m_echeverria@espoch.edu.ec
Transitional dynamics of paramo grassland ecosystem and the primary
activities through neural networks in an Andean lake complex, Sangay
National Park
Dinámica de transición del ecosistema herbazal de páramo frente actividades primarias
mediante el uso de redes neuronales en un complejo Lacustre Andino, Parque Nacional Sangay
Magdy Echeverría Guadalupe
PhD, Environmental Sciences, Universidad San
Marcos Perú, Doctor of Chemistry Escuela
Superior Politécnica de Chimborazo.
m_echeverria@espoch.edu.ec
https://orcid.org/0000-0003-4487-9208
Franklin Enrique Cargua Catagña
Forestry Engineer, Escuela Superior Politécnica de
Chimborazo, Riobamba Ecuador.
Franklin.carguac@espoch.edu.ec
https://orcid.org/0000-0003-2691-5578
Carlos Rolando Rosero Erazo
Master in Biodiversity and Climate Change,
Escuela Superior Politécnica de Chimborazo,
Riobamba, Ecuador, Universidad Santiago de
Compostela, Compostela, Spain.
carlos.roseroe@espoch.edu.ec
https://orcid.org/0000-0001-6005-8915
Abstract
The present study shows the change in three decades, where gains and losses of cover of the páramo
ecosystem were analyzed in the Ozogoche lake complex; whose surface area is 9397.97 ha, an
altitudinal range from 3792 to 4096 m.a.s.l., average temperature from 4 to 12°C and precipitation 1000-
1200 mm. A Land Use Change Modeler (LCM) was used for land cover change; six categories were
established, forest (B), páramo (Pr), crops (C), pasture (Ps), water bodies (A) and forest plantations
(Pf); during the period 1991-2001, Pr showed gains of 503 ha and losses of 305 ha, a net change of
808.8 ha; the forest category (B), suffered a loss of 41 ha and a gain of 19 ha with a net change of 60
ha, this value shows a moderate and slow change trend; category changes for the decade from 2001 to
2011, show a change for (B) with loss of 30 ha and gain of 13.92 ha, a net change of 43.92 ha; the Pr
category shows losses of 2201.2 ha and gain of 1334.8 ha, a total change of 3536 ha, the change between
categories in this time is systematic, but it is much more marked in terms of loss of fragile ecosystems;
the trend of páramo 0.94 shows the probability of displacement to other categories, mainly to pasture,
crops and forest plantations.
http://centrosuragraria.com/index.php/revista, Published by: Edwards Deming Institute,
Quito - Ecuador, July - September vol. 1. Num. 14. 2022, This work is licensed under a
Creative Commons License, Attribution-NonCommercial-ShareAlike 4.0 International.
https://creativecommons.org/licenses/by-nc-sa/4.0/deed.es
Received January 01, 2022
Approved: June 28, 2022
50
Key words: Conservation, land management, páramo, resilience, land use.
Resumen
El presente estudio muestra el cambio en tres décadas, donde se analizaron ganancias y pérdidas de
cobertura del ecosistema páramo en el complejo lacustre Ozogoche; cuya superficie es 9397.97 ha, un
rango altitudinal de 3792 a 4096 m.s.n.m, temperatura promedio de 4 a 12°C y precipitación 1000-1200
mm. Para el cambio de cobertura vegetal se empleó un Modelador de cambio de Uso del Suelo (LCM,
Land Change Modeler); Se establecieron seis categorías, bosque (B), páramo (Pr), cultivos (C), pastos
(Ps), cuerpos de agua (A) y plantaciones forestales (Pf); durante el período 1991-2001, el Pr mostró
ganancias 503 ha y pérdidas de 305 ha, un cambio neto de 808.8 ha; la categoría bosque (B), sufrió una
pérdida de 41 ha y una ganancia de 19 ha con un cambio neto de 60 ha, este valor muestra una tendencia
de cambio moderada y lenta; los cambios de categoría para la década del 2001 hasta el 2011, muestran
un cambio para (B) con pérdida de 30 ha y ganancia de 13.92 ha, un cambio neto de 43.92 ha; la
categoría de Pr muestra pérdidas 2201.2 ha y ganancia de 1334.8 ha, un cambio total de 3536 ha, el
cambio entre categorías en este tiempo es sistemático, pero es mucho más marcado en función a pérdida
de ecosistemas frágiles; la tendencia de páramo 0,94 muestra la probabilidad de desplazamiento a otras
categorías, principalmente a pasto, cultivos y plantaciones forestales.
Palabras clave: Conservación, ordenamiento territorial, páramo, resiliencia, usos de suelo.
Introduction
The páramos are semi-humid and cold ecosystems located on the Andes Mountains, above
the upper limit of the forest, characterized by having a vegetation of low vegetation and a
low humidity. (Mena, Medina, & Hofstede, 2001)In biological terms, the páramos
constitute an important part of the extraordinary ecological diversity of a relatively small
country like Ecuador, but with an environmental and biological variety greater than that of
countries with much larger areas. (Mittermeier, Myers, Thomsen, Da Fonseca, & Olivieri,
1998). The diversity of these ecosystems is attributed to the adaptation of species in adverse
ecosystems that have interacted over time. (Azócar, 1981).
In Ecuador, the páramo covers about 1 250 000 ha, approximately 6% of the national
territory and contains almost 30% of Ecuador's vascular plant species. (Medina & Mena,
2001)This demonstrates the presence of this ecosystem and the importance of diversity in
these ecosystems. (Magurran, 1988)This kind of information will serve as a starting point
for the monitoring of climate change, and thus be able to define how the flora behaves in
the face of this type of variation. (Rodríguez, Armando, Guido, Guillermo, & Enrique,
2016)..
Echeverría, Cargua, Rosero 2022
July - September vol. 1. Num. 14 2022
The degradation of the ecosystems that constitute the páramo zones, determine the need to
broaden the perception of their current and future conditions, in order to propose the
precepts for the definition of a management plan for environmental planning and zoning of
the territory in accordance with the potential of the ecosystems found in these zones.
(Alvizu, 2004).
The present study aims to determine the current state of the páramo ecosystems using a
Land Use Change Modeler (LCM) that is incorporated in the Geographic Information
System (GIS), IDRISI developed by Clark Labs, which is based on a network-based GIS.
(Eastman, 2012)The LCM is based on artificial neural networks to obtain information in a
systematic and continuous way from large extensions of land impossible to cover with any
other kind of methodology. (Paegelow & Olmedo, Spatio-temporal simulation models and
remote sensing, the segmentation method for chronological land use mapping, 2010)..
The community of Ozogoche is of great importance because it is formed mostly by a paramo
ecosystem, which is under anthropic threat, that is why it is important to carry out this type
of study to help determine the floristic composition and serve as a basis for projects that
help protect the resources available and in turn become a source of economic income for
the local inhabitants.
The stratification was carried out based on the land use classes present in the study area,
considering the size of the micro-watershed where three strata were differentiated (MAE
Continental Ecological Classification 2012). The study area that encompasses the Ozogoche
River micro-watershed has an area of 9397.97 ha, these results are of utmost importance for
the care and preservation of ecosystems vulnerable to climate change.
The information generated in this type of studies is of vital importance in the management
of these ecosystems, making it possible to locate the areas with the greatest degree of
affectation over time and to identify the possible causes that generate it. (Wilson & Weng,
2011)These aspects will eventually serve as input in the development of management plans
and conservation policies aimed at the protection and sustainable management of these
sensitive areas. (Izco, 2007).
Materials and methods
The research was carried out in an area corresponding to the upper micro-watershed of the
Ozogoche River, where two communities, Ozogoche Alto and Bajo, are located, with an
area of 9397.97 ha and an altitudinal range between 3792 - 4096 meters above sea level.
The data obtained from the yearbooks of the meteorological station (M396 INHAMI 2012)
show an average temperature of 8ºC, with ranges of (4 - 12) °C; a precipitation of 1077 with
ranges of (1000 - 1200) mm. The soils are sandy loam of great depth, (50-100) cm; with
52
high organic matter content. In the upper micro-basin there is a lake complex made up of
thirty main lagoons. At the local, national and regional levels, this area is considered of the
utmost importance because of the environmental services it provides to the surrounding
communities, including water supply for human consumption, irrigation and hydroelectric
plants.
Figure 1. Study area Upper Ozogoche micro-watershed
Our analysis was focused, under a double stratification sampling design, in the first part the
strata were selected considering the land uses of the study area, classifying in quadrants of
(1*1) km; the second part consisted in estimating the weight in proportion of the strata (%),
where plots were implemented for each stratum, for which equations were used, modified
from forest inventories. (Fonseca, Alice, & & Rey, 2009)..
𝑛 =
𝑡
!
∗ 𝐶𝑉
!
𝐸
"
!
Echeverría, Cargua, Rosero 2022
July - September vol. 1. Num. 14 2022
Base cartographic inputs were used, generated by the Instituto Geográfico Militar (IGM) at
a scale of 1:50,000 (Beltrán et al. 2009). In addition, Rapid Eye and Landsat 8 images
provided by the Ministry of Environment of Ecuador (MAE) were used, with a resolution
of 5 meters and 30 meters respectively.
The supervised classification was performed with the Maximum Likelihood method, which
uses numerical manipulation of the images, which can be interpreted and classify the digital
numbers that represent each pixel and convert them to a language that can manipulate and
work in different enhancements, with this method to achieve a different mapping. (Aguayo,
Pauchard, Azócar, & Parra, 2009)..
This procedure allowed the creation of a database for the years (1991-2001-2011), with
numerical and geographic data that were visually evaluated through coloration scales of
vegetation masses. The changes in vegetation cover and the gain or loss of páramo area as
a consequence of a natural or anthropogenic phenomenon were evaluated. (Ruiz, Savé, &
Herrera, 2013)..
For the land cover change study, a Land Use Change Modeler (LCM, Land Change
Modeler) that is embedded in the Geographic Information System (GIS), called IDRISI
developed by Clark Labs, was used. The Land Use and Land Cover Change Modeling
(LUCC) (Mas, Kolb, Paegelow, Camacho Olmedo, & Houet, 2014) using past and current
land cover information, provides information for environmental and land use planning.
(Wilson & Weng, 2011).
In the following flowchart we can observe the inputs that must be prepared for the input to
the model, various geographic information systems such as ERDAS, ARCGIS, IDRISI, can
be used for the correct input to the model. Other "raster" data useful in modeling are
obtained from these data. All the layers are necessary for the LCM in IDRISI with which
the respective outputs of the subprocesses and the products of the modeler are obtained.
54
Figure 2. Soil Change Modeler Flowchart
Source: Authors,2022
Vegetation cover changes in the area were included to manage different scenarios. This
allowed predicting impacts derived from anthropogenic intervention. The LCM is based on
neural networks, the maps of transitions generated by this modeler have presented optimal
levels of accuracy. (Oñate & Bosque, 2010). The comparison between the maps of potential
change generation with observed deforestation (LUCC Land Use and Cover Change), using
the method of weights in categories of 0-1, which allows to identify the transition potential
and the data entropy. (Pérez & F, 2012).
Echeverría, Cargua, Rosero 2022
July - September vol. 1. Num. 14 2022
In order to use this methodology, it was necessary to take into account 4 important factors.
(Sohl & Claggett, 2013).:
• Document the source code of the model.
• Have reference scenarios to frame uncertainties.
• Improve methods for the propagation of key processes in land use and change.
• Adopt scientifically rigorous measures for uncertainty quantification and model
validation.
The LUCC model used input values as if they were only one (global input), these values are
represented as weights of the input values, these are not restricted but change as the
influence they have on the input values was evaluated, i.e. the input variables (output) of
roads, slope, settlements and uses, are evaluated and distributed according to their influence
within a system, to obtain the Skill measure, and interact with the hidden layers to have
accuracy rates according to the certainty values in order to establish the ideal characteristics
of the change and the prediction model generated by it. (Pineda Jaimes, Bosque Sendra,
Gómez Delgado, & Plata Rocha, 2009)..
Input function of LUCC neural networks.
𝑖𝑚𝑝𝑢𝑡
#$
=
(
𝑖𝑛
#% $
∗ 𝑤
#% $
)
∗
(
𝑖𝑛
#!$
∗ 𝑤
#!$
)
∗ … .
(
𝑖𝑛
#&$
∗ 𝑤
#&$
)
There are different types of vegetation cover classification, but no thresholds of data
accuracy have been determined. The present study determines the influence of variables on
change trends and the measures that should be considered when evaluating a cover.
Result
For a long time methodologies have been created to make the modeler work correctly, the
primary information "INPUT", must be evaluated so that it complies with the logical
characteristics, to establish interactions with the neural networks, which show variations of
temporality according to reality, formats, spatial consistency in its extension, reference
systems, pixel resolution, limit values to zero (0), area and categories must be evaluated
before implementing these models. (Liu & Deng, 2010). The (Figure 03) shows the
Ozogoche micro-watershed (9397.97 ha), which has the nearby communities of Ozogoche
Alto and Bajo, a 12 km (91) third order road. In addition to 10% of it are water bodies (about
27 perennials), among the main ones are the Cubillín and Magtallán lagoons. It has an
altitudinal range between 3792-4096 meters.
56
Figure 3. Land use 1991/2001/201
Source: Authors,2022
The maximum likelihood supervised classification is established with an algorithm that
categorizes by segmenting areas defined by points in space (north and east coordinates),
called training areas, (Paegelow & Olmedo, Spatio-temporal simulation models and remote
sensing, the segmentation method for chronological land use mapping, 2010).The
classification established six categories, forest, moorland, crops, pasture, water body,
plantations; being the moorland ecosystem of greater surface occupying 78.69% in relation
to the total; followed by water bodies (water/lake), with an average value in the three time
periods with a value of 10.36%; pastures with an average of 6.75 and the other uses that
together add up to 4.21% of the total surface.
Table 1. Land Use Categories 1991-2001-2011
Category
1991 has
2001 ha
2011 ha
1
65.84
43.23
26.66
7544.80
7742.7
9
6876.3
4
42.51
177.54
630.44
767.27
390.93
743.85
5
966.52
961.20
989.48
0.090
71.16
127.36
Echeverría, Cargua, Rosero 2022
July - September vol. 1. Num. 14 2022
Source: Authors,2022
The change of category in high Andean zones has become alarming in recent years. Lack
of training, inadequate management of soil resources, lack of land management policies and
inadequate training have become a trigger for the increase of the agricultural frontier.
Access to mechanization is one of the most important factors of change. (Paruelo,
Guerschman, & Verón, 2005).In our study we were able to determine the changes detected
by the maximum likelihood algorithm during three decades (1991-2001-2011).
Table 2. Changes in land use categories 1991-2001-2011
Category
Legend
1991
has
%
2001 ha
%
2011 ha
%
1
B
65.84
0.70
43.23
0.46
26.66
0.28
Pr
7544.8
80.3
7
7742.79
82.49
6876.3
4
73.2
C
42.51
0.45
177.54
1.89
630.44
6.71
Ps
767.27
8.17
390.93
4.16
743.85
7.92
5
A
966.52
10.3
0
961.20
10.24
989.48
10.5
3
Pf
0.09
0.00
1
71.16
0.76
127.36
1.36
Source: Authors,2022
During the first decade, it is clear to observe the change that arises in the grass stratum,
which our classification also shows fragmented wetland areas due to its coloration in the
combination of bands at the time of classification, with a value corresponding to 4% of the
area corresponding to its category, while the transition of the other categories is constant in
its proportionality, excluding exotic forest plantations that had an exponential increase at a
rate of 1:8 times its initial area.
For the decade of 2011, different behaviors were observed in the transition of the categories,
the most relevant change was observed in the moorland category, with a value of 9.79%,
which shows a marked decrease in terms of area; in contrast, the cultivation category
increased by 4.82%, showing a strong anthropic intervention.
58
Profit and loss
Several LUCC models have been developed, so it is difficult to compare which one gives
an accurate representation; there are a number of land use modeling tools and techniques,
the most widely used models is the integrated modeling technique in IDRISI. The Land Use
Change Modeler (LCM) and Markov chains. (Mishra, Rai, & & & Mohan, 2014).. But LCM
is widely used modeling tool. Land Change Modeler that performs the comparison of trend
and land use changes as a function of change factors, roads, settlements, the LCM module
works on neural networks and evaluates levels of accuracy, but this is highly dependent on
the influencing variables, land use changes were evaluated by gains and losses by different
classes.
During the 1991-2001 period, the páramo showed a gain of 503 hectares and a loss of 305
hectares, with a net change of 808 hectares; a fragile ecosystem such as the Andean brow
forest suffered a loss of 41 ha and a gain of 19 ha with a total net change of 60 ha; the
pasture category suffered a loss of 461 ha and a gain of 85 ha, with a net change of 546 ha;
this loss will be analyzed in the transitions submodel; the páramo category was also
representative in terms of its area since it suffered a gain of 71 ha, with a similar value of
total net change. The total net change in all categories shows a moderate and slow change
trend in relation to the area.
Category changes for the decade from 2001 to 2011, showed a representative change for
forest with a loss of 30 ha and a gain of 13.92 ha, with a net change of 43.92 ha; the paramo
category had a notable loss 2201.2 ha and a gain of 1334.8 ha, with a total net change of
3536 ha; the crop category had a gain of 526.59 and a loss of 73.8 ha, with a total net change
of 600.39 ha; the pasture category suffered a loss of 432.3 ha and a gain of 79.71 with a net
change of 512.01 ha; the forest plantation category suffered a loss of 56.14 ha and a loss of
2.1 ha, with a total net change of 58.24 ha; the change between categories in this time period
is more systematic, but is much more marked in terms of the loss of fragile ecosystems
analyzed in previous chapters.
Table 3. Profit and loss 1991-2001-2011
Category
Legend
1991-2001
2001-2011
Los
ses
Earnings
2001
Losses
Earnings
2011
1
B
0.7
0
-41
0.46
-30.0
13.92
0.28
Pr
80.
37
-
305
503
82.4
9
-2201.2
1334.8
73.20
C
0.4
5
-22
1.89
-73.8
526.59
6.71
Echeverría, Cargua, Rosero 2022
July - September vol. 1. Num. 14 2022
Ps
8.1
7
-
461
4.16
-432.3
79.71
7.92
5
A
10.
30
-17
10.2
4
-59.4
63.79
10.53
Pf
0.0
0
0
0.76
2.1
56.14
1.36
Source: Authors,2022
Annual Percentage Rate of Change
The values of change are reflected with two positive trends (+) that show that the final area
is greater than the initial area, represented by the annual or total variation in a given time,
while the negative sign (-) shows that the final area is less than the initial area, the calculated
value should be considered according to the total area of the study area and the vulnerable
ecosystems that may exist in it; Generally, anthropic activities cause severe changes in the
natural ecosystems; the opening of roads and population centers are the causes of changes
in the categories; the main effects are the loss of ecosystem goods and services in the micro-
watershed.
Table 4. Annual rate of change 1991-2011
Category
Legend
Changes
(ha)
1991/2001
Changes
(ha)
2001/2011
Annual rate of
change (ha/year)
1
B
-22.61
-16.57
-1.96
Pr
197.99
-866.45
-33.42
C
135.03
452.90
29.40
Ps
-376.34
352.92
-1.17
5
A
-5.32
28.28
1.15
Pf
71.07
56.20
6.36
Source: Authors,2022
The trend showed a variation in two decades, manifesting the high values in the decade
2011-2017, in the páramo ecosystem with a loss of 866.45 ha, with an average annual value
of 33.42 ha/year; while anthropogenic changes show a high growth, showing a value of
452.90 ha for the crop category, with an average annual gain of 29.40 ha; the smallest
changes but corresponding to sensitive ecosystems are given by the loss of forest at a ratio
of 22.61 ha in the decade 1991-2001, with an annual rate of 1.96 ha/year.
Transition and persistence, have a structure so that the multilayer perceptron neural network
(MLP), which is one of the most widely used Artificial Neural Networks (ANN). The
60
training of multilayer perceptron neural networks is based on Backpropagation (BP) which
is a supervised training algorithm. It is a common method of training Artificial Neural
Networks. Starting from a desired output, the network interacts with various inputs. An
(MLP) is an automated process, which maps input data sets to an output set. An MLP
consists of multiple layers of nodes in a directed graph, with each layer being fully
connected to the next. The perceptron is an algorithm for supervised classification of an
input into one of several possible outputs. MLP performs classification of remotely sensed
images through multilayer perceptron neural network classifier using back propagation.
(Matich Damian, 2001)..
Table 5. Analysis of variables according to the transition and persistence submodel
Input
Files
Forcing a Single Independent
Variable to be Constant
Model Skill Breakdown by
Transition & Persistence
Variable
Accu
racy
(%)
Skill
measure
Class
Skill
measu
re
DEM
75.80
0.71
T Paramo to Pastures
0.200
DEM
59.19
0.51
T Bosque to Páramo
0.887
Changes
(91/01)
48.91
0.39
T Crops to Páramo
0.998
Changes
(01/11)
27.42
0.13
T Paramo to Crops* T
0.889
Villages
2001
81.35
0.78
T Pastures to Páramo
0.200
Villages
2011
46.13
0.35
T Pastures to Crops
0.703
Class 91
81.18
0.77
P Páramo
0.887
Class 01
72.90
0.67
P Forest
0.857
Rivers
81.21
0.77
P Crops
0.200
Rivers
79.68
0.76
P Páramo
0.502
2001
81.09
0.77
P Pastures
0.499
Roads
2011
80.32
0.76
P Pastures
0.741
Analysis 1991/2001 Bold, 2001/2011 italic; *significance
The analysis of the transition (T) and persistence (P) submodel is based on the study of the
crossed matrix of the weights of the neurons of the hidden layer and the neurons of the
output layer, six variables that LUCC considers influential in the process of change were
Echeverría, Cargua, Rosero 2022
July - September vol. 1. Num. 14 2022
used, for the 91/01 change; a precision of 81.18% was determined, with an ability measure
of 77.42; the order of influence to the change was established for the populated variable,
this is due to the fact that the settlements at the beginning of the 90's, increased due to the
opening of the facility of rural land acquisition in the projects of IERAC (Ecuadorian
Institute of Agrarian Reform and Colonization), which resulted in the displacement towards
the moorlands, contingency projects like FACE (Forest Absorbing Carbon Dioxide),
implemented extensive areas with forest plantations as a measure of "reforestation" in areas
that have no forestry aptitude. The changes occurred in the decade 2001/2011; determined
a total precision of 80.32%, with a measure of ability of 0.76; the variable that has
influenced in greater proportion was that of roads, the opening of new roads is a factor in
the change of coverage, the settlements have access to the mechanization of the soil, the
agricultural frontier increases systematically and in all uses; for the analysis of change tests
were performed ignoring areas of transition between 0.1:0.1:1:1 ha and 1:1:1:13 ha; among
which it is evident that there is a relationship to change by ignoring changes of less than 14
hectares, thus reducing from 22 to 6 transitions.
From the previous analysis, we can see the tendency of change to know which area generates
the greatest change in coverage, and it is directly related to the introduction of roads,
communities and influence of the geomorphology by the fluvial network, the changes in
volume of the water bodies are negligible, around the Cubillín and Magtayan lagoon, but
the tendency of all the variables have direct influence towards the zones that have been
analyzed whose described indexes show a high floristic diversity, with emphasis on the
paramo ecosystems.
Figure 4. Trends in change 1991-2011
1991-2001
2001-2011
62
Source: Authors,2022
The IDRISIS module implements Markov chains (MARKOV) and simulates the prediction
of the state of a system at a given time from two preceding states. This means that the
modeling does not take into account explanatory and descriptive variables, but is based
exclusively on the analysis of the internal dynamics of the system, which, in our case,
corresponds to the evolution of land use. It is a discrete procedure in discrete time, where
the value at time t1 depends on the values at times t0 and t1 (second order Markovian chain).
The algorithm compares two chronologically successive land cover maps, and estimates and
configures a transition probability matrix. The prediction is materialized in a series of land
cover maps (one for each category) for a future time, where the digital level of each pixel
expresses the probability of belonging to the category under analysis. (Paegelow, Camacho
Echeverría, Cargua, Rosero 2022
July - September vol. 1. Num. 14 2022
Olmedo, & Menor Toribio, Markov chains, multi-criteria evaluation and multi-objective
evaluation for prospective landscape modeling, 2003)..
Table 6. MARKOV Chain Analysis
Probability of changing to:
CUT's
B
Pr
C
Ps
A
Pf
Forest
0.4077
0.5198
0.0068
0.0178
0.026
0.0219
Paramo
0.0017
0.9492
0.0198
0.0105
0.0012
0.0176
Crops
0.0127
0.3835
0.4746
0.1292
0
0
Pastures
0.0029
0.5664
0.0254
0.4046
0.0004
0.0004
Water Bodies
0.0065
0.0139
0.0007
0
0.9763
0.0026
Forest plantations
0.2
0.2
0.2
0.2
0.2
0.001
Source: Authors,2022
The MARKOV matrix shows the probability that an entire category changes or remains in
its category, analyzing the directionality of change in previous analyses, the trend of
moorland was observed with a 0.94 probability of change in other categories, thus showing
us the transition submodels, which show that the trend of change towards other categories,
mainly pasture and crops; the change generated on the basis of stochastic processes occurs
in greater proportion in the categories of forest, pasture and crops, due to their dynamics in
the analysis and the interaction generated in the changes in land use; Water bodies have a
tendency of permanence according to the area; forest plantations will not suffer alterations
according to their cutting season; this category has a particularity because when they reach
their season they are extracted and the changes they can generate are not foreseeable, since
they can be replaced by agricultural zones and in some cases they can be considered zones
of assisted natural regeneration.
Conclusions
A supervised classification was established with the Maximum Likelihood algorithm,
establishing 6 categories, forest (B), paramo (Pr), crops (C), pastures (Ps), water bodies (A)
and forest plantations (Pf); the largest area corresponds to the Pr category, while the BPf
categories occupy a smaller area. During the period 1991-2001, the páramo showed a gain
of 503 ha and a loss of 305 ha, with a net change of 808 ha.8 ha; a fragile ecosystem such
as the Andean brow forest (B), suffered a loss of 41 ha and a gain of 19 ha with a total net
change of 60 ha; the Ps category suffered a loss of 461 ha and a gain of 85 ha, with a net
change of 546 ha; the Pr category was also representative in terms of its area since it suffered
a gain of 71 ha, with a similar value of total net change. The total net change in all categories
shows a moderate and slow trend of change in relation to area.
64
Category changes for the decade from 2001 to 2011, show a representative change for B
with a loss of 30 ha and a gain of 13.92 ha, with a net change of 43.92 ha; the Pr category
has a notable loss 2201.2 ha and a gain of 1334.8 ha, with a total net change of 3536 ha; the
crop category has a gain of 526.59 and a loss of 73.8 ha, with a total net change of 600.39
ha; the Ps category suffered a loss of 432.3 ha and a gain of 79.71 with a net change of
512.01 ha; the Pf category suffered a loss of 56.14 ha and a loss of 2.1 ha, with a total net
change of 58.24 ha; the change between categories in this time span is more systematic, but
is much more marked in terms of loss of fragile ecosystems.
The analysis of the transition and persistence submodel is based on the study of the crossed
matrix of the weights of the neurons of the hidden layer and the neurons of the output layer,
using six variables that LUCC considers influential in the process of change, for the 91/01
change; a precision of 81.18% was determined, with an ability measure of 77.42; the order
of influence to the change was established for the population variable, this is due to the fact
that the settlements at the beginning of the 90's, increased due to the opening of the facility
of rural land acquisition in the IERAC (Ecuadorian Institute of Agrarian Reform and
Colonization) projects, which resulted in the displacement towards the moorlands,
contingency projects such as FACE (Forest Absorbing Carbon Dioxide), implemented
extensive areas with forest plantations as a measure of "reforestation" in areas that do not
have forest aptitude. The probability that an entire category will change or remain the same
is the result of analyzing the directionality of change; the trend of moorland 0.94 shows the
probability of displacement to other categories; the transition submodels show that the trend
of change towards other categories, mainly to pasture and crops; the change generated on
the basis of stochastic processes occurs in greater proportion in the categories of forest,
pasture and crops, due to their dynamics in the analysis and the interaction generated in the
changes in land use; Water bodies have a tendency of permanence according to the area;
forest plantations will not suffer alterations according to their cutting season; this category
has a particularity because when they reach their season they are extracted and the changes
they can generate are not foreseeable, since they can be replaced by agricultural zones and
in some cases they can be considered zones of assisted natural regeneration.
They express the synthesis of the main result obtained by the research process or study
carried out, they are a consequence of the same and therefore the conclusions are directly
related to the stated objective.
References
Aguayo, M., Pauchard, A., Azócar, G., & Parra (2009). Land use change in south-central Chile
in the late twentieth century: Understanding spatial and temporal landscape dynamics
... . Chilean journal of natural history, 3(82), 361-374.
Echeverría, Cargua, Rosero 2022
July - September vol. 1. Num. 14 2022
Alvizu, P. (2004). Complexity and functional response of páramo vegetation along altitudinal
gradients. Merida-Venezuela.
Azócar, A. (1981). The flora of the páramos. In: Los páramos venezolanos. Venezuela: Karl
Weidman. Edi. Alemena Bookstore.
Eastman, J. R. (2012). IDRISI Selva. (MA: Clark University. ed.) Worcester.
Fonseca, W., Alice, F., & Rey, J. M. (2009). Models for estimating native species biomass in
plantations and secondary forests in the Caribbean zone of Costa Rica. Valdivia
Forests, 36-47.
Galeas, C. J., Navarro, G. F., Cornejo, X., M. H., & Toasa11, G. (2012). Propuesta
Metodológica para la representación cartográfica de los ecosistemas del Ecuador
continental. Ministry of the Environment of Ecuador, 86-88.
Izco, e. a. (2007). Floristic study of the southern pajonal moorlands of Ecuador. Revista
Peruana de Biología, Vol.14, n.2, pp. 237-246.
Josse, C. P. (2000). La Biodiversidad de los Páramos. Quito: Abya Yala.
Liu, J., & Deng, X. (2010). Progress of the research methodologies on the temporal and spatial
process of LUCC. Chinese Science Bulletin, 55(14), 1354-1362.
Magurran, A. E. (1988). Ecological diversity and its measurement. Zaragoza: Princeton:
Princeton university press.
Mas, J.-F., Kolb, M., Paegelow, M., Camacho Olmedo, M. T., & Houet, T. (01 2014). Inductive
pattern-based land use/cover change models: A comparison of four software packages.
Environmental Modelling & Software, 51(0), 94 - 111.
doi:http://dx.doi.org/10.1016/j.envsoft.2013.09.010
Matich Damián, J. (2001). Redes Neuronales (First ed.). Rosario: Universidad Tecnológica
Nacional.
Medina, G., & Mena, P. (2001). The moorlands of Ecuador. In: Mena, P., G. Medina & R.G.M.
Hofstede (eds.). Los Páramos del Ecuador. Páramo Project. Quito: Abya Yala.
Mena, P., Medina, G., & Hofstede, R. (2001). Los Páramos del Ecuador. Quito: Abya Yala.
Mishra, V. N., Rai, P. K., & & Mohan, K. (2014). Prediction of land use changes based on land
change modeler (LCM) using remote sensing: a case study of Muzaffarpur (Bihar),
India. Journal of the Geographical Institute" Jovan Cvijic", 64(1), 111-127.
66
Mittermeier, R. A., Myers, N., Thomsen, J. B., Da Fonseca, G. A., & Olivieri, S. (1998).
Biodiversity hotspots and major tropical wilderness areas: approaches to setting
conservation priorities. Conservation biology, 12(3), 516-520.
Oñate, F., & Bosque, J. (December 15, 2010). Application of GIS and remote sensing
techniques in generation of land use scenarios for hydrological modeling. Journal of
Hydrology, 395(3-4), 256 - 263. doi:http://dx.doi.org/10.1016/j.jhydrol.2010.10.033
Ordoñez. (2009). Consultancy report, Survey of information in the southern area of the
province of Manabí, Ecuador, in support of the regional program of innovative
financial mechanisms for sustainable land management through environmental
livestock. Manabí_Ecuador.
Paegelow, M., & Olmedo, M. C. (2010). Spatio-temporal simulation models and remote
sensing, the segmentation method for chronological land use mapping. Geographic
Series, 16(1), 19-34.
Paegelow, M., Camacho Olmedo, M. T., & Menor Toribio, J. (2003). Markov chains, multi-
criteria evaluation and multi-objective evaluation for prospective landscape modeling.
International Journal of Geographic Information Science and Technology, 1(3), 22-44.
Paruelo, J. M., Guerschman, J. P., & Verón, S. R. (2005). Agricultural expansion and land use
change. Science Today, 15(87), 14-23.
Pauli H., M. G. (2003). Fieldwork manual for the GLORIA project. Institute for ecology and
biological conservation. Vienna: University of Vienna.
Pauli, H., Gottfried, M., Hhenwallner, D., Reiter, K., & Grabherr, G. (2003). Fieldwork manual
for the GLORIA project. In Initiative for global research and monitoring of alpine
environments as a contribution to the Global Terrestrial Observing System (GTOS)
(pp. 1-67). Vienna: University of Vienna and Pyrenean Institute of Ecology.
Pérez, A., & F, M. (2012). Comparing two approaches to land use/cover change modeling and
their implications for the assessment of biodiversity loss in a deciduous tropical forest.
Environmental Modelling & Software, 29(1), 11 - 23.
doi:http://dx.doi.org/10.1016/j.envsoft.2011.09.011
Perrone, A. (2001). La sostenibilidad del ecoturismo en el Ecuador. Quito: Serie Páramo
(Ecoturismo) 9: 23-37.
Pineda Jaimes, N. B., Bosque Sendra, J., Gómez Delgado, M., & Plata Rocha, W. (2009).
Analysis of land use change in the State of Mexico using geographic information
Echeverría, Cargua, Rosero 2022
July - September vol. 1. Num. 14 2022
systems and multivariate regression techniques: An approach to deforestation
processes. Investigaciones Geográficas(69), 33-52.
Reyes, A. P., Ramos, P. C., López, S. A., & López, I. H. (2016). STUDY OF THE FLORISTIC
DIVERSITY OF THE PÁRAMO OF THE GUANGOPUD COMMUNITY,
CHIMBORAZO PROVINCE, ECUADOR. REVISTA PERSPECTIVA(16), 1-2.
Rodríguez, R. L., Armando, D. D., Guido, S. L., Guillermo, R. M., & Enrique, C. C. (2016).
Forest biodiversity in Andean brow and its relation to the carbon content in the above-
ground biomass. Acta Agronómica, 66(1), 35-41.
Ruiz, V., Savé, R., & Herrera, A. (2013). Multitemporal analysis of land use change, in the
Protected Terrestrial Landscape Miraflor Moropotente Nicaragua, 1993-2011.
Ecosystems, 22(3).
Sierra, M. (1999). Preliminary proposal for a vegetation classification system for continental
Ecuador. Inefan/Gef-Birf and Ecociencia Project.
Smith, S. R. (2007). Ecology: Communities. Eds. Capella, F. 4 ed. Madrid, ES. P 350-356.:
Editorial Pearson Educacion S.A.
Sohl, T. L., & Claggett, P. R. (11/15/2013). Clarity versus complexity: Land-use modeling as
a practical tool for decision-makers. Journal of Environmental Management, 129(0),
235 -243. doi:http://dx.doi.org/10.1016/j.jenvman.2013.07.027
Vargas, O. P. (2002). Effect of grazing on vegetation structure in a humid páramo of Colombia.
ECOTROPICOS_Venezuelan Society of Ecology, Vol: 15 (1):35-50.
Wilson, C. O., & Weng, Q. (2011). Simulating the impacts of future land use and climate
changes on surface water quality in the Des Plaines River watershed, Chicago
Metropolitan Statistical Area, Illinois. Science of The Total Environment, 409(20), 4387
- 4405. doi:http://dx.doi.org/10.1016/j.scitotenv.2011.07.001