Modeling stream flow and sediment yield using the SWAT model: a case study of Ankara River basin,Turkey

The Soil and Water Assessment Tool (SWAT) was tested for prediction of stream flow and sediment yield in the Ankara basin, Turkey. The overall objective of this study was to evaluate the performance and applicability of the SWAT and generate a soil erosion map. Thirteen years of daily/monthly flow and monthly sediment data were used for calibration and validation. Model performance was evaluated using statistical measures to assess the applicability of the model in simulating stream flow and sediment yield during calibration (1989–1996) and validation (1982–1984) periods. Nash Sutcliffe efficiency (NSE), relative error (RE), and R² (coefficient of determination) for daily flow were computed as 0.61, ?0.55, and 0.78, respectively; and as 0.79, ?0.58, and 0.89 for monthly flow during the calibration. Statistical comparisons of sediment yield produced values for NSE, RE, and R² of 0.81, ?1.55, and 0.93, respectively, during the calibration. The resulting map suggests that significant portions of urbanized and highly cultivated areas in the vicinity of stream channels are particularly vulnerable to soil erosion. SWAT satisfactorily simulated hydrology and sediment yield and can be used as a tool in decision-making for water resources planning in a basin with similar characteristics.     

Radial Basis Function Network for Ore Grade Estimation

This paper highlights the performance of a radial basis function (RBF) network for ore grade estimation in an offshore placer gold deposit. Several pertinent issues including RBF model construction, data division for model training, calibration and validation, and efficacy of the RBF network over the kriging and the multilayer perceptron models have been addressed in this study. For the construction of the RBF model, an orthogonal least-square algorithm (OLS) was used. The efficacy of this algorithm was testified against the random selection algorithm. It was found that OLS algorithm performed substantially better than the random selection algorithm. The model was trained using training data set, calibrated using calibration data set, and finally validated on the validation data set. However, for accurate performance measurement of the model, these three data sets should have similar statistical properties. To achieve the statistical similarity properties, an approach utilizing data segmentation and genetic algorithm was applied. A comparative evaluation of the RBF model against the kriging and the multilayer perceptron was then performed. It was seen that the RBF model produced estimates with the R ² (coefficient of determination) value of 0.39 as against of 0.19 for the kriging and of 0.18 for the multilayer perceptron.     

A framework for uncertainty assessment in simulation models

In this article, we introduce a conceptual framework for systematic identification and assessment of sources of uncertainty in simulation models. This concept builds on a novel typology of uncertainty in model validation and extends the GIScience research focus on uncertainty in spatial data to uncertainty in simulation modelling. Such a concept helps a modeller to interpret and handle uncertainty in order to efficiently optimise a model and better understand simulation results.

To illustrate our approach, we apply the proposed framework for uncertainty assessment to the TREE LIne Model (TREELIM), an individual-based model that simulates forest succession at the alpine tree line. Using this example, uncertainty is identified in the modelling workflow during conceptualisation, formalisation, parameterisation, analysis and validation. With help of a set of indicators we quantify the emerging uncertainties and assess the overall model uncertainty as a function of all occurring sources of uncertainty.

An understanding of the sources of uncertainty in an ecological model proves beneficial for: (1) developing a structurally valid model in a systematic way; (2) deciding if further refinement of the conceptual model is beneficial for the modelling purpose; and (3) interpreting the overall model uncertainty by understanding its sources. Our approach results in a guideline for assessing uncertainty in the validation of simulation models in a feasible and defensible way, and thus functions as a toolbox for modellers. We consider this work as a contribution towards a general concept of uncertainty in spatially explicit simulation models.     

基于SWAT模型的乌鲁木齐河流域径流模拟

为了更好地研究西北干旱区高山河流对气候变化的响应,在新疆乌鲁木齐河上游汇水区构建SWAT分布式水文模型,结合Sufi-2算法对模型进行参数率定、敏感性分析以及不确定性分析,对乌鲁木齐河英雄桥水文站以上2000-2011年日径流进行模拟。结果表明:(1)河岸调蓄的基流α因子、土壤饱和导水率、地下水的时间延迟、气温降水垂直变率以及相关融雪参数比较敏感。(2)模型模拟结果与观测流量过程线拟合程度较好,日值模拟结果Ens>0.75,PBIAS介于±25%,R²达到0.8以上。(3)较高的P-factor及较低的R-factor显示模型不确定性较小。总体来说,SWAT模型在乌鲁木齐河上游区的适用性比较理想。     

Evaluation of210Pb dating in lake sediments using stable Pb,Ambrosia pollen,and137Cs

We used sediment chronology data from fourteen published studies of lake cores across much of North America and Scandinavia in order to make a large scale comparison of the different dating techniques. The uncertainty of²¹⁰Pb derived dates was determined using common sediment event markers: the stable Pb rise, the¹³⁷Cs rise, and theAmbrosia pollen rise. For all data combined, the 95% confidence intervals for the stable Pb rise and theAmbrosia rise, were approximately 30 years. These 95% confidence intervals are slightly higher than those derived by First-Order Error analysis performed by others on²¹⁰Pb derived dates. When comparing the concordance of two²¹⁰Pb models (CRS and CIC) against markers of known history, we found that the CRS model dates (constant rate of supply) had consistently better agreement than the CIC model dates (constant initial concentration). Major discrepancies between¹³⁷Cs and²¹⁰Pb were common, but were consistently more severe in sediments of soft water lakes pointing to an inability of sediments with low mineral content to immobilize Cs.This is the ninth of a series of papers to be published by this journal following the 20 ^th anniversary of the first application of²¹⁰Pb dating of lake sediments, Dr. P. G. Appleby is guest editing this series.     

Paleoecological evidence of major declines in total organic carbon concentrations since the nineteenth century in four nemoboreal lakes

A decade of widespread increases in surface water concentrations of total organic carbon (TOC) in some regions has raised questions about longer term patterns in this important constituent of water chemistry. This study uses near-infrared spectroscopy (NIRS) to infer lake water TOC far beyond the decade or two of observational data generally available. An expanded calibration dataset of 140 lakes across Sweden covering a TOC gradient from 0.7 to 24.7 mg L⁻¹ was used to establish a relationship between the NIRS signal from surface sediments (0–0.5 cm) and the TOC concentration of the water mass. Internal cross-validation of the model resulted in an R ² of 0.72 with a root mean squared error of calibration (RMSECV) of 2.6 mg L⁻¹. The TOC concentrations reconstructed from surface sediments in four Swedish lakes were typically within the range of concentrations observed in the monitoring data during the period represented by each sediment layer. TOC reconstructions from the full sediment cores of four lakes indicated that TOC concentrations were approximately twice as high a century ago.     

Influence of averaging interval on shear velocity estimates for aeolian transport modeling

Recent investigations of aeolian transport have focused on increasingly short time scales because of growing recognition that wind unsteadiness is a major factor in the dynamics of sediment transport. However, the statistical reliability of shear velocity (u_*) estimates becomes increasingly uncertain as averaging interval is decreased. This study provides an empirical assessment of the influence of averaging interval on the reliability of u_* estimates. The data consist of 15-min wind-speed profiles (1 Hz sampling) collected at four coastal sites. Each profile was subdivided into progressively shorter fixed-length time intervals, and estimates of u_* and the 95% confidence interval for u_* were determined for each time-block using standard statistical techniques.The logarithmic model accurately represents the measured wind-speed profiles, even with relatively brief averaging intervals. Mean r² values remain robust down to block lengths as short as 10–20 s, typically retaining better than 98% of the r² value found for the full-length data sets. Fewer than 2% of the individual 10-s blocks had r² values less than 0.9. However, mean confidence intervals typically expanded by 70–80% of the full-record value as block length decreased from 900 to 10 s. For highly log-linear profiles, this amounted to an absolute increase from about ±8% to only ±14% of u_*, so that the additional information gained through the use of shorter averaging intervals may outweigh the increase in statistical uncertainty. Nevertheless, given that rates of aeolian transport are generally modeled as a function of u_*³, this increase in uncertainty may be significant for transport modeling. Thus, very short averaging intervals should be used with caution when predicting aeolian sediment flux. It is proposed that transport modeling should incorporate the shear velocity confidence interval as an indicator of the potential error associated with this source of uncertainty.     

A model for inferring dissolved organic carbon (DOC) in lakewater from visible-near-infrared spectroscopy (VNIRS) measures in lake sediment

We developed an inference model to infer dissolved organic carbon (DOC) in lakewater from lake sediments using visible-near-infrared spectroscopy (VNIRS). The inference model used surface sediment samples collected from 160 Arctic Canada lakes, covering broad latitudinal (60–83°N), longitudinal (71–138°W) and environmental gradients, with a DOC range of 0.6–39.6 mg L⁻¹. The model was applied to Holocene lake sediment cores from Sweden and Canada and our inferences are compared to results from previous multiproxy paleolimnological investigations at these two sites. The inferred Swedish and Canadian DOC profiles are compared, respectively, to inferences from a Swedish-based VNIRS-total organic carbon (TOC) model and a Canadian-based diatom-inferred (Di-DOC) model from the same sediment records. The 5-component Partial Least Squares (PLS) model yields a cross-validated (CV) R_CV² R_{CV}^{2}  = 0.61 and a root mean squared error of prediction (RMSEP _CV ) = 4.4 mg L⁻¹ (11% of DOC gradient). The trends inferred for the two lakes were remarkably similar to the VNIRS-TOC and the Di-DOC inferred profiles and consistent with the other paleolimnological proxies, although absolute values differed. Differences in the calibration set gradients and lack of analogous VNIRS signatures in the modern datasets may explain this discrepancy. Our results corroborate previous geographically independent studies on the potential of using VNIRS to reconstruct past trends in lakewater DOC concentrations rapidly.     

改进SWAT模型水库模块及其在水库控制流域径流模拟中的应用

为在径流模拟中充分考虑水库的综合调蓄作用,改善了SWAT（The Soil and Water Assessment Tool）模型的水库算法,在水库模块中增加基于水库供水发电调度规则的水库出流模拟算法,以提高径流模拟精度。选择东江流域作为典型流域,分别应用改进前后的SWAT模型进行径流模拟,并分析模型改进的效果,结果表明：修改后的SWAT模型解决了水库模拟出流量在非汛期过低、汛期过大的问题,能更有效地模拟枯季径流和洪峰流量,明显提高了月径流及日径流的模拟精度,其中,月径流模拟中率定期内R2提高了0.06~0.22,NS提高了0.06~0.52,验证期内R2提高了0.1~0.19,NS提高了0.12~1.22;日径流模拟中率定期内R2提高了0.04~0.16,NS提高了0.04~0.46,验证期内R2提高了0.11~0.15,NS提高了0.15~1.21。     

Palaeolimnological validation of estimated reference values for a lake profundal macroinvertebrate metric (Benthic Quality Index)

Modern assessment and monitoring of aquatic ecosystems is increasingly based on biota and the “reference condition” approach, in which the observed values (O) of biological variables are compared to those expected in the absence of human disturbance (E). To use this approach, correct estimation and validation of reference conditions are critical. Because appropriate modern or historical data are never available for this approach, palaeolimnological data offer an alternative. We used a calibration data set from 73 profundal sites in semi-pristine Finnish lakes to construct a regression model for estimating expected values for the chironomid Benthic Quality Index (BQI)—a macroinvertebrate metric widely used in bioassessment—from environmental variables that are insensitive to human disturbance. For comparison, reference values were estimated using the European legislative rationale based on a priori lake typology. Performance of the alternative approaches was assessed by internal ‘leave-one-out’ cross-validation using the calibration set and by external cross-validation using independent palaeolimnological data on BQI values representing the historical pristine status of 24 lake basins. Additionally, for 19 of these sites, which vary in their degree of human impact, the ratio of present BQI to that in pristine condition, which shows the degree of actual change, if any, was calculated from palaeolimnological data and compared with the O/E ratios based on the present chironomid data and estimated E. A linear regression model with mean depth and mean/maximum depth ratio as independent variables estimated the reference values of BQI much closer to the observed ones (r ² = 0.58, RMSEP = 0.65 and r ² = 0.71 RMSEP = 0.55; for internal and external cross-validation, respectively) than did the typology approach (r ² = 0.28, RMSEP = 0.86; r ² = 0.10, RMSEP = 0.97). The regression approach also yielded O/E ratios more similar to the actual ones (r ² = 0.79, RMSEP = 0.09) than did the typology approach (r ² = 0.62, RMSEP = 0.23). Our results strongly support the use of lake morphometric variables and modelling instead of categorical lake typology for the establishment of reference conditions for profundal macroinvertebrate communities and demonstrate the utility of palaeolimnological data in the validation of reference values and assessment methods.     

Water-chemistry transfer functions for epiphytic diatoms in standing freshwaters and a comparison with models based on littoral sediment assemblages (Flanders, Belgium)

Quantitative inference models for water-chemistry variables are derived from epiphytic diatom assemblages in 186 lentic and mostly shallow freshwaters in lower Belgium (Flanders). When the complete pH range is considered (pH 3.4–9.3), robust transfer functions are obtained for median pH (jack-knifed r ² = 0.88, RMSEP = 0.38 pH units or 6.4% of the observed range) and dissolved inorganic carbon concentration (jack-knifed r ² = 0.86, RMSEP = 0.194 log₁₀ mg DIC l⁻¹ or 10.2% of the observed range) by means of weighted-averaging partial least squares regression (WA-PLS). For these variables, the calibration models are as reliable as those based on sedimentary diatom assemblages. Inferences of pH may be improved by combining estimates from epiphytic and sediment assemblages. In circumneutral and alkaline conditions, WA-PLS calibration of maximum or median total phosphorus is possible (log-transformed; jack-knifed r ² = 0.64 or 0.66 and RMSEP = 14% or 12.3% of the observed range, respectively). It makes little difference if taxa showing no response to TP are taken into consideration or not. These models considerably expand the prospects of using historical herbarium materials to hindcast environmental conditions and also allow more accurate interpretation of current compositional changes in epiphytic communities. Compared to littoral sediment assemblages, fewer water-column variables can be inferred reliably from epiphyton. This probably results from differences between the effective gradients in both habitats, together with lower in situ species diversity and less effective spatial integration (i.e. lower recruitment of phytoplankton) in the epiphyton. A comparison of the HOF response-model types and WA-optima of diatom taxa for epiphytic and sediment assemblages shows that the relationship to individual variables, and in particular to those related to trophic status, may differ with habitat. Thus, the combination of samples from both habitat types in the same calibration model is not recommended. Electronic Supplementary Material Supplementary material is available and is accessible for authorised users in the online version of this article at     

Difficulties in interpreting fast mixing in the radiometric dating of sediments using 210Pb and 137Cs

²¹⁰Pb geochronologies should be validated with independent tracers such as ¹³⁷Cs. In the cases with constant ²¹⁰Pb activity in the topmost sediments, the presence of a distinct ¹³⁷Cs peak within the ²¹⁰Pb plateau has been used as a definitive demonstration of acceleration (increase in the sedimentation rate in recent years) versus fast mixing. Nevertheless, some limitations can be identified in the use of semiquantitative arguments, and a global understanding of the whole ¹³⁷Cs activity profile is then required. Particularly, the incomplete mixing within the top sediment zone (described through the Incomplete Mixing Zone model) can explain quantitatively and simultaneously the ¹³⁷Cs peak and the flattening in the ²¹⁰Pb activity profile. This is demonstrated using selected examples from literature data. Thus, measured constant ²¹⁰Pb activities in the top 6 cm of a sediment core from Lake Zürich. Nevertheless, they found ⁷Be only in the uppermost layer, the distinct ¹³⁷Cs maximum at 6 cm depth, and undisturbed varves. The fast mixing seemed then opposed to common sense. The constant rate of supply model shows acceleration and it adequately matches the position of the 1963 ¹³⁷Cs peak. Nevertheless it fails to explain the whole ¹³⁷Cs profile when handling time series of ¹³⁷Cs atmospheric deposition. Finally, it is shown how the incomplete mixing of the activity (through the pore water) over a certain mass depth at the top sediment, with a finite value of the mixing coefficient, can quantitatively explain the whole activity versus mass depth profiles of ¹³⁷Cs and ²¹⁰Pb, and the presence of ⁷Be only in the uppermost sediment layer. A further validation of these ideas is presented from other literature data.     

Predicting sediment inputs to aquatic ecosystems across England and Wales under current environmental conditions

The awareness of water quality issues has never been higher. As part of its continuing strategic diffuse pollution policy support, ADAS recently undertook to identify catchments across England and Wales that could potentially fail recently proposed suspended sediment yield targets under current environmental conditions. The total suspended sediment loads (SSL) delivered to all rivers were assumed to comprise contributions from diffuse sources in the agricultural and urban sectors, as well as from eroding channel banks and point sources represented by sewage treatment works (STWs). Diffuse agricultural sediment loss to rivers was predicted using the PSYCHIC model. Corresponding inputs from diffuse urban sources were estimated on the basis of an Event Mean Concentration (EMC) methodology. Channel bank sediment inputs were calculated using a prototype national scale model, while point source sediment contributions were based on a register of consented effluent discharges. Modelled SSL were validated (r²=68%) against PARCOM data (1999–2003) for the delivery of sediment to different regions of the UK maritime area. The results of the validation were considered to be realistic for a national scale predictor. The modelling exercise suggested that those catchments currently at risk of exceeding proposed suspended sediment yield critical thresholds are largely confined to upland areas across Wales and northwest England and the chalklands of southern and eastern England.     

Diatom and chrysophyte functions and inferences of post-industrial acidification and recent recovery trends in Killarney lakes (Ontario, Canada)

The surface sediment diatom and chrysophyte assemblages from 33 Sudbury lakes were added to our published 72 lake data set to expand and refine the diatom and chrysophyte-based inference models that we had earlier developed for this region. Our calibration data set now includes 105 lakes, representing gradients for multiple environmental variables (e.g., lakewater pH, metals, and transparency). The revised models are based on the weighted averaging calibration and regression approach and include bootstrap error estimates. The pH model was the strongest (r² _boot = 0.75, RMSE _boot = 0.50). The chrysophyte-inferred pH model (r² _boot = 0.79, RMSE _boot = 0.48) that we developed was as robust as the diatom pH model. Diatom and chrysophyte inferred pH models were then applied to top (surface sediments representing current conditions) and bottom (generally from > 30 cm deep representing pre-industrial conditions) sediment diatom and chrysophyte assemblages of 19 Killarney area lakes near Sudbury. The top and bottom inferred pH results were compared to early-1970s measured pH data. These data suggest that, although many of the poorly buffered Killarney lakes had experienced acidification, marked pH recovery has occurred in many lakes within the last 25 years. Despite the stunning pH recovery, the present-day diatom and chrysophyte assemblages are significantly different from assemblages present during pre-industrial times. Our results suggest that biological recovery may require more time than chemical recovery. It is also likely that these lakes may never recover biologically because other anthropogenic stressors (e.g., climate warming and increased exposure to UV-B radiation) may now have greater influence on biological communities in Killarney/Sudbury area lakes than acidification.     

SWAT分布式流域水文物理模型的改进及应用研究

SWAT (Soil and Water Assessment Tool) 模型是一个集成遥感 (RS)、地理信息系统 (GIS) 和数字高程模型(DEM)技术的先进的分布式流域水文物理模型。为了推动该模型在中国的适应性研究及应用,并改进模型以提高水文模拟的精度,针对模型在中国西北寒旱区的黑河流域和中西部温润的汉江流域的水文模拟中发现的问题进行了扩充和改进,增加了土壤粒径转换模块和天气发生器(WGEN)数据预处理模块,改进了模型中的WGEN算法、潜在蒸散量模拟算法以及气象参数的空间离散方法。利用扩充和改进后的模型对汉江褒河上游江口流域的降雨-径流过程进行了系统的研究。结果表明,不仅模型的使用效率有明显提高,而且改进后模型的效率系数和相关系数也比改进前有较大改善。     

Uncertainty due to DEM error in landslide susceptibility mapping

Terrain attributes such as slope gradient and slope shape, computed from a gridded digital elevation model (DEM), are important input data for landslide susceptibility mapping. Errors in DEM can cause uncertainty in terrain attributes and thus influence landslide susceptibility mapping. Monte Carlo simulations have been used in this article to compare uncertainties due to DEM error in two representative landslide susceptibility mapping approaches: a recently developed expert knowledge and fuzzy logic-based approach to landslide susceptibility mapping (efLandslides), and a logistic regression approach that is representative of multivariate statistical approaches to landslide susceptibility mapping. The study area is located in the middle and upper reaches of the Yangtze River, China, and includes two adjacent areas with similar environmental conditions – one for efLandslides model development (approximately 250 km²) and the other for model extrapolation (approximately 4600 km²). Sequential Gaussian simulation was used to simulate DEM error fields at 25-m resolution with different magnitudes and spatial autocorrelation levels. Nine sets of simulations were generated. Each set included 100 realizations derived from a DEM error field specified by possible combinations of three standard deviation values (1, 7.5, and 15 m) for error magnitude and three range values (0, 60, and 120 m) for spatial autocorrelation. The overall uncertainties of both efLandslides and the logistic regression approach attributable to each model-simulated DEM error were evaluated based on a map of standard deviations of landslide susceptibility realizations. The uncertainty assessment showed that the overall uncertainty in efLandslides was less sensitive to DEM error than that in the logistic regression approach and that the overall uncertainties in both efLandslides and the logistic regression approach for the model-extrapolation area were generally lower than in the model-development area used in this study. Boxplots were produced by associating an independent validation set of 205 observed landslides in the model-extrapolation area with the resulting landslide susceptibility realizations. These boxplots showed that for all simulations, efLandslides produced more reasonable results than logistic regression.     

基于SWAT的典型沼泽性河流径流演变的气候变化响应研究——以三江平原挠力河为例

SWAT（Soil and Water Assessment Tool）是流域尺度的分布式水文模型,具有评价气候变化对径流影响的优势。利用SWAT模拟了三江平原典型沼泽性河流——挠力河流域3个水文站（上游的宝清站、保安站和中游的菜嘴子站）1974～1992年年径流量演变特征及变化趋势。在对模型参数敏感性分析的基础上,对模型的参数进行了率定和验证,率定期为1975～1982年,验证期为1983～1992年。率定期的模型效率指数ENS都大于0．85,皮尔逊相关系数都大于0．9,相对误差都小于10％;验证期,模型效率指数ENS有所减小,但也都大于0．61,模型对年径流的模拟结果令人满意。将率定的SWAT应用于气候变化的水文响应研究,结果发现,1995～2004年相对1975—1985年的年径流量变化只有部分是由气候因素引起的,气候因素对3个水文站（宝清站、保安站和菜嘴子站）的年径流量变化的影响率分别为25．7％、11．4％、39．9％,说明还有其他因素影响研究区的年径流量。     

Reconstructing changes in macrophyte cover in lakes across the northeastern United States based on sedimentary diatom assemblages

Macrophytes are a critical component of lake ecosystems affecting nutrient and contaminant cycling, food web structure, and lake biodiversity. The long-term (decades to centuries) dynamics of macrophyte cover are, however, poorly understood and no quantitative estimates exist for pre-industrial (pre-1850) macrophyte cover in northeastern North America. Using a 215 lake dataset, we tested if surface sediment diatom assemblages significantly differed among lakes that have sparse (<10% cover; group 1), moderate (10–40% cover; group 2) or extensive (>40% cover; group 3) macrophyte cover. Analysis of similarity indicated that the diatom assemblages of these a priori groups of macrophyte cover were significantly different from one another (i.e., difference between: groups 1 and 3, R statistic = 0.31, P < 0.001; groups 1 and 2, R statistic = 0.049, P < 0.01; groups 3 and 2, R statistic = 0.112, P < 0.001). We then developed an inference model for macrophyte cover from lakes classified as sparse or extensive cover (145 lakes) based on the surface sediment diatom assemblages, and applied this model using the top-bottom paleolimnological approach (i.e., comparison of recent sediments to pre-disturbance sediments). We used the second axis of our correspondence analysis, which significantly divided sparse and extensive macrophyte cover sites, as the independent variable in a logistic regression to predict macrophyte cover as either sparse or extensive. Cross validation, using 48 randomly chosen sites that were excluded from model development, indicated that our model accurately predicts macrophyte cover 79% of the time (r ² = 0.32, P < 0.001). When applied to the top and bottom sediment samples, our model predicted that 12.5% of natural lakes and 22.4% of reservoirs in the dataset have undergone a ≥30% change in macrophyte cover. For the sites with an inferred change in macrophyte cover, the majority of natural lakes (64.3%) increased in cover, while the majority of reservoirs (87.5%) decreased in macrophyte cover. This study demonstrates that surface sediment diatom assemblages from profundal zones differ in lakes based on their macrophyte cover and that diatoms are useful indicators for quantitatively reconstructing changes in macrophyte cover.     

Environmental factors controlling spatial variation in sediment yield in a central Andean mountain area

A large spatial variability in sediment yield was observed from small streams in the Ecuadorian Andes. The objective of this study was to analyze the environmental factors controlling these variations in sediment yield in the Paute basin, Ecuador. Sediment yield data were calculated based on sediment volumes accumulated behind checkdams for 37 small catchments. Mean annual specific sediment yield (SSY) shows a large spatial variability and ranges between 26 and 15,100 Mg km^− 2 year^− 1. Mean vegetation cover (C, fraction) in the catchment, i.e. the plant cover at or near the surface, exerts a first order control on sediment yield. The fractional vegetation cover alone explains 57% of the observed variance in ln(SSY). The negative exponential relation (SSY = a × e^{−b C}) which was found between vegetation cover and sediment yield at the catchment scale (10³–10⁹ m²), is very similar to the equations derived from splash, interrill and rill erosion experiments at the plot scale (1–10³ m²). This affirms the general character of an exponential decrease of sediment yield with increasing vegetation cover at a wide range of spatial scales, provided the distribution of cover can be considered to be essentially random. Lithology also significantly affects the sediment yield, and explains an additional 23% of the observed variance in ln(SSY). Based on these two catchment parameters, a multiple regression model was built. This empirical regression model already explains more than 75% of the total variance in the mean annual sediment yield. These results highlight the large potential of revegetation programs for controlling sediment yield. They show that a slight increase in the overall fractional vegetation cover of degraded land is likely to have a large effect on sediment production and delivery. Moreover, they point to the importance of detailed surface vegetation data for predicting and modeling sediment production rates.