Chinese). https://tjj.henan.gov.
cn/2018/02-27/1373044.html (accessed: July 18, 2023).
Hsu, P., Xie, J., Lee, J., Zhu, Z., Li, Y., Chen, B., Zhang, S.
(2023). Multiscale interactions driving the devastat-
ing oods in Henan Province, China during July 2021.
Weather. Clim. Extremes, 39, 100541. DOI: 10.1016/j.
wace.2022.100541
Huang, D., Huang, J.., Liu, T. (2019). Delimiting urban
growth boundaries using the CLUE-S model with vil-
lage administrative boundaries. Land Use Policy, 82,
422–435. DOI: 10.1016/j.landusepol.2018.12.028
Hinz, R., Sulser, T.B., Huefner, R., Mason‐D’Croz, D., Dun-
ston, S., Nautiyal, S., et al. (2020). Agricultural devel-
opment and land use change in India: A scenario analy-
sis of trade‐os between UN Sustainable Development
Goals (SDGs). Earth’s Future, 8, e2019EF001287. DOI:
10.1029/2019EF001287
Jin, K., Wang, F., Zong, Q., Qin, P., Liu, C. (2020). Impact of
variations in vegetation on surface temperature change
over the Chinese Loess Plateau. Sci. Total Environ., 716,
136967. DOI: 10.1016/j.scitotenv.2020.136967
Kanianska, R. (2016). Agriculture and its impact on land-
use, environment and ecosystem services. IntechOpen,
1–25. DOI: 10.5772/63719
Li, C., Li, F., Wu, Z., Cheng, J. (2017). Exploring spatially
varying and scale-dependent relationships between soil
contamination and landscape patterns using geograph-
ically weighted regression. Appl. Geogr., 82, 101–114.
DOI: 10.1016/j.apgeog.2017.03.007
Li, B., Cao, X., Xu, J., Wang, W., Ouyang, S., Liu, D.
(2021). Spatial–temporal pattern and inuence fac-
tors of land used for transportation at the county lev-
el since the implementation of the reform and open-
ing-up policy in China. Land, 10, 833. DOI: 10.3390/
land10080833
Li, S., Qin, Z., Zhao, S., et al. (2022). Spatiotemporal vari-
ation of land surface temperature in Henan Province of
China from 2003 to 2021. Land, 11, 1104. DOI: 10.3390/
land11071104
Magliocca, N.R., Dhungana, P., Sink, C.D. (2023). Review
of counterfactual land change modeling for causal in-
ference in land system science. J Land Use Sci, 18 (1),
1–24. DOI: 10.1080/1747423X.2023.2173325
Meyfroidt, P., de Bremond, A., Ryan, C.M., Archer, E., As-
pinall, R., Chhabra, A., et al. (2022). Ten facts about
land systems for sustainability. Proc. Natl. Acad. of Sci.
U.S.A, 119(7). DOI: 10.1073/PNAS.2109217118
Mu, B., Mayer, A.L., He, R., Tian, G. (2016). Land use dy-
namics and policy implications in Central China: A case
study of Zhengzhou. Cities, 58, 39–49. DOI: 10.1016/j.
cities.2016.05.012
Okrah, A., Prempeh, N.A., Mensah, C., John, R., Kumi, N.,
Otu-Larbi, F., Kyere-Boateng, R. (2020). Impact of spa-
tio-temporal land cover changes on land surface tem-
perature over Dormaa from 1990–2020. North American
Academic Research, 6(4), 87–104. DOI: 10.5281/zeno-
do.7838837
Reay, D.S. (2020). Land use and agriculture: Pitfalls and
precautions on the road to net zero. Front. Clim., 2, 4.
DOI: 10.3389/fclim.2020.00004
Sakieh, Y., Salmanmahiny, A. (2016). Performance assess-
ment of geospatial simulation models of land-use change
– A landscape metric-based approach. Environ Monit
Assess, 188, 169. DOI: 10.1007/s10661-016-5179-5
Sarfo, I., Qiao, J., Yeboah, E., Okrah, A. , El Rhadiouini, Ch., Osibo, B.K., Boah, A., Ben Amara, D. (2024). Causal effects and prediction of land
use systems in rural landscapes.. . Acta Sci. Pol., Formatio Circumiectus, 23 (3), 27–56. DOI: http://dx.doi.org/10.15576/ASP.FC/190971
49
www.acta.urk.edu.pl
Sarfo, I., Shuoben, B., Beibei, L., et al. (2022). Spatiotem-
poral development of land use systems, inuences and
climate variability in Southwestern Ghana (1970–2020).
Environ Dev Sustain, 24, 9851–9883. DOI: 10.1007/
s10668-021-01848-5
Turner, B.L., Meyfroidt, P., Kuemmerle, T., Müller, D.,
Chowdhury, R. (2020). Framing the search for a theo-
ry of land use. J Land Use Sci, 15(4), 489–508. DOI:
10.1080/1747423X.2020.1811792
Ullah, S., Ahmad, K., Sajjad, R.U., Abbasi, A.M., Nazeer, A.,
Tahir, A.A. (2019). Analysis and simulation of land cov-
er changes and their impacts on land surface temperature
in a lower Himalayan region. Journal of Environmen-
tal Management, 245, 348–357. DOI: 10.1016/j.jen-
vman.2019.05.063
Wang, Y., Vliet, J.V., Debonne, N., Pu, L., Verburg, P. (2021).
Settlements changes after peak population: Land systems
projections for China until 2050. Landsc Urban Plan, 209,
1–12. DOI: 10.1016/j.landurbplan.2021.104045
Xi, J., Zhou, R., Bu, R., Na, R., Guo, E. (2023). Analysis
of the causal relationship between the spatial change of
cultivated land conversion and economic development
in North China, using Hohhot City in Inner Mongolia as
an example. Pol. J. Environ. Stud., 32 (4), 3373–3383.
DOI: 10.15244/pjoes/162548
Xu, D., Zhang, K., Cao, L., Guan, X., Zhang, H. (2022).
Driving forces and prediction of urban land use change
based on the geodetector and CA-Markov model: A case
study of Zhengzhou, China. Int. J. Digit, 15, 1, 2246–
2267. DOI: 10.1080/17538947.2022.2147229
Yang, Q., Huang, X., Li, J. (2017). Assessing the relation-
ship between surface urban heat islands and landscape
patterns across climatic zones in China. Scientic Re-
ports, 7, 9337. DOI: 10.1038/s41598-017-09628-w
Zhao, H., Chang, J., Havlík, P., et al. (2021). China’s fu-
ture food demand and its implications for trade and en-
vironment. Nat Sustain, 4, 1042–1051. DOI: 10.1038/
s41893-021-00784-6
ABSTRAKT
ZALEŻNOŚCI PRZYCZYNOWO-SKUTKOWE I PRZEWIDYWANIE SYSTEMÓW UŻYTKOWANIA GRUNTÓW
W KRAJOBRAZACH WIEJSKICH: PRZYKŁADY Z PROWINCJI HENAN
Cel pracy
W rozwoju obszarów wiejskich i rolnictwa grunty odgrywają kluczową rolę w kontekście zwiększania
produktywności. Aby zrozumieć wpływ konkretnych czynników (przyczyn) lub kombinacji czynników na
wyniki, istotne jest zidentykowanie i ustalenie wyraźnych związków przyczynowych. W naszych bada-
niach analizujemy związki przyczynowo-skutkowe pomiędzy różnymi systemami użytkowania gruntów
a temperaturą powierzchni ziemi (LST) w prowincji Henan. Ponadto opracowujemy prognozy użytko-
wania gruntów w oparciu o obserwowane na bieżąco trendy. Zrozumienie tej dynamiki jest niezbędne do
poprawy stopnia informatyzacji rolnictwa, lepszego zarządzania środowiskiem i inteligentnych wyborów,
korzystnych dla klimatu – w lokalnych okręgach, powiatach i wioskach, w ważnych rolniczo regionach
Chin i poza nimi.
Materiał i metody
W badaniu wykorzystano zintegrowane dane, techniki teledetekcji i podejście przyczynowo-skutkowe do
zbadania systemów użytkowania gruntów LUS i LST w prowincji Henan. Następnie zaś Moduły do Oceny
Zmiany Użytkowania Gruntów (MOLUSCE) oraz automaty komórkowe – sztuczną sieć neuronową (CA-
-ANN), aby przewidzieć, jak będą się kształtować systemy LUS w najbliższej przyszłości (2023–2053).
Wyniki i wnioski
Wyniki badań wskazują, że tereny zabudowane (+500%), lasy (+50,88%) i zbiorniki wodne (+83,56%)
znacznie powiększyły swoją powierzchnię w ciągu ostatnich 40 lat. Jednocześnie powierzchnia obszarów
uprawnych (–20,81%) i jałowych (–60,53%) stale spada. Analiza wnioskowania czasowego wykazała silną
zbieżność między udziałem obszarów zabudowanych a temperaturą powierzchni gruntu (LST), co potwier-
dza głęboki wpływ zabudowy na intensywność LST. Analiza przestrzennych wnioskowań przyczynowych
pokazuje pozytywną korelację – od umiarkowanej do silnej – w zakresie pośrednich powiązań mapowania
Sarfo, I., Qiao, J., Yeboah, E., Okrah, A. , El Rhadiouini, Ch., Osibo, B.K., Boah, A., Ben Amara, D. (2024). Causal effects and prediction of land
use systems in rural landscapes.. . Acta Sci. Pol., Formatio Circumiectus, 23 (3), 27–56. DOI: http://dx.doi.org/10.15576/ASP.FC/190971
50 www.acta.urk.edu.pl
między terenami zabudowanymi (ρ = 0,63) i nieużytkami (ρ = 0,32) w odniesieniu do LST. Prognozy doty-
czące użytkowania gruntów (2023–2053) wskazują na zmniejszenie powierzchni lasów i zbiorników wod-
nych, zaś odwrotną tendencję w przypadku gruntów uprawnych. Informacje te są szczególnie ważne przy
formułowaniu ukierunkowanych dyrektyw politycznych niezbędnych do uregulowania zaburzonej równo-
wagi procesów użytkowania gruntów i unikania niepożądanych kompromisów gospodarczych.
Słowa kluczowe: analiza przyczynowo-skutkowa, geodetektor, użytkowanie gruntów i pokrycie terenu, tem-
peratura powierzchni lądu (LST), Chiny
APPENDIX
Table A.1. Description of land cover types identied in Henan Province (source: Authors’ own elaboration)
Class Denition
Forests Closely interwoven trees and lush vegetation dominate these areas. It also includes all vegetative
regions with no exposed soil.
Built-up areas Urban, business, and industrial regions. This category also includes community green spaces,
playing elds, and truck terminals.
Bare land Bare sections of soil or rocks that have not been covered by greenery. In and around built-up
regions, barren areas are noticeable. It constitutes terrains that have been cleared in preparation for
redevelopment or cultivation.
Farmlands and shrubs Widely distributed trees, hedges or bushes, secluded thickets, and non-tree crops.
Water bodies Rivers, lagoons, lakes, and other bodies of water are all part of this ecosystem.
Table A.2 presents k1 and k2 becoming coecients
determined by eective wavelength of a satellite sen-
sor based on these constants.
Table A.2. ETM+ and TM thermal band calibration con-
stants (source: according to Coll, 2010)
K1 (Wm–2 sr–1 μm–1) k2 (Kelvin)
Landsat 5 –TM 607.76 1260.56
Landsat 7 –ETM+ 666.09 1282.71
Conversion of Spectral radiance (Lλ) to Kelvin with
emissivity value from Landsat 8
Removal of atmospheric distortions from the thermal
infrared data was performed using ENVI 5.0 software
for the correction of thermal band 10 (Table A.3).
Table A.3. Band 10’s thermal constants (source: Avdan and
Jovanovska, 2016)
K11321.08
K2777.89
COMPUTATION OF REMOTELY-SENSED INDICES
Enhanced Vegetation Index (EVI)
Synonymous to Normalized Dierence Vegetation In-
dex (NDVI) is the EVI, used to quantify vegetation
greenness. However, EVI elaborates by making fur-
ther corrections to some atmospheric conditions and
canopy background noise, which are more sensitive
in areas with dense vegetation. It incorporates an “L”
value to adjust for canopy background, “C” values
Sarfo, I., Qiao, J., Yeboah, E., Okrah, A. , El Rhadiouini, Ch., Osibo, B.K., Boah, A., Ben Amara, D. (2024). Causal effects and prediction of land
use systems in rural landscapes.. . Acta Sci. Pol., Formatio Circumiectus, 23 (3), 27–56. DOI: http://dx.doi.org/10.15576/ASP.FC/190971
51
www.acta.urk.edu.pl
as coecients for atmospheric resistance, and values
from the blue band (B). These enhancements allow for
index calculation (Eqn. S1) as a ratio between the R
and NIR values, while reducing the background noise,
atmospheric noise, and saturation in most cases:
EVI G
NIRR
NIRC RC BL
12
... (S1)
Where:
G – gain factor (set at 2.5),
C1 and C2 – correct for aerosol resistance (set at 6 and
7.5, respectively),
L – adjusts for canopy background (set at one),
NIR, red and blue – reectance in the near-infrared,
red and blue wavelengths, respectively.
The gain factor (G), C1, C2, and L coecients were
all derived and optimized for the MODIS sensor on
the Terra and Aqua Satellites. Here, EVI is measured
as in Eqn. S2:
EVI
Band Band
Band Band Band
25 67
51
43
43 1
.
()
(.
)...
(S2)
Normalized Difference Built-Up Index (NDBI)
The NDBI is another signicant index used to detect
built-up areas, including urban and constructed surfac-
es. It utilizes reectance values from the near-infrared
(NIR) and short-wave infrared (SWIR) bands.
NDBI
SWIR NIR
SWIR NIR
()
()
... (S3)
We computed NDBI using the expression (Eqn. S4):
NDBI
Band Band
Band Band
54
54
... (S4)
Normalized Difference Bareness Index (NDBaI)
Similarly, NDBaI is used to identify bare surfaces, in-
cluding barren areas or soil using Eqn. S5:
NDBal
Band Band
Band Band
56
56
... (S5)
Normalized Difference Water Index (NDWI)
The NDWI is applied to examine water bodies. It takes
advantage of the RS image’s green and near-infrared
regions. It is vulnerable to land development and re-
sults in overestimated water bodies. NDWI products,
according to Xu (2007), can be used in connection to
vegetation index change products in order to exam-
ine the backdrop of a zone’s noticeable change. Water
bodies have low reectance. Only the visible fraction
of the electromagnetic spectrum is reected. Water
tends to reect more blue light (0.4–0.5 µm) than
green light (0.5–0.6 µm) and red light (0.6–0.7 µm).
Clear water has the highest reection in the visible
blue spectrum. As a result, water appears blue. In the
visible spectrum, turbid water has a greater reectiv-
ity. In the Near Infrared (NIR) and above, there is no
reectance. The following equations (Eqns. S6–S7)
were used to quantify water index:
NDWI NIRSWIR
NIRSWIR
()
()
... (S6)
MNDWIGreen SWIR
Green SWIR
()
()
... (S7)
EVI, NDBI, NDBaI and NDWI values range from –1
to 1, with higher values indicating high density and
lower values representing areas with least density for
each class.
Landscape metrics
Contextually, supervised classication was used to
categorize the land cover data into dierent classes.
This step involves assigning each pixel or area to a
specic land cover type (Ahmed et al., 2019). Depend-
ing on the study objectives, relevant land metrics like
the percentage of land cover types, fragmentation in-
dices, edge density, or diversity indices were calculat-
ed using Percentage of landscape (PLAND) (Eqn. S8)
(Yang et al., 2017).
PLAN
DA
pa
m
ip
100 / ... (S8)
where:
m – the number of patches in the landscape for class p;
aip – the area of patch ip;
Sarfo, I., Qiao, J., Yeboah, E., Okrah, A. , El Rhadiouini, Ch., Osibo, B.K., Boah, A., Ben Amara, D. (2024). Causal effects and prediction of land
use systems in rural landscapes.. . Acta Sci. Pol., Formatio Circumiectus, 23 (3), 27–56. DOI: http://dx.doi.org/10.15576/ASP.FC/190971
52 www.acta.urk.edu.pl
A – the total landscape area, which is a measure
of the proportion of the total area occupied by
a particular land-use type.
These metrics facilitate the understanding of spa-
tial distribution (Asgarian et al., 2014; Yang et al.,
2017) and conguration of land cover types in the
study area. The PLAND metrics signies the propor-
tion or relative contribution of a specic land cover
class or feature in the overall landscape. It provides
valuable information about the spatial distribution
and dominance of dierent land cover types within a
dened area. This metric is commonly used in envi-
ronmental studies, land use planning, and landscape
ecology to assess and understand the composition
and structure of landscapes. While correlations serve
as a useful tool in uncovering associations between
variables, it is crucial to acknowledge their inherent
symmetry, indicating they do not inherently indicate
the direction of inuence or establish causality (Ha-
gan et al., 2019).
Pearson’s correlation coefcient (r) analysis
A relationship between land metrics for dierent land
cover types and surface temperature was determined
using Pearson’s correlation coecient (r). The r coef-
cient was quantied using Eqn. S9:
r
xy
xi y
ii
i
()
()
(_ )( )
22
(S9)
where:
r – correlation coecient,
xi – values of the x-variable in the sample,
μ – average values of the x-variable,
yi – values of the y-variable in the sample,
δ – average values of the y-variable.
Correlation heatmap (Ullah et al., 2019) was fur-
ther generated to visualize the relationships between
the multiple variables in the dataset. It provides a vi-
sual representation of the correlation coecients (r)
between the pairs of variables, helping to identify pat-
terns, trends, and dependencies (Okrah et al., 2020).
Correlation heatmaps help to quickly identify which
variables are positively or negatively correlated. Pos-
itive correlations (values close to 1) denote two vari-
ables that tend to increase or decrease simultaneous-
ly, whilst negative correlations (values close to –1)
show that one variable tends to increase as the other
decreases.
Convergent Cross Mapping (CCM) model for
Temporal Causal Inference
Using Python-based Jupyter Notebook, we conducted
Convergent Cross Mapping (CCM) to infer temporal
indirect causality between LULCC variables and LST.
The initial step involved transforming the time series
data of the LULCC variable and LST into higher-di-
mensional spaces through data embedding, a crucial
process for revealing the system’s dynamics. Follow-
ing this, we carefully selected appropriate lag param-
eters, including embedding dimensions and time de-
lays, to optimize the CCM analysis. Subsequently, we
applied the CCM algorithm, calculating cross-map-
ping values for the target variable based on the histor-
ical time series of the potential explanatory variables
(i.e., for LULCC). Additionally, we computed cross-
mapped values for the potential explanatory variables
using the historical time series of the target variable
(LST) (Gao et al., 2023).
In the presented scenario, a specic nonlinear sys-
tem is explicitly dened, where 𝑋 inuences 𝑌, and
vice versa. The impact of 𝑋 on 𝑌 is determined by
a factor 𝛽𝑦, 𝑥, while the eect of 𝑌 on 𝑋 is inuenced
by a factor 𝛽𝑥, 𝑦. The constants 𝑟𝑥 and 𝑟𝑦 introduce
chaos to the system, with higher values leading to
more unpredictable behaviour, as expressed by equa-
tions S10 and S11:
𝑋(𝑡+1) =𝑋(𝑡) [𝑟𝑥−𝑟𝑥𝑋𝑡−𝛽𝑥, 𝑦𝑌(𝑡)] (S10)
𝑌(𝑡+1) =𝑌(𝑡) [𝑟𝑦−𝑟𝑦𝑌𝑡−𝛽𝑦, 𝑥𝑋(𝑡)] (S11)
Geographical Convergent Cross Mapping (GCCM)
model for Spatial Causal Inference
Furthermore, we employed Geographical Convergent
Cross Mapping (GCCM) in Python (Jupyter Note-
book) to evaluate the indirect causal relationships be-
tween LULCC variables and LST in a spatial context.
GCCM involves generating cross-maps to visualize
the interactions between the variables under inves-
tigation (Gao et al., 2023). Initially, we constructed
embeddings, setting the dimension of the embeddings
Sarfo, I., Qiao, J., Yeboah, E., Okrah, A. , El Rhadiouini, Ch., Osibo, B.K., Boah, A., Ben Amara, D. (2024). Causal effects and prediction of land
use systems in rural landscapes.. . Acta Sci. Pol., Formatio Circumiectus, 23 (3), 27–56. DOI: http://dx.doi.org/10.15576/ASP.FC/190971
53
www.acta.urk.edu.pl
as M. Subsequently, we iterated through each spatial
unit, identifying its spatial lags of dierent orders. We
then predicted Y based on X by dening a sequence of
library sizes and compiling those vectors into a matrix
based on their spatial orders. For each library size, we
predicted Y, as illustrated in Eqn. S12, by searching
for nearby points in the state space and constraining
them by the library size:
ˆ
YM wY M
sx si si x
i
M
()
...
1
1 (S12)
where:
s – a spatial unit at which the value of Y needs to be
predicted,
ˆ
Ys – the prediction result,
M – the number of dimensions of the embedding,
si – the spatial unit used in the prediction,
Ysi – the observation value at si and simultaneously
the rst component of a state in My, noted as ψ
(y, si).
Further, ψ (y, si) is determined by its one-to-one
mapping point ψ (x, si), which is in turn one of the
M + 1 nearest neighbours of the focal state ψ (x, s)
in Mx. wsi is the corresponding weight dened in Eqn.
S13.
wsi |M
weight xs xs
weight xs xs
x
i
i
i
M
((,),(,))
((,),(,))
...
1
1 (S13)
where weight (*,*) is the weight function between two
states in the shadow manifold, dened as Eqn. S14:
weight xs xs disxsxs
disxs
i
i
((,),(,))exp ((,),(,))
((,),
1(, ))xs
(S15)
where exp is the exponential function, and dis (*,*)
represents the distance function between two states in
the shadow manifold dened in Eqn. S16..
disxsxs
M
hx hx abs hx
hx
i
si ssik sk
((,),(,))
() () ()
,(
)
() ()
1
k
M
1
1
(S16)
where (*) means the absolute value of a real number,
and abs (*,*) represents the distance function between
two vectors, as the rst element hsi (x) in ψ(x, si) corre-
sponds to the spatial focal units, while other elements
in ψ(x, si) respectively correspond to a vector with se-
veral spatial units.
The concrete form of abs (*,*) for raster data and
polygon data are specied as absr and asbv in Eqn. S17
and Eqn. S18, respectively.
abs hxhx
D
ux
ux
rsik sk
si kd skd
d
D
() ()
(, )(,)
(),()
() ()
1 (S17)
abs hxhx
Dux
D
ux
vsik sk
si kd skd
d
D
() ()
(,
)(
,)
(),()
()
()
11
12
221 d
D
(S18)
where:
usi(k, d)(x) – the spatial unit of the kth-order spatial lags
of si in the direction d,
D – the number of spatial units (or directions)
in the kth-order,
The skill of cross-mapping prediction is measured
by the Pearson correlation coecient between the true
observations and corresponding predictions, dened
in Eqn. S19.
CovY Y
VarY VarY
(, )
()
()
(S19)
where Cov(Y, Y
ˆ ) represents covariance, and Var (Y, Y
ˆ )
represents variance.
We further examined the cross-maps to identify
patterns of convergence, searched for consistent pat-
terns in which changes in LULCC variables (causes)
precede changes in LST intensity (eects).
Other ndings
Rural and urban dynamics in Henan Province
Figure S1 depicts spatial distribution, and Figure S2
illustrates trends of rural and urban settings over the
ˆ
ˆ
Sarfo, I., Qiao, J., Yeboah, E., Okrah, A. , El Rhadiouini, Ch., Osibo, B.K., Boah, A., Ben Amara, D. (2024). Causal effects and prediction of land
use systems in rural landscapes.. . Acta Sci. Pol., Formatio Circumiectus, 23 (3), 27–56. DOI: http://dx.doi.org/10.15576/ASP.FC/190971
54 www.acta.urk.edu.pl
past 40 years in the studied domain. The illustrations
present area changes pertaining to the spatial distri-
bution of the region’s rurality and urbanism. Spatial
analysis shows urban areas have been on ascendency,
having expanded from 415.70 km2 in 1983 to 8,656.57
km2 (2023). Surprisingly, rural areas have undergone
constant uctuations over the same period. A spike in
rural areas’ trajectory could be observed between 1993
(1, 281.16 km2) and 2013 (2,221.52 km2), followed by
a sharp decline in 2023 (1,775.33 km2) which can be
attributed to some socio-political and economically or
policy-driven factors.
Dark red spots indicate high-density urban con-
centration hotspots like Zhengzhou, Huanglongsi,
Shangqiu, Luoyang, Nanyang, Xinyang, Xinxiang,
Anyang, etc. Rural or low-density urban concentration
zones are characterized by light green patches, located
around southern and western parts of Henan Province.
Rural areas are predominated by economic activities
such as granary/agriculture and industrial activities,
and are mostly characterized by elevation, terrain/
slope, aspects and water bodies in the aforementioned
areas.
Land metrics and LST
Statistical interpretations of land metrics/PLAND
analysis show quite a signicant inverse correlation
(–0.63, p < 0.05) between farmlands/shrubs (i.e.,
crops, orchards, thickets, etc.) and LST (Fig. S3).
Here, farmlands/shrubs obtained 46.1% coverage. As
Fig. S1. Spatial distribution of Henan Province’s rural and urban areas (source: Authors’ own elaboration)
Sarfo, I., Qiao, J., Yeboah, E., Okrah, A. , El Rhadiouini, Ch., Osibo, B.K., Boah, A., Ben Amara, D. (2024). Causal effects and prediction of land
use systems in rural landscapes.. . Acta Sci. Pol., Formatio Circumiectus, 23 (3), 27–56. DOI: http://dx.doi.org/10.15576/ASP.FC/190971
55
www.acta.urk.edu.pl
Fig. S2. Spatial trends in Henan Province’s rural and urban areas over the past 40 years (source: Authors’ own elaboration)
Fig. S3. Correlation analysis between the understudied land cover types and surface temperature (source: Authors’ own
elaboration)
Sarfo, I., Qiao, J., Yeboah, E., Okrah, A. , El Rhadiouini, Ch., Osibo, B.K., Boah, A., Ben Amara, D. (2024). Causal effects and prediction of land
use systems in rural landscapes.. . Acta Sci. Pol., Formatio Circumiectus, 23 (3), 27–56. DOI: http://dx.doi.org/10.15576/ASP.FC/190971
56 www.acta.urk.edu.pl
the percentage of farmlands/shrubs declines, surface
temperature amplies. Agricultural practices could
be designed to optimize cooling eects by promoting
agroforestry, i.e. green alternative livelihood projects
that do not adversely impact the forests. By consid-
ering the impacts of land cover on local LST, policy-
makers and stakeholders can make informed decisions
to create more sustainable and climate-friendly en-
vironments. A moderate negative correlation (–0.58,
p < 0.05) was generated for forests and LST. Forests
covered an area of about 24.3%. Thus, expansion in
forest areas resulted in the reduction of LST. Forests
serve as natural LST regulators, as they minimize heat
stress and create more pleasant microclimatic condi-
tions (Asgarian et al., 2014). Additionally, a signi-
cant negative correlation (–0.4, p < 0.05) was obtained
for water bodies (such as lakes, rivers, or reservoirs)
and LST. Areas covered by water amounted to 10.7%.
These correlations demonstrate the signicant inu-
ence of land cover types on prevailing LST patterns
(Ahmed et al., 2019). Understanding these relation-
ships is crucial for eective climate resilience plan-
ning and sustainable land use management, consider-
ing the physical and socio-economic characteristics of
Henan Province.
Similarly, a signicant positive correlation (0.9,
p < 0.05) between bare land and LST was obtained.
Barren areas obtained a 9.2% in area coverage. Find-
ings indicate an increment in the percentage of barren
areas induce the amplication of LST. This suggests
that areas with little or no vegetation, such as bare
surfaces, deserts, rocky surfaces or limited greenery,
may contribute to higher local temperatures, as they
absorb and retain more heat/energy (Al et al., 2020).
The piecemeal of evidence presented in Fig. S3, ex-
hibits a moderate positive correlation (0.69, p < 0.05)
between built-up and LST. Here, built environment
obtained 9.7% coverage. The results indicate the
percentage of built environment amplied in tandem
with LST. Findings indicate areas with high density
of built-up such as Zhengzhou, Shangqiu, Anyang,
Hebi, Nanyang, Xinyang, Puyang, Jiangguanchi and
Luohe experience higher temperatures due to their
composition or land use structure, compared to areas
with low density hotspots. As cities or urban areas
expand, they replace natural surfaces with impervi-
ous surfaces (traditional or impermeable concretes,
pavers and asphalts), which absorb and store heat,
resulting in higher LST compared to the surrounding
rural areas.