Assessing soil erosion risk at national scale in developing countries: The technical challenges, a proposed methodology, and a case history

https://doi.org/10.1016/j.scitotenv.2019.135474Get rights and content

Highlights

  • Large evidence on the urgency to asses soil erosion in developing countries (DC)

  • Most DC have insufficient observations to quantify erosion at country scale.

  • RUSLE-GGS successfully tackles with the uncertainty in quantifying erosion in DC.

  • RUSLE-GGS can potentially standardize erosion evolution assessment in DC.

  • Attaining Goal 15 from UN 2030 Agenda demands standardizing such assessment.

Abstract

Through an extensive bibliographic review, this contribution underlines the urgency and challenges to quantify soil erosion rates (ERs) in developing countries. It subsequently elaborates on the combined application of GIS-based RUSLE, generalized likelihood uncertainty estimation (GLUE) principles and sediment delivery ratio functions (SDR) to quantify ERs at country scale for these countries, as they commonly have limited measurements to that purpose. The methodology, termed RUSLE-GGS (RUSLE-GIS-GLUE-SDR) herein, comprises the following sequence: (1) construction of ER samples using RUSLE-GIS based on freely available local/global geoenvironmental observations and field relations, (2) construction of area-specific sediment yield samples utilizing SDR transfer functions, and (3) assessment of the most behavioral samples by means of bias analysis and cross validation. Its application to Peru allows obtaining 5-km resolution ER and potential erosion maps for the years 1990, 2000, and 2010. RUSLE-GGS is highly replicable and could potentially be used as an initial standard and systematic method to estimate ERs in developing countries through the active participation of local scientists. Thus, it potentially can contribute to improve the capacity building in such countries and set an initial frame to compare the evolution of soil erosion in their territories towards attaining Goal 15 of the UN 2030 Agenda for Sustainable Development.

Introduction

Soil erosion is a natural phenomenon mainly induced by site meteorological, topographical, geological, land cover conditions (e.g., soil disturbances related to deforestation, mining, agriculture, construction, urbanization, population growth, etc.), and underlying geomorphological processes such as hill slope erosion, mass movement, and channel erosion. Soil erosion will very likely be intensified by large scale anthropogenic controls such as global warming (Nearing et al., (2004), Lal et al., (2011)). As a consequence, soil erosion represents a global societal concern because: (1) it often degrades soil and water resources and triggers economic losses in several countries all around the World (Ribaudo, (2009), Ayele et al., (2015)), and (2) plays an important role in the global carbon cycle (Yang et al., (2003), Van Oost et al., (2007), Ito, (2007)).

Some initiatives have been launched in recent years to improve World's social, economic, environmental conditions. The UN 2030 Agenda for Sustainable Development (United Nations, 2015) has set 17 goals for the year 2030 to that end. In specific, Goal 15 — life on land (“ by 2030 governments need to protect, restore and promote sustainable use of terrestrial ecosystems, sustainably manage forests, combat desertification, and halt and reverse land degradation and halt biodiversity loss”) is closely related to soil erosion. Lu et al. (2015) identified the following 5 priorities to accomplish these goals: (1) devising metrics so that the goals can be measurable, comparable and achievable: (2) establishing monitoring mechanisms to decide which values need to be tracked, and set up systems to acquire the data; (3) evaluating progress; (4) enhancing infrastructure, i.e. expanding Earth observation, ground-based monitoring and information processing capabilities; and (5) standardizing and verifying data, e.g. presenting the data as open access information. Soil erosion rates (ERs) have been profusely estimated in developed countries through field, experimental and numerical modeling approaches, and at a wide range of spatio-temporal scales Kirkby and Cox, 1995; (Dedkov and Gusarov, (2006), Bellin et al., (2011), Morgan and Nearing, (2011), Cerdà et al., (2013)). Conversely, a very limited number of such studies have been conducted in developing countries Onyando et al., (2005), Shamshad et al., (2008), Labrière et al., (2015), even though there is a large suite of scientific evidence that (1) ERs are steadily increasing in their territories and likely reaching dramatics levels (Pimentel et al., (1995), Pham et al., (2001), Ananda and Herath, (2003), , (2006), Labrière et al., (2015), Borrelli et al., (2017)), and (2) soil erosion is currently one of the major environmental and geomorphic hazards exhibiting higher impacts in these countries (Alcantara-Ayala, (2002), Mondal et al., (2017)). Thus, to attain Goal 15, the quantification of ERs in developing countries probably needs to be addressed with particular urgency.

Developing countries are mostly located in humid tropical regions (Sachs, 2001) and commonly face technical, financial, regulatory, and capacity-building challenges to improve the availability of: (1) spatio-temporal measurements and field relations to estimate ERs (Millward and Mersey, 1999; (Labrière et al., 2015); and (2) soil erosion observations (e.g., ERs, frequency and extent of erosion, sediment yield) to calibrate or validate erosion models. In particular, sediment yield data is usually only available for large rivers, and, in many instances is insufficient in length, consistency, and continuity; and moreover, it is rarely publicly available (Labrière et al., 2015).

Several models to estimate ERs exist. They have been characterized as follows: (1) empirical or statistical models (e.g., SEDD, PSIAC) which are mainly based on the Revised Universal Soil Loss Equation, RUSLE; (2) conceptual models (e.g., SEDNET, SWAT), which commonly describe catchment processes without providing specific details of their interactions; and (3) physically based models (e.g., WEPP, PESERA, EUROSEM) which are based on the equations of conservation of mass and momentum for flow and the equations of conservation of mass for sediment (de Vente et al., (2013), Hajigholizadeh et al., (2018)). The distinction between models is however diffuse for they couple modules from each of these categories (Ranzi et al., (2012), de Vente et al., (2013)). Likewise, past research has highlighted the strong dependency of empirical, conceptual and physically based models on the availability of high resolution spatio-temporal input and calibration data, and the critical need of long and continuous simulations to reliably predict soil erosion (Merritt et al., (2003), Nearing, (2004), Ranzi et al., (2012), de Vente et al., (2013), Borrelli et al., (2017)). Therefore, the selection of the most suitable model is subjected to the intended use and available data.

RUSLE was basically developed to estimate long-term average soil loss (i.e. gross erosion) and has been applied not only at small scales, but also at large scales, i.e. national, continental, and global scales (de Vente et al., 2005, de Vente et al., (2008); Jetten and Maneta, (2011), Naipal et al., (2015), Panagos et al., (2015), Martin-Fernandez and Martinez-Nu nez, (2011)). Typically, the main purpose of national scale estimations has been showing historical average erosion risk information to be used by policy-makers and territorial planning authorities, and to identify critical soil erosion prone areas that might need institutional attention and/or require finer spatio-temporal assessment (Van der Knijff et al., (2000), Šúri et al., (2002), Terranova et al., (2009)). Some of these estimates (Šúri et al., (2002), Terranova et al., (2009), Ranzi et al., (2012)) were obtained by adopting Geographical Information System (GIS) techniques to treat data for the application to the RUSLE model.

In order to achieve Goal 15 of the UN 2030 Agenda for Sustainable Development, developing countries must firstly concentrate their efforts in the first and fifth priorities proposed by Lu et al. (2015). We posit that the studies to that end should be gradually conducted by local scientists to improve their capacity-building in these countries as well. Thus, it is reasonable to state that on the basis of Priority 1, there is a need to: (1) estimate soil erosion rates at both national scale and annual scale by following a standard method, (2) set a standard base line year for future comparison. The application of the RUSLE-GIS model based on publicly available local and satellite observations appears to be the most accessible mean to meet this necessity. That nevertheless demands generalizing detail in data and coping with the structural paucity of soil erosion measurements, which makes model validation challenging and imposes higher uncertainty into the model outputs.

Several studies have tackled RUSLE uncertainty. For instance, at global scale RUSLE-based ERs were validated using spatial extrapolation of plot experiments data from the NRI database for the USA and erosion estimates for Europe, and subsequently they were compared with global sediment yield observations from the World's major rivers (Pham et al., (2001), Van Oost et al., (2007), Ito, (2007), Naipal et al., (2015)). Borrelli et al. (2017), meanwhile, used Markov Chain Monte Carlo approach. At catchment scale, RUSLE-based ERs have been validated using sediment delivery ratio (SDR) equations, in which SDR was used as a proxy parameter to estimate catchment sediment yield from gross erosion (Catari Yujra and Saurí i Pujol, (2010), Lee et al., (2014), Swarnkar et al., (2017)). Likewise, Swarnkar et al. (2017) coupled Monte Carlo, RUSLE and SDR at catchment scale in India and obtained ER estimates with acceptable level of uncertainty.

In recent years, Generalized Likelihood Uncertainty Estimation (GLUE) principles have been adopted to estimate the uncertainty of erosion models. For example de Vente et al., (2008), Jetten and Maneta, (2011) coupled GLUE and SDR estimates to validate physically based erosion models at regional scale. GLUE considers that in field applications it is very difficult to specify a consistent model of the output errors due to our imperfect knowledge of the system and the associated uncertainty of the input data, and by virtue of that, different parameter sets can produce acceptable results (Freer et al., (1996), Brazier et al., (2000), Brazier et al., (2001), Aronica et al., (2002), Wei et al., (2008), Beven et al., (2008), Quinton et al., (2011)). Far from the prevalent approach that parametrizes the RUSLE fundamental parameters and calibrate the outputs with local observations, an application of GLUE into RUSLE-based models would estimate the likelihood of a given set of models, parameters and variables. That would also agree with a body of evidence that suggests that model predictions that are produced through the random generation of parameter values can perform better than those produced by classical calibration (Brazier et al., (2000), Beven and Brazier, (2011)).

This contribution aims to present a novel method termed RUSLE-GGS (RUSLE-GIS-GLUE-SDR) and has the following specific objectives: (1) describing the technical details of RUSLE-GGS, which unlike previous methodologies can potentially provide reliable ERs estimates at country scale to address the urgency to quantify the dynamics of soil erosion in developing countries in accordance to Goal 15 from the UN 2030 Agenda for Sustainable Development; and (2) elaborating on the application of RUSLE-GGS to Peru for the years 1990, 2000 and 2010.

Section snippets

Geoenvironmental conditions

Peru is located on the Neotropic ecoregion (Fig. 1a). It occupies 1.29 × 106 km2 and traditionally has been divided into three main natural regions (Fig. 1b and c), namely: coastal (western), Andean(central), and Amazonian(eastern), which occupy 12%, 28%, and 60% of the Peruvian territory, respectively. The main biomes in Peru (Fig. 1c) are deserts and xeric shrublands (coastal region), montane grasslands and shrublands (Andean region), and tropical and subtropical moist broadleaf forests

RUSLE-GGS efficiency

Equations A.1 through A.5 were used to build 24 ER samples (Ey[1],,Ey[24]), and subsequently, by using 6 SDR transfer functions (Table 3), 144 area-specific sediment yield samples (SSY*) were obtained for each year y1990,2000,2010, and for each station in Table 2. Thus, this study is based on 1728SSY* samples whose likelihood were evaluated by using the Bias function (Eq. (2)) and the Nash-Sutcliffe index (Eq. (3)).

A custom computer program was built to analyze the spatio-temporal likelihood

RUSLE-GGS, the proposed methodology

The critical affectation of many developing countries by soil erosion has been reported in several studies (Pimentel et al., (1995), Pham et al., (2001), Alcantara-Ayala, (2002); Ananda and Herath, (2003), , (2006), Labrière et al., (2015), Mondal et al., (2017), Borrelli et al., (2017)). As described in Section 2.1, Peru, an upper-middle-income economy, epitomizes such situation. It is therefore reasonable arguing that it might be much worse in poorer countries.

Soil erosion studies require

Conclusions and implications

The profuse literature review presented in this study indicates that soil erosion in developing countries is a matter of serious concern and that Peru, an upper-middle-income economy, presents erosion features that exemplify such situation. Thus, it is reasonably grounded to state that the quantification of soil erosion rates (ERs) in developing countries needs to be addressed with particular urgency. This is however challenging because they commonly suffer from an inherent paucity of

CRediT authorship contribution statement

Miluska A. Rosas: Methodology, Software, Validation, Formal analysis, Investigation, Data curation, Writing - original draft, Writing - review & editing, Visualization. Ronald R. Gutierrez: Conceptualization, Methodology, Software, Validation, Formal analysis, Investigation, Resources, Data curation, Writing - original draft, Writing - review & editing, Visualization, Supervision, Project administration, Funding acquisition.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgments

This project was funded by CONCYTEC within the framework of the 012-2013-FONDECYT Agreement. The authors started this study under the guidelines of GERDIS-PUCP (Pontificia Universidad Católica del Perú). We would like to thank the Servicio Nacional de Meteorología e Hidrología and the Instituto Geofísico del Perú for providing valuable data for this study. The authors also appreciate the technical discussions with Dr. Waldo Lavado and Dr. Sergio Morera. Dr. Gutierrez thanks to the Universidad

References (1)

    Cited by (0)

    View full text