Land use in Ecuador: a statistical analysis at different aggregation levels

https://doi.org/10.1016/S0167-8809(98)00151-0Get rights and content

Abstract

Land use in Ecuador was investigated by means of statistical analysis with the purpose of deriving quantitative estimates of the relative areas of land use types on the basis of biogeophysical, socio-economic and infrastructural conditions. The smallest spatial units of investigation were 5 by 5 minute (9.25×9.25 km) cells of a homogenous geographical grid covering the whole country. Through aggregations of these cells, a total of six artificial aggregation levels was obtained with the aim of analysing spatial scale dependence of land use structure. For all aggregation levels independent multiple regression models were constructed for the estimation of areas within cells of the land use/cover types permanent crops, temporary crops, grassland and natural vegetation. The variables used in the models were selected from a total of 23 variables, that were considered proxies of biogeophysical, socio-economic and infrastructural conditions driving Ecuadorian land use. A spatial stratification was applied by dividing the country into three main eco-regions. The results showed that at higher aggregation levels, the independent variables explained more of the variance in areas of land use types. In most cases, biogeophysical, socio-economic as well as infrastructural variables were important for the explanation of land use, although the variables included in the models and their relative importance varied between land use types and eco-regions. Also within one eco-region, the model variables varied with aggregation level, indicating spatial scale effects. It is argued that these types of analyses can support the quantitative multi-scale understanding of land use, needed for the modelling of realistic future land use change scenarios that take into account local and regional conditions of actual land use.

Introduction

World-wide changes in land use and resulting land cover have caused important effects on natural resources through deterioration of soil and water quality, the loss of biodiversity, and by changing global climate systems (Turner II et al., 1994; Ojima et al., 1994). This has stimulated research aiming at a better understanding of the factors driving land use and cover change, and the effects of these changes on the environment. It is recognised that biogeophysical as well as human drivers have to be taken into account (Turner II et al., 1995; Bilsborrow and Okoth Ogendo, 1992; Riebsame et al., 1994). In order to support explorative modelling of future land use and cover changes and their effects, quantitative information is needed about the way interacting driving forces relate to (changes in) land use/cover. Because of the complex nature of these relations, empirical statistical analysis of land use/cover and its drivers has been proposed (Turner II et al., 1995; Bawa and Dayanandan, 1997; Walsh et al., 1997; Veldkamp and Fresco, 1997a). Empirical models based on these statistical relations can complement process-based land use modelling. Process-based models aim at more explanatory power but have difficulties in linking biogeophysical and human drivers. Furthermore scale problems may arise when applying point models to higher spatial scales (Veldkamp and Fresco, 1997b). Empirical explorative land use models based on the analysis of actual land use present possible scenarios that aim at a relatively limited future time scale (say 20 years), but are especially useful in situations where the actual production is still considerably below the biophysical potential, indicating strong limitations from socio-economic conditions.

It has been recognised that the type and effect of drivers of land use/cover may vary with spatial scale, because of the occurrence of patterns in land use/cover that disappear or emerge going from one scale to another (Walsh et al., 1997; Veldkamp and Fresco, 1997a). In order to investigate these scale dependent patterns in land use, an analysis at different spatial scales is necessary. Veldkamp and Fresco (1997a)have analysed land use in Costa Rica at artificial spatial aggregation levels, created by aggregating uniformly sized cells of a geographical grid. They concluded that the contribution of biogeophysical and socio-economic factors to the explanation of land use/cover in Costa Rica shows a scale dependence.

In the present study a statistical analyses is presented of current land use in Ecuador with the objective of finding quantitative estimates of (proxies of) land use/cover drivers. The methodology, a statistical analyses at different sub-national artificial spatial aggregation levels in order to investigate spatial scale effects, builds on the methodology proposed by Veldkamp and Fresco (1997a). However, some major adaptations were made in order to improve the methodology. A spatial stratification of land use through the definition of three main eco-regions was applied. Furthermore, different sets of drivers were considered at the various aggregation levels.

Ecuador was chosen for this research because it is a country with a high agro-ecological diversity. It has a dynamic and expanding agricultural land use, which has major effects on the natural resources of the country and the sustainability of land use (Southgate and Whitaker, 1994; Bebbington, 1993). In many ways Ecuador can be considered indicative for the Andean countries in general.

Strictly, land use and land cover should be separated, land cover being the biophysical state of the earth's surface and immediate subsurface, and land use involving both the manner in which the biophysical attributes of the land are manipulated and the intent underlying that manipulation – the purpose for which the land is used (Turner II et al., 1995). However, land use and cover can often not be clearly separated. In this paper only the term land use will be used.

Section snippets

Land use in Ecuador and its potential drivers

Ecuador is a country characterised by great agro-ecological diversity. Cañadas (1983) has identified 25 Holdridge lifezones within Ecuador. The country can be divided into three broad eco-regions. The Andean eco-region consists of the north-south orientated Andean mountain range, with peaks to around 6000 m above sea level. West of this eco-region, the tropical lowlands bordering the pacific ocean comprise the coastal eco-region. East of the Andes, the Amazonian eco-region is located, still

Spatial resolution and aggregation levels

The lowest aggregation level used in this study was a homogeneous geographical grid with a grid cell size of 5 by 5 minutes (approximately 9.25 by 9.25 km), covering the Ecuadorian territory according to the protocol of Rio de Janeiro 1942, excluding the Galapagos islands. A cell was considered an Ecuadorian land cell when at least 50% of its surface was Ecuadorian territory excluding sea. The total number of Ecuadorian land cells was 2982. Biogeophysical and socio-economic data were collected

Results

Fig. 5 presents graphically the adjusted coefficients of determination (r2) of the multiple regression models for the four land use types and six aggregation levels, for the eco-regions Coast, Andes and Amazon. All models were significant at the 0.001 level. The general pattern in all three eco-regions is that of an increasing r2 with higher aggregation levels, as expected given the reduction of extreme values.

In eco-region Coast, the model explains 24% to 46% of the variance of the land use

Discussion

The statistical analysis of Ecuadorian land use has resulted in significant multiple regression models for all combinations of land use type, eco-region and aggregation level. In most cases the models give rather satisfying fits, taking into account the highly complex nature of agro-ecosystems and the limited number of variables used. Not only the r2 and the variables selected in the multiple regression models vary with the aggregation level, but also the standardised betas of the variables. In

Acknowledgements

Part of this study was executed by the first author at the research station of the International Potato Centre (CIP) in Quito. The authors thank CIP (in particular C. Crissman) for the facilities and co-operation offered. Thanks are expressed to J. Maldonado (INEC, Quito) and C. Larrea (Frente Social, Quito) for sharing their valuable information.

References (36)

  • R Bromley

    The colonization of humid tropical areas in Ecuador

    Singapore Journal of Tropical Geography

    (1981)
  • Cañadas, L., 1983. El mapa bioclimático y ecológico del Ecuador. Banco Central del Ecuador, Quito, Ecuador, 210...
  • Cuvi, M., Urriola, R., 1988. Oleaginosas, cereales y agroindustria en la Costa Ecuatoriana. In: Gondard, P., León,...
  • FAO, 1997. FAOSTAT, statistical database....
  • Frère, M., Rea, J., Rijks, J.Q., 1975. Estudio agroclimatológico de la zona Andina. FAO, 375...
  • González, A., Maldonado, F., Mejı́a, L., 1986. Memoria explicativa del mapa general de suelos del Ecuador. Sociedad...
  • INEC (Instituto Nacional de Estadı́stica y Censos), 1974a. Censo Agropecuario...
  • INEC (Instituto Nacional de Estadı́stica y Censos), 1974b. Censo de población...
  • Cited by (85)

    • Spatial differentiation characteristics of internal ecological land structure in rural settlements and its response to natural and socio-economic conditions in the Central Plains, China

      2020, Science of the Total Environment
      Citation Excerpt :

      The spatial differentiation of the ELRS structure is the result of multiple factors of the natural environment and socio-economic conditions. The spatial pattern characteristics of RS are related to the scale of the study (de Koning et al., 1998). On a large scale, the structure of RS is mainly influenced by natural conditions (Tian et al., 2012).

    • Assessment of sediment yield using RS and GIS at two sub-basins of Dez Watershed, Iran

      2016, International Soil and Water Conservation Research
      Citation Excerpt :

      In fact, cultivated areas can act as both a source and a pathway for transporting nutrients (Ouyang & Bartholic, 1997). Many studies have shown that human activity is the major cause of recent changes in the land use (Bennett, 1939; De Koning, Veldkamp, & Fresco, 1998; Jain & Das, 2010; Le Bissonnais, Montier, Jamagne, Daroussin, & King, 2002; Martınez-Casasnovas, 2003; Pandey, Chowdary, & Mal, 2007). Also in non-residential areas biophysical conditions of the land, such as lithology, soil characteristics, hydrology, topography, and ground cover largely determine the spatial pattern of the land use and its temporal changes (Estrany, Garcia, & Walling, 2010; Veldkamp & Fresco, 1996).

    • Landscape changes in Norwegian mountains: Increased and decreased accessibility, and their driving forces

      2016, Land Use Policy
      Citation Excerpt :

      On the one hand, ‘upscaling’ can occur as a description of general trends and an explanation of how trends can be related to the study area in question (Brandt et al., 1999; Kristensen et al., 2009), on the other hand it can be performed as an identification of driving forces on different actor levels (Hersperger and Bürgi, 2009; Schneeberger et al., 2007). Among the structure-based approaches, only the two studies not explicitly dealing with landscape change (de Koning et al., 1998; Verburg and Chen, 2000) regard different scales (regional and national). In the upscaling process, values for the structures are aggregated in larger grid cells.

    View all citing articles on Scopus
    1

    Current address: SDR, FAO, Rome, Italy.

    View full text