Environmental changes in biodiversity hotspot ecosystems of
South Ecuador: RESPonse and feedback effECTs (FOR2730)

Abstract:

Soils of Páramo ecosystems regulate the water supply to many Andean populations. In spite of being a necessary input to distributed hydrological models, regionalized soil water retention data from these areas are currently not available. The investigated catchment of the Quinuas River has a size of about 90 km2 and comprises parts of the Cajas National Park in southern Ecuador. It is dominated by soils with high organic carbon contents, which display characteristics of volcanic influence. Besides providing spatial predictions of soil water retention at the catchment scale, the study presents a detailed methodological insight to model setup and validation of the underlying machine learning approach with random forest. The developed models performed well predicting volumetric water contents between 0.55 and 0.9 cm3 cm? 3. Among the predictors derived from a digital elevation model and a Landsat image, altitude and several vegetation indices provided the most information content. The regionalized maps show particularly low water retention values in the lower Quinuas valley, which go along with high prediction uncertainties. Due to the small size of the dataset, mineral soils could not be separated from organic soils, leading to a high prediction uncertainty in the lower part of the valley, where the soils are influenced by anthropogenic land use.

Abstract:

This paper focuses on the analysis of precipitation patterns, using a Local Area Weather Radar to collect information about the precipitation distribution in an Andean region of southern Ecuador (cities of Loja, Zamora and Catamayo). 54 representative events were selected to develop daily precipitation maps and to obtain their relevant characteristics, which were related to the topography and the season. The results showed that a strong correlation between the areas covered by precipitation (RA coefficient) and the season exists. In general, humid air masses come from the east (Amazon Basin), but during the main rainy season (December to April), humidity also frequently enters the study region from the west (Pacific Ocean). The rainy season is characterized by convective precipitation, associated with higher evaporation rates during austral summer. The relatively dry season is formed between May and November, but considerable precipitation amounts are registered, too, due to advective moisture transport from the Amazon Basin, a result of the predominant tropical easterlies carrying the humidity up the eastern escarpment of the Andes, generally following the natural course of the drainage systems.

Abstract:

Tropical forests are significant carbon sinks and their soils’ carbon storage potential is immense. However, little is known about the soil organic carbon (SOC) stocks of Tropical mountain areas whose complex soil-landscape and difficult accessibility pose a challenge to spatial analysis. The choice of methodology for spatial prediction is of high importance to improve the expected poor model results in case of low predictor-response correlations. Four aspects were considered to improve model performance in predicting SOC stocks of the organic layer of a tropical mountain forest landscape: Different spatial predictor settings, predictor selection strategies, various machine learning algorithms and model tuning. Five machine learning algorithms: random forests, artificial neural networks, multivariate adaptive regression splines, boosted regression trees and support vector machines were trained and tuned to predict SOC stocks from predictors derived from a digital elevation model and satellite image. Topographical predictors were calculated with a GIS search radius of 45 to 615 m. Finally, three predictor selection strategies were applied to the total set of 236 predictors. All machine learning algorithms—including the model tuning and predictor selection—were compared via five repetitions of a tenfold cross-validation. The boosted regression tree algorithm resulted in the overall best model. SOC stocks ranged between 0.2 to 17.7 kg m-2, displaying a huge variability with diffuse insolation and curvatures of different scale guiding the spatial pattern. Predictor selection and model tuning improved the models’ predictive performance in all five machine learning algorithms. The rather low number of selected predictors favours Forward compared to backward selection procedures. Choosing predictors due to their indiviual performance was vanquished by the two procedures which accounted for predictor interaction.

Abstract:

With respect to sampling for regression-based digital soil mapping (DSM), the above all aim is to ensure that the spatial variability of the soil is well-captured without introducing any bias, while the design remains feasible according to operational constraints such as accessibility, man power and cost. Representativeness of the sample concerning the population to be sampled needs to be guaranteed in any regression-based modelling approach. Four selected sampling designs were adapted to show that basically any design may be optimised to represent the n-dimensional predictor space of a particular area, while selecting points is only permitted from a small accessible sub-area or from outside the area. Sampling efficiency may be evaluated based on the representation of the predictor space. However, not only each predictor’s probability function but also the interaction between predictors may have to be considered, to select a representative sample. Instead of sampling a previously un-sampled area with limited accessibility, the four sampling designs may also be used to subsample an existing dataset and, thereby, optimise a suboptimal dataset based on the predictor space of the area which shall be mapped by DSM.

Abstract:

A digital soil mapping approach is applied to a complex, mountainous terrain in the Ecuadorian Andes. Relief features are derived from a digital elevation model and used as predictors for topsoil texture classes sand, silt, and clay. The performance of three statistical learning methods is compared: linear regression, random forest, and stochastic gradient boosting of regression trees. In linear regression, a stepwise backward variable selection procedure is applied and overfitting is controlled by minimizing Mallow’s Cp. For random forest and boosting, the effect of predictor selection and tuning procedures is assessed. 100-fold repetitions of a 5-fold cross-validation of the selected modelling procedures are employed for validation, uncertainty assessment, and method comparison. Absolute assessment of model performance is achieved by comparing the prediction error of the selected method and the mean. Boosting performs best, providing predictions that are reliably better than the mean. The median reduction of the root mean square error is around 5%. Elevation is the most important predictor. All models clearly distinguish ridges and slopes. The predicted texture patterns are interpreted as result of catena sequences (eluviation of fine particles on slope shoulders) and landslides (mixing up mineral soil horizons on slopes).

Abstract:

The sloping mire landscape of the investigation area, in the southern Andes of Ecuador, is dominated by stagnic soils with thick organic layers. The recursive partitioning algorithm Random Forest was used to predict the spatial water stagnation pattern and the thickness of the organic layer from terrain attributes. Terrain smoothing from 10 to 30m raster resolution was applied in order to obtain the best possible model. For the same purpose, several model tuning parameters were tested and a prepredictor selection with the R-package Boruta was applied. Model versions were evaluated and compared by 100 repetitions of the calculation of the residual mean square error of a five-fold cross-validation. Position specific density functions of the predicted soil parameters were then used to display prediction uncertainty. Prepredictor selection and tuning of the Random Forest algorithm in some cases resulted in an improved model performance.We therefore recommend testing prepredictor selection and tuning to make sure that the best possible model is chosen.This needs particular emphasis in complex tropical mountain soil-landscapes which provide a real challenge to any soil mapping approach but where Random Forest has proven to be successful due to the testing of model tuning and prepredictor selection.

Abstract:

The World Reference Base for Soil Resources (WRB) (FAO, IUSS Working Group WRB, 2007) at present does not acknowledge the spatial soil continuum, but provides a sound basis to do so. Using methods from statistical learning theory to develop digital soil maps is much more efficient and precise while regionalising soil diagnostic properties instead of complex entities such as the soil units assigned by the WRB. Particularly in providing spatial soil information displayed in digital soil maps, any aggregation of this spatial soil information to soil units means a loss of information. The soil landscape can be systematically described in its spatial continuum simply by the vertical order and extent of the WRB diagnostic horizons. The diagnostic horizons are related in their thickness to a standard depth and listed from top to bottom in order of appearance. Typical diagnostic horizon thickness and occurrence probability were predicted from terrain parameters by classification and regression trees (CART), throughout the research area in southern Ecuador. The two disadvantages of CART, abrupt prediction class boundaries and dependence on the dataset, were addressed by hundredfold model runs on different data subsets, leading to a range of possible predictions. Prediction uncertainty was included in the digital soil maps by calculating these predictions' means and standard deviations as well as by horizon occurrence probability prediction. Model performance was evaluated by means of hundredfold external cross validation. Terrain parameters were found to have a strong influence on diagnostic topsoil properties. However, no influence on the vertical profile differentiation was observed. Hence predicting horizon thickness and subsoil diagnostic properties was difficult. The systematic description of the soil continuum of this particular soillandscape resulted in histic and stagnic soil parts dominating the first 100 cm of the soil column for most of the area.

Abstract:

Soil-landscapes are diverse and complex due to the interaction of pedogenetic, geomorphological and hydrological processes. The resulting soil profile reflects the balance of these processes in its properties. Early conceptual models have by now resulted into quantitative soil-landscape models including soil variation and its unpredictability as a key soil attribute. Soils in the Andean mountain rainforest area of southern Ecuador are influenced by hillslope processes and landslides in particular. The lack of knowledge on the distribution of soils and especially physical soil properties to understand slope failure, resulted in the study of this particular soil-landscape by means of statistical models relating soil to terrain attributes, i.e. predictive soil mapping. A 24 terrain classes comprising sampling design for soil investigation in mountainous areas was developed to obtain a representative dataset for statistical modelling. The soils were investigated by 56 profiles and 315 auger points. The Reference Soil Groups (RSGs) Histosol, Stagnosol, Umbrisol, Cambisol, Leptosol and Regosol were identified according to the World Reference Base for Soil Resources (WRB). Classification tree models and a probability scheme based on WRB hierarchy were applied to include RSG prediction uncertainty in a digital soil map. Histosol probability depended on hydrological parameters; highest Stagnosol probability was found on slopes < 40° and above 2146 m a.s.l. Poor model performance, probably due to the prediction of complex categories (RSGs) and WRB inconsequence (absolute and relative value criteria), led to the proposal of ?incomplete soil classification? by relating the thickness of the WRB?s diagnostic horizons as percentage to the upper 100 soil centimetres, including the organic layer. Typical diagnostic horizons histic, humic, umbric, stagnic and cambic were regionalised in their thickness and occurrence probability by classification and regression trees (CART). Prediction uncertainty was addressed with hundredfold model runs based on different random Jackknife partitions of the dataset. Whether the first mineral soil horizon displays stagnic properties or not, likely depends on physical soil properties in addition to terrain parameters. Incomplete soil classification resulted in histic and stagnic soil parts dominating the first 100 cm of the soil volume for most of the research area. While soil profiles and auger points were described in their horizon composition, thickness, Munsell colour and soil texture by finger method (FAO, 2006), soil cohesion, bulk density and texture by pipette and laser were analysed in soil profiles only. Texture results by pipette compared to laser method, showed the expected shift to higher silt and lower clay contents. Linear regression equations were adapted. Pedotransfer functions to predict physical soil properties from the bigger auger dataset analysed by field texture method only, could not be developed, because field texture analysis did not provide satisfying results. It was therefore not possible to correct its results with the more precise laboratory data. Comparing CART and Random Forest (RF) in their model performance to predict topsoil texture and bulk density as well as mineral soil thickness by hundredfold model runs with random Jackknife partitions, RF predictions resulted more powerful. Altitude a.s.l. was the most important predictor for all three soil parameters. Increasing sand/ clay ratios with increasing altitude, on steep slopes and with overland flow distance to the channel network are caused by shallow subsurface flow removing clay particles downslope. Deeper soil layers are not influenced by the same process and therefore showed different texture properties. Terrain parameters could only explain the spatial distribution of topsoil properties to a limited extent, subsoil properties could not be predicted at all. Other parameters that likely influence soil properties within the investigation area are parent material and landslides. Strong evidence was found that topsoil horizons did not form from the bedrock underlying the soil profile. Parent material changes within short distance and often within one soil profile. Landslides have a strong influence on soil-landscape formation in shifting soil and rock material. Soil mechanical and hydrological properties in addition to terrain steepness were hypothesized to be the major factors in causing soil slides. Thus, the factor of safety (FS) was calculated as the soil shear ratio that is necessary to maintain the critical state equilibrium on a potential sliding surface. The depth of the failure plane was assumed at the lower boundary of the stagnic soil layer or complete soil depth, depending on soils being stagnic or non-stagnic. The FS was determined in dependence of soil wetness referring to 0.001, 0.01, 0.1 and 3 mm/h net rainfall rate. Sites with a FS &#8805; 1 at 3 mm/h (complete saturation) were classified as unconditionally stable, sites with a FS < 1 at 0.001 mm/h as unconditionally unstable. The latter coincided quite well with landslide scars from a recent aerial photograph.

Abstract:

Landslides are a common phenomenon within the Ecuadorian Andes and have an impact on soil-landscape formation. Landslide susceptibility was determined in a steep mountain forest region in Southern Ecuador. Soil mechanical and hydrological properties in addition to terrain steepness were hypothesised to be the major factors in causing soil slides. Hence, the factor of safety (FS) was calculated as the soil shear ratio that is necessary to maintain the critical state equilibrium on a potential sliding surface. Regression tree (RT) and Random Forest (RF) models were compared in their predictive force to regionalise the depth of the failure plane and soil bulk density based on terrain parameters. The depth of the failure plane was assumed at the lower boundary of the stagnic soil layer or soil depth respectively, depending on soils being stagnic or nonstagnic. FS was determined in dependence of soil wetness referring to 0.001, 0.01, 0.1 and 3 mm h&#8722;1 net rainfall rates. Sites with FS&#8805;1 at 3 mm h&#8722;1 (complete saturation) were classified as unconditionally stable; sites with FSb1 at 0.001 mm h&#8722;1 as unconditionally unstable. Bulk density and the depth of the failure plane were regionalised with RF which performed better than RT. Terrain parameters explained the spatial distribution of soil bulk density and the depth of the failure plane only to a relatively small extent which is reasonable due to frequent translocation of soil material by landslides. Nevertheless, their prediction uncertainty still allowed for a reasonable prediction of nconditionally unstable sites.

Abstract:

The number of previous hydrological studies concerning water flow paths in tropical montane rainforest is small. However, due to the increasing pressure of deforestation and land use change comprehensive knowledge of these natural ecosystems is needed if sustainable land use strategies should keep negative effects of human impacts on water flow paths as low as possible. In this context, present work addresses the identification, characterisation, and modelling of water flow paths in soils of an undisturbed and landslide affected natural Andean forest ecosystem in the south of Ecuador whose deforestation rate is one of the highest in South America. In an investigation area situated in the Andes of South Ecuador, in gentler slopes and altitudes above 2100 m ASL mainly Stagnosols and Histosols with stagnic colour pattern and low to negligible rock fragment content prevail. With increasing altitude the abundance of these soils increase, while the presence of Cambisols and Regosols is most pronounced below 2100 m ASL and clearly correlated with the slope angle. Therefore, these soils were mainly encountered in steeper, particularly landslide affected sites often resembling a melange of fine soil and high contents of rock fragments. Aside the investigation of the influence of the rock fragment content on soil hydrological- and physical parameters such as the relationship between rock fragment content and saturated hydraulic conductivity of the soil, present study aims particularly to investigate flow paths of water in soils of landslide affected and unaffected hillslopes. Therefore, we employed conventional field- and laboratory methods, dye tracer experiments including an appropriate image processing technique, as well as statistical models. Results show that both rock fragment content and bulk density control significantly, but not largely the saturated hydraulic conductivity of the mineral soils. Dye tracer experiments and soil parameters document a deeper percolation in the landslide affected hillslopes than in the landslide unaffected hillslopes, where we found preferential flow in root channels with low soil matrix interaction as dominant flow mechanism. A surface near quasi impervious layer along the interface between topsoil and subsoil limits percolation of the water giving the prerequisites of a lateral shallow subsurface flow along the interface between topsoil and organic layer. This is in line with previous studies performed in the same investigation area which already proved indirectly the existence of this flow. However, in none of these studies the shallow subsurface flow was assigned to certain slope inclinations or altitudes. Due to a recently published digital soil map and the results we obtained from the landslide unaffected sites, we know that particularly in hillslopes of less than 30 ° above 2100 m ASL prerequisites are given for spatially extended shallow subsurface flow. However, even if these prerequisites are not evident for the landslide affected hillslopes, we cannot exclude the possibility of shallow subsurface flow occurrence here since soil cover of the steep terrain is relatively shallow while rainfall is high throughout the year. Therefore, and given that key parameters such as permeability of subsoil and bedrock, interception and evaporation remain unclear or were investigated exclusively such as the spatial variability of the saturated hydraulic conductivity, we conducted a series of virtual experiments in order to assess the potential occurrence of shallow subsurface flow in Cambisols below 2100 m ASL. In these experiments we also included the organic layer being highly abundant in the investigation area, whose hydraulic parameters were estimated by means of inverse numerical modelling. The virtual experiments were based on a two dimensional finite element model representing a steep forested hillslope transect of ~54 m length. Aside soil properties, evapotranspiration and interception, the model included the spatial variability of the saturated hydraulic conductivity, the pressure head and their spatial trends. The results of virtual experiment series show that a sound evidence of the key parameters aforementioned is obligate if process conceptualisation regarding shallow subsurface flow generation, but also landslide initiation, solute and matter transport is in the spotlight.

Abstract:

Soil landscape modelling is based on understanding the spatial distribution patterns of soil characteristics. A model relating the soil?s properties to its position within the landscape is used to predict soil properties in other similar landscape positions. To develop soil landscape models, the interaction of geographic information technology, advanced statistics and soil science is needed. The focus of this work is to predict the distribution of the different soil types in a tropical mountain forest area in southern Ecuador from relief and hydrological parameters using a classification tree model (CART) for soil regionalisation. Soils were sampled along transects from ridges towards side valley creeks using a sampling design with 24 relief units. Major soil types of the research area are Histosols associated with Stagnosols, Cambisols and Regosols. Umbrisols and Leptosols are present to a lesser degree. Stagnosols gain importance with increasing altitude and with decreasing slope angle. Umbrisols are to be found only on slopes <30°. Cambisols occurrence might be related to landslides.The CART model was established by a data set of 315 auger sampling points. Bedrock and relief curvature had no influence on model development. Applying the CART model to the research area Histosols and Stagnosols were identified as dominant soil types. Model prediction left out Cambisols and overestimated Umbrisols, but showed a realistic prediction for Histosols, Stagnosols and Leptosols.

Abstract:

Ref.: Ms. No. EJFOR-D-08-00181R1 accepted on 16-06-2009 Ecosystem analysis is typically done by determination of species composition, structural ex-ploration, determination of matter and energy fluxes and/or system analyses based on deter-ministic or probabilistic/stochastic model approaches. However, regarding ecosystem dynam-ics, temporal structure, information content, complexity of signals, and their modifications when subsequently passing through different subsystems, have not intensively been studied to date. Structure in time series characterised by information and complexity measures may pro-vide additional, powerful tools to analyse state and dynamics of ecosystems. Along their path through ecosystem compartments e.g. hydrological signals are transformed in several ways, comprising changes in randomness, autocorrelation structures, and smoothness. Thus, time series analyses with complexity and information measures is of interest for a holistic under-standing of ecosystem behaviour and early indications of natural and anthropogenic distur-bances of ecosystems like ecosystem degradation and climate change. Further, these measures provide additional criteria for the calibration of model parameters tests of model validity, and determination of the necessary degree of complexity of process models. In this paper we pre-sent the outcome from applications of information and complexity measures to hydrological time series in two climatically different forest ecosystems in South Germany and Southern Ecuador. The values of information and complexity measures have a clearly different range for the different hydrological parameters but for ecosystems of the same type like mountain forest the values of information and complexity measures exhibit similar behaviour for the hydrological parameters. We hypothesize that complexity of hydrological time series in-creases with the number of abiotic and biotic variables involved in the generating process of the time series. Thus, complexity should have a minimum in the precipitation signal which is controlled by abiotic, atmospheric factors only, and reach a maximum in the root zone where the interaction of abiotic and biotic variables is high. Hydrological time series under study cover the sequence of hydrological signals from open precipitation, throughfall, sapflow, wa-ter fluxes in the soil compartment and system discharge. We detected pronounced data aggre-gation and transformation effects of hydrological signals along their path through subsystems in terms of information propagation. We further found similar patterns in different ecosystems of the same general type. As a result of intensive abiotic and biotic interactions, a pronounced maximum of complexity was found in the moisture signal of the soil compartment.