Skip to main content

Effects of landscape attributes and climate variables on catchment hydrology

Abstract

Catchments characteristics, such as geomorphology, geology, soil, land use, and climatic variables, play an important role in total stream flow responses, a critical resource for people and the environment. Most of the previous literatures were applied a conventional statistical regression model to assess the relationship between landscape-climate descriptors, and streamflow and PET. However, a conventional statistical regression model didn’t consider dependence of explanatory variables that were collected or extracted across both space and time. This paper investigated the impacts of landscape attributes and climate variables on catchment scale temporal variation of total streamflow and spatio-temporal variation of potential evapotranspiration (PET) in the Mille catchment using multiple linear regression techniques, and the importance of this study was to test spatial autocorrelation in the spatial regression model which is required to properly assess and quantify the relationship between hydrological regime response components and Landscape-climate descriptors in a catchment with topographically complex, and high spatio-temporal climatic variation like in our case study area, the Mille catchment. Statistical regression analysis revealed significant relationships between streamflow and climate variables, especially with rainfall. Mean maximum temperature is the most dominant factor controlling temporal variation of potential evapotranspiration at a monthly scale, whereas NDVI is the most significant factor that controls the spatial variability of PET. The multiple regression model shows that 91.1% of temporal variation in streamflow was accounted for rainfall, whereas, 96.6% and 78.4% of temporal and spatial variation in potential evapotranspiration was accounted for in maximum temperature and NDVI, respectively. Methods also can be applied to catchments with similar landscape attributes and climate variables.

Introduction

Catchment landscape components, such as geomorphology, geology, soil, and climatic characteristics, play a significant role in total stream flow, base flow and surface runoff spatiotemporal variation (Zhu and Day 2009). Land use and land cover, which is an integral component of the landscape, critically alter the hydrological regime response descriptors like surface runoff and evaporation (Mao and Cherkauer 2009). The land use and land cover change affect the rainfall and temperature pattern, the fundamental driving forces of the hydrological cycle (Kittel 2000). Subsequently, it alters the catchments’ water balance that exists between evaporation, groundwater recharge, and stream flow (DeFries and Eshleman 2004).

Streamflow with various magnitudes is extremely disturbed by topographic attributes like elevation, drainage network, and climate variables (Mohamoud 2004; Wei and Zhang 2010), and watershed soil moisture variability (Qiu et al. 2001). Hydrological response descriptors like baseflow are also significantly affected by climate descriptors like rainfall, and landscape components like geology, relief, land use, and soil attributes (Price 2011). According to Woods et al. (1997), spatial variability of soil moisture content strongly affects catchments’ hydrology, which may be predominantly controlled by catchment topography. Evaluating the impacts of landscape-climate descriptors on catchments’ hydrological regime response is important for the management and development of catchments.

One of the major challenges with climate variables and their changes is their impact on water resources and extreme hydrological events (Andersson et al. 2011). Semiarid regions of developing countries, including Ethiopia, which faces major water resource stress, are likely to be the most severely impacted (Misra 2014; Nations 2007). Developing countries like Ethiopia, where water is a stressed resource, and their river basins of arid and semi-arid provinces, are facing extreme land use land cover changes as a result of rapidly increasing population number and socio-economic developments (Chatterjee 2018; Kassas 1976), and consequently, they drastically affect stream flow variability (Pervez and Henebry 2015). Additionally, in recent years many of the river basins in developing countries are facing extremely adverse stream flow conditions, such as floods and droughts (Garg et al. 2017).

Understanding the factors that drive the temporal fluctuation of catchments’ streamflow and potential evapotranspiration is essential for predicting how those descriptors will respond to landscape-climate predictors (Chiverton et al. 2015; Ersi et al. 2022; Hatfield and Prueger 2014; Xiao et al. 2019), and helps water resource planning and management related issues like the delivery of drinking and agricultural irrigation water, and industrial water use (Verstraeten et al. 2008). Most of the aforementioned literature used a conventional statistical regression model at the regional scale to assess and quantify the landscape-climate explanatory variables as a factor that affects hydrological regime response descriptors like stream flow and potential evapotranspiration. However, the conventional statistical regression model was not considered for the spatially referenced data that was collected or extracted across both space and time, because there is an assumption that the explanatory variables are independent, but in reality that is not (Lei Ji and A. J.Peters 2004). Instead, a spatial regression method that can correct for spatial autocorrelation in the regression model (Lei Ji and Peters 2004; Tiefelsdorf 2000) is required to properly assess and quantify the relationship between hydrological regime response components and Landscape-climate descriptors in a catchment with topographically complex, and high spatiotemporal climatic variation like in our case study area, Mille catchment.

This research aims to analyze mean monthly stream flow and potential evapotranspiration (PET) with limited landscape-climate descriptors such as mean monthly rainfall, potential dryness index (PDI), mean monthly temperature, maximum temperature, land use land cover-normalized difference vegetation index (LULC-NDVI), soil water content (SWC), baseflow index (BFI), easting and elevation (for spatial PET) using temporal and spatial regression method that adjusted for spatial autocorrelation.

This research attempts to answer the following questions regarding stream flow-potential evapotranspiration and limited landscape-climate predictors at the catchment scale: (1) which landscape-climate descriptors are likely to affect the stream flow temporal variation? (2) Which landscape-climate descriptors are the factors that drive spatiotemporal variation of potential evapotranspiration within the Mille catchment? and (3) which landscape or climate predictor variable has the dominant impact on streamflow and potential evapotranspiration? The results can help improve our understanding of how limited landscape attributes and climate variables affect the stream flow and potential evapotranspiration variability, thus assisting water resource management and planning throughout the catchment.

Materials and methods

Study area

Mille River catchment is located in the Awash river basin of Northeast Ethiopia lies between 11°26 and 11°46N latitude and 39°38 to 40°46E longitude, and covers an area of 442 km2 (Figure 1). Complex mountainous terrain and various climatic conditions, including humid to arid weather, characterize the catchment. The intra-annual spatial rainfall distribution in the study area is bimodal, which is mainly affected by the Intertropical Convergence Zone (ITCZ) (Fekadu 2015). Bati et al. (2022) showed that the combination of altitude and easting as an explanatory variable highly affects the spatial pattern of the Mille catchment’s rainfall and temperature.

Fig. 1
figure 1

Study area, Mille catchment

From Fig. 2, we can observe that the catchment’s water loss as potential evapotranspiration is on average ten folds than the mean areal rainfall. Except for July and August, the rest months from September to June the catchment were under water deficiency. The lowest PET is in December which is 116.58 mm, while the highest PET was in June, 173.60 mm.

Fig. 2
figure 2

Mean monthly streamflow (m.3/s), areal rainfall, and PET (mm)

Datasets and methods

Datasets

Daily meteorological datasets were obtained (Fig. 1) from the National Meteorological Agency (NMA), Ethiopia, from 1 January 1983 to 31 December 2002. These meteorological datasets include rainfall (R), maximum and minimum air temperature (Tmin and Tmax), relative humidity (RH), solar radiation (SR), and wind speed (Ws). According to Hargreaves and Samani (1985), a recommendation based on two main reasons such as the problem related to the quality of climatological data, and the possible errors in connection with the more sophisticated methods, Potential evapotranspiration at each meteorological station was estimated using the Hargreaves method stated as:

$$ET_{0} = 0.00023*RA* TD^{0.50} \left( {T^{0} C + 17.8} \right)$$
(1)

where \({ET}_{0}\) is potential evapotranspiration (reference crop evapotranspiration), \(RA\) is extraterrestrial radiation, TD is mean maximum minus mean minimum temperature, and \({T}^{0}C\) is mean temperature

The mean monthly areal rainfall (mm), temperature(°c), and potential evapotranspiration (mm) were estimated using kriging with external drift (KED) (Goovaerts 2000; Bati et al. 2022; Webster 2015), a geostatistical interpolation technique with elevation as an auxiliary variable, respectively. The potential dryness index (PDI) was calculated as the ratio between mean monthly areal potential evapotranspiration and mean monthly areal rainfall. Average daily stream flow at the catchment outlet (Lat: 11°25′ N and Long: 40°46′ E) from 1983 to 2002 was provided by the Department of Hydrology, Ministry of Water and Energy, Ethiopia. Average soil water content (SWC) (mm) extracted from SWAT+ model calibrated dataset (not shown here).

Monthly point and spatial areal NDVI was extracted from MOD13A1-006 MODIS/Terra Vegetation Indices having 500m spatial and 16 days temporal resolution from NASA (https://doi.org/10.5067/MODIS/MOD13A1.006), accessed 12 November 2022 using Application for Extracting and Exploring Analysis Ready Samples (AρρEEARS) software (https://appeears.earthdatacloud.nasa.gov). The data was extracted from 2000 to 2002, and then the monthly mean NDVI was averaged.

Baseflow (mm) was estimated from total stream flow records using the Base flow Digital Filter Program (https://swat.tamu.edu/software/), an automated base flow separation technique(Arnold et al. 1995; Arnold, 1999), and then baseflow index (BFI) was calculated from the historical records as the ratio of base flow to total streamflow used as a surrogate for geological attributes.

Monthly landscape-climate descriptors datasets such as mean monthly areal soil water content (SWC), NDVI, mean monthly areal rainfall, mean monthly areal temperature, potential dryness index (PDI), and hydrological regime response like mean monthly total streamflow and potential evapotranspiration were prepared for regression analysis (Table 1). In addition, Elevation and Easting were extracted from 30m SRTM DEM provided by NASA/USGS (https://earthexplorer.usgs.gov), accessed 8 June 2021.

Table 1 Statistical results for all collected data in the study

Statistical analysis and modeling technique

Statistical analysis and modeling were conducted using R programming software (https://www.R-project.org/). Regression analysis was computed to examine relationships between hydrological regime responses like stream flow and potential evapotranspiration (PET), and landscape-climate descriptors like elevation, easting, PDI, mean monthly rainfall, mean monthly temperature, LULC_NDVI, BFI and S_SWC (Average). The regression analysis procedure involved the following two steps:

  1. (a)

    A multiple linear regression analysis was performed using the response variable such as mean monthly streamflow and potential evapotranspiration, and a set of explanatory variables.

The multiple linear regression model mathematical equations are expressed as:

$$y = \beta_{0} + \beta_{1} x_{1} + \beta_{2} x_{2} + \ldots + \beta_{i} x_{i}$$
(2)

where y is the response variable. \({\beta }_{0}\) is the intercept, \({x}_{1}\) is the 1st explanatory variable, \({\beta }_{1}\) is the slope coefficient, \({\beta }_{i}\) is the slope coefficient for the ith explanatory variable, \(\varepsilon\) is the remaining unexpected noise in the data (error)

The values for β0, β1, β2…, βi were estimated using the ordinary least square method, which minimizes the sum of squared residuals (RSS):

$$RSS = { }\sum \left( {Y_{i} - \hat{Y}_{i} } \right)^{2}$$
(3)

where: yi: The actual response value for the ith observation. ŷi: The predicted response

Multiple regression models were developed and fitted with ordinary least squares (OLS) Procedures, a technique used to estimate the regression model parameters (Kroll and Song 2013).

  1. (b)

    Checking multicollinearity

One important assumption of multiple linear regressions is that no explanatory variable in the regression model is highly correlated with another explanatory variable in the same model but in reality that is not. Multicollinearity is the condition where at least one explanatory variable is closely related to one or more other explanatory variables (Helsel et al. 2020), which influences the coefficient estimates and the p-values. Such adverse consequences are magnified when sample sizes are smaller (like in our case), correlations between variables are higher, and model error variances are higher (Kroll and Song 2013). Before the selection of the most useful predictor variables, we detected multicollinearity in the model using the variance inflation factor (VIF) (Kroll et al. 2004). Predictors that have VIF values that are greater than 10 are the cause of multicollinearity, and removed one or more of the highly correlated variables (Kroll and Song 2013).

$${\mathbf{VIF}}_{{\mathbf{i}}} = \frac{1}{{\left( {1 - {\mathbf{R}}_{{\mathbf{i}}}^{2} } \right)}}$$
(4)

where \({R}_{i}^{2}\) is the R2 from a regression of the ith explanatory variable on the other explanatory variables, and the equation is used for adjustment of \({x}_{i}\) in multiple regression equation (Eq. 1).

R tool was used to do multiple linear regression analyses because it is free, powerful, and widely available (https://www.R-project.org/).

Results

Streamflow regression analysis

Table 2 shows Pearson correlation coefficients between mean monthly stream flow and limited landscape-climate explanatory variables. Stream flow is correlated positively to rainfall, SWC, NDVI, and temperature, and negatively correlated to PDI, and BFI.

Table 2 Correlation coefficients of streamflow versus landscape-climate explanatory variables

The estimation of the full multiple linear regression model (Eq. 2) is shown in Table 3. The analysis of variance (ANOVA) indicates that overall the model was performed significantly (p-value < 0.002305, Adjusted R2 = 0.9131). Also, the predictor variable, rainfall is statistically significant (p-value = 0.025) at the 0.05 significance level, while the PDI,\({T}_{mean}\), SWC, NDVI, and BFI are not. However, we can see that the variance inflation factor (VIF) value for BFI is greater than 10 (Fig. 3). This indicates that it is highly correlated with other explanatory variables in this regression, which means that multicollinearity is likely to be a problem in this regression model.

Table 3 Coefficient estimates of the full regression model for stream flow
Fig. 3
figure 3

VIF values of explanatory variables for the full regression model

In such cases, the coefficient estimates and p-values in the regression output are likely unpredictable. Thus, based on Kroll and Song (2013), we remove one variable at a time, which is the cause for multicollinearity, and repeat the process until all problem-causing explanatory variables are removed.

From Fig. 3, BFI has a VIF value greater than 10 that causes multicollinearity to have been removed from the regression. So, finally, we were left with a list of variables that have no or very weak correlation between them (Table 4).

Table 4 Coefficient estimates of the subset regression model for streamflow

From the regression model output, the adjusted R-Squared value for the model is 0.9171, the overall F-statistic is 25.35 and the corresponding p-value is 0.0005753<0.001 significance level, which shows that the overall regression model is significant with the VIF value for all predictor variables were less than 10. Nevertheless, predictor variables such as PDI, Tmean, SWC, and NDVI are still showing less significance on the response variable, stream flow, whereas rainfall is highly significant at α = 0.001 level of significance. The regression model equation using the final regression model output is written as:

$$Q = - 0.52 + 0.05R + - 0.25{\text{T}}_{{{\text{mean}}}} + 0.18PDI + 0.0{\text{8SWC }} + {2}.{\text{81NDVI}}$$
(5)

where Q is streamflow (mm)

In multiple regression Model 5, a per unit increase of mean rainfall depth, potential dryness index, soil water content and normalized difference vegetation index produce an increase in percentage stream flow values of 0.05, 0.18, 0.08, and 2.81, respectively, whilst per unit decreases of stream flow value attributed to mean temperature is 0.25%.

Potential evapotranspiration

Regression model for temporal variation

The relationship between mean monthly potential evapotranspiration and predictor variables in the Mille catchment has been analyzed by developing the correlation coefficients. Table 5 showed that the correlation coefficient of mean monthly PET is positively correlated to rainfall, SWC, and mean Tmean, negatively correlated to PDI and BFI, and almost no relationship between NDVI.

Table 5 Correlation coefficients of the PET and predictor variables

Table 6 indicated that the VIF value for PET of the temporal regression model was like that of the stream flow regression model (see Fig. 3). Overall, the full temporal regression model for PET estimated with the OLS procedure was significant at F-statistics = 43.27, Adjusted R-squared = 0.9584 and p-value = 0.0003762 < 0.001 significance level, respectively.

Table 6 Statistic results of the full regression model for mean monthly PET

From the regression model output last column, the VIF value for the BFI predictor variable is greater than 10, and this shows that the predictor variable is likely suffering from a multicollinearity problem, which affects the regression coefficient and the p-value of the predictor.

By removing the predictor in the model that has a high VIF value from the dataset (Kroll and Song 2013) and by doing so, it can solve multicollinearity (Table 7).

$${\text{PET}} = { } - 49.63 + { }7.74{\text{Temp}} + { } - 0.051{\text{R}} + 0.26{\text{SWC}} - 18.35{\text{NDVI }} - 0.196{\text{PDI}}$$
(6)
Table 7 Statistical estimates of the subset regression model for the temporal regression model

Overall performance of the subset regression model was highly significant (F-statistic = 60.95, p= 4.62E-05< 0.001, Adj_R2 = 0.9646), specifically the mean monthly temperature is statistically highly significant at the 0.001 significance level, and the rainfall is statistically less significant at 0.1 significance level while SWC, NDVI, and PDI are not statistically significant predictors. The final regression model is written as:

Spatial multiple regression model

To improve a better understanding of the correlation between the response variable (in our case PET) and spatial explanatory variables, we created a correlation matrix to view the linear correlation coefficients between each pair of variables (Table 8). From the correlation matrix, we can conclude that PET is strongly positively correlated with easting and temperature, and negatively correlated with rainfall, elevation, and land cover (NDVI). Each predictor variable exhibits a high correlation with the other, which shows there is a multicollinearity problem.

Table 8 Correlation matrix of PET versus landscape-climate predictor variables

Table 9 shows that all predictors have a VIF value greater than 10, and all predictors are statistically less significant at the 0.05 significance level. So, we removed a predictor that has a high VIF value sequentially until the predictor/s left with a VIF value of less than 10 (Kroll and Song 2013).

Table 9 Statistical results for full regression of the relationship between spatial PET and landscape-climate predictors for the mean annual. The climatic variables included in the analysis were mean elevation, mean annual rainfall, easting, NDVI, and maximum temperature

For instance, if we look at the coefficient estimated for the temperature, the model is telling us that for each additional one unit increase in maximum temperature, the average increase in PET is -5.84E+01 mm, assuming rainfall, mean Elevation, easting, and NDVI are held constant. This doesn’t seem to make sense, considering we would expect the mean temperature with high magnitude and thus have a higher PET. The reason behind this effect was a multicollinearity that affects the estimates of coefficients (Helsel et al. 2020). Following the same procedure applied for stream flow and temporal PET, we removed predictors in the model having a high VIF value step by step from the datasets (Kroll and Song 2013), and lastly, we solved an issue of multicollinearity (Table 10). Accordingly, land use land cover NDVI is statistically highly significant (F-statistic = 21.78, Adj.R2 = 0.75, p-value = 0.003443 < 0.01 significance level). The final spatial regression model is written as:

$$PET_{spatial} = 2508.2 - 1986.3{\text{NDVI}}$$
(7)

where \({PET}_{spatial}\) is the spatial potential evapotranspiration

Table 10 Statistical estimates of the subset regression model for the linear regression model

Discussion

Landscape-climate predictors’ effect

Explaining the spatiotemporal variability of catchment hydrological regime response in connection with catchment landscape attributes and climate variables is a major concern of catchment hydrology. In this study, further analysis and discussion on the results of these multiple linear regression models are summarized below.

Streamflow temporal variation

In this paper, a multiple regression equation, for long year-based mean monthly stream flow for the Mille River Catchment was constructed using multiple linear regression techniques. This regression model expressed the quantitative relationship between mean monthly stream flow, and landscape-climate predictors that were used to quantify the impacts of limited landscape-climate predictors on stream flow temporal variation at catchment scale (see Table 4). The multiple linear regression analysis of mean monthly stream flow with limited landscape attributes and climate variables for mean monthly stream flow prediction (Fig. 4) was analyzed, and our regression analyses demonstrated that the regression model was overall statistically significant (\({\mathrm{F}}_{\mathrm{statistics}}=25.35,\mathrm{ Adj}. {\mathrm{R}}^{2}\) = 0.9171, P-value = 0.0005753 < 0.001 significance level). However, among all the explanatory variables, only rainfall is strongly significant for explaining temporal variation in stream flow.

Fig. 4
figure 4

Mean monthly streamflow chart a, comparison of observed and estimated mean monthly streamflow b

The regression model showed that mean monthly rainfall is the most important contributor to stream flow temporal variation, while mean temperature, PDI, SWC, NDVI, and BFI are relatively less important contributors. As shown in Figure 5, the mean monthly streamflow increased with increasing observed mean monthly rainfall, with a coefficient of determination of 0.911. The analysis showed that the temporal variation of the response variable, mean stream flow was significantly influenced by rainfall temporal variation. This could be because of the orographic effect of the catchment on mountainous terrain on spatial and temporal variation of rainfall (Goovaerts 2000; Bati et al. 2022).

Fig. 5
figure 5

Relationship between mean monthly estimated streamflow (Q) with mean monthly rainfall

PET temporal variation

Temporal variation of potential evapotranspiration in the topographic complex catchment depends on a few dominant landscape-climate descriptors (Hatfield and Prueger 2014; Zhang et al. 2010). In this paper, we combined a correlation and regression analysis to identify the dominant influence factors on potential evapotranspiration temporal variation. Specifically, we developed a regression model to analyze the influence of mean monthly areal rainfall, temperature, PDI, SWC, NDVI, and BFI on mean monthly PET. The regression model depicts that the mean monthly temperature is the most significant contributor to potential evapotranspiration deviation (p< 0.001) (see Table 7).

From Fig. 6, 96.6% of temporal variation in PET accounted for the mean monthly maximum temperature. However, no other predictors like rainfall, SWC, PDI, and NDVI were statistically significant for the regression model. The reason lack of significance may be because of the dominance of aridity to semi-aridity regime across the study area, Mille catchment. Figure 7a, b represent the observed and estimated mean monthly PET chart and scatterplot that shows the estimated and observed PET relationship, respectively. Thus, variation among months for potential evapotranspiration in the Mille catchment depends on the mean monthly air temperature across the regime, and this study agrees with conclusions drawn by other studies (Ye et al. 2014; H. Zhang and Wang 2021).

Fig. 6
figure 6

Plot of the true linear relation between the estimated PET and the observed mean monthly temperature, and 12 estimations of the PET for mean monthly maximum temperature values from 1 through 12

Fig. 7
figure 7

Estimated and observed monthly time series of the PET a, and Comparisons between PET predicted by Landscape-climate predictors and that derived from the measured temperature time series b

PET spatial variation

Spatial variation of potential evapotranspiration within the Mille catchment is strongly significant based on the results shown in Fig. 8a, b. We observed that spatial variation of potential evapotranspiration across the regime significantly increases from the uppermost head to the lower part of the landscape in the Mille catchment. We found that the spatial pattern of mean annual PET varied from 900 mm in the western study area to 2200 mm in the eastern part of the catchment. To investigate the potential drivers of the spatial PET variation, two meteorological variables (rainfall and maximum temperature), two topographic attributes (elevation and easting), and one land use land cover (NDVI) attribute were analyzed using OLS regression analysis.

Fig. 8
figure 8

Spatial patterns of observed mean annual PET in Mille catchment during 1983–2002 a and scatterplot between the observed PET and the land use land cover mean annual NDVI at the catchment scale b. As depicted in Fig. 8a, the mean annual PET is elevated from the head of the catchment to the lower part of the region

As a result, the effect of the mean annual NDVI on the spatial distribution of PET is highly significant (see Table 10).

The location of the catchment lay in the core of the Great Rift Valley, which may be the main contributor to high diversity in the Agro ecological zone in the region, which resulted in high PET (Resources, 2009). Thus, we conclude that mean annual NDVI is the main factor influencing the spatial distribution of PET in the Mille catchment.

We developed a variogram model to show the occurrence of spatial autocorrelation of the regression residuals with a sample variogram (Fig. 9). The mean annual PET residuals experimental variogram was fitted with theoretical variograms using a spherical variogram model with a range of 13.99 km, a sill of 17601 mm2, and a nugget effect of 14447 mm2.

Fig. 9
figure 9

Empirical variogram fitted with theoretical variogram using spherical variogram model for mean annual PET regression residuals

From Fig. 9, the residuals in regression analysis between any pair of weather stations within a distance range are spatially correlated to each other which violate the assumptions of linear regression, which states that there is no spatial relationship between consecutive residuals in time series data.

Conclusions

This research applied multiple linear regression techniques to explore the relationship between the catchment’s landscape-climate predictors and hydrological regime response spatiotemporal variation, specifically stream flow and potential evapotranspiration at monthly and annual time scales. The results show that the dominant climate predictor on mean monthly stream flow temporal variation was the mean monthly areal rainfall that accounts for 95.5% of the stream flow variation. Other studies using multiple regression models have also demonstrated that the precipitation is the main climatic factor for streamflow temporal variation (e.g. Ali et al. 2021), which supports our findings. The complex mountainous area of the upper catchment resulted in high spatial areal rainfall (Bati et al. 2022), and this may possibly be a source for a significant relationship with lowland gauged stream flow (Viviroli et al. 2007), in our case the Mille River.

A comprehensive assessment of the effects of landscape attributes and climate variables on the spatio-temporal variation of potential evapotranspiration was also presented via regression analysis that provides the basis for the identification of dominant factors which affect the spatio-temporal variation of the response variable. In temporal regression analysis, we found that among limited landscape-climate descriptors, only long year-based mean monthly temperature has a significant influence on potential evapotranspiration temporal variation showing that the climate component is a key for hydrological cycle temporal alteration, and these findings are like that reported by other authors (e.g. Neubert and Rannow 2014). An increase in mean monthly temperature increases the potential evapotranspiration which means that it enhances the moisture-holding capacity of the atmosphere thus temperature is the key change in the hydrological system, specifically alteration in the temporal variation of potential evapotranspiration at a monthly scale whereas, spatial variability in catchment’s land cover depicted that the spatial correlation between potential evapotranspiration and land use land cover (NDVI) component is highly significant.

In multiple linear regressions, there is the possibility that some of the independent variables are correlated with one another, so it is important to check these before developing the regression model. In this paper, multicollinearity in the model was detected by using the variance inflation factor (VIF) so that one or more of the highly correlated predictors that have VIF values which are greater than 10 were removed.

Overall, understanding factors causing spatio-temporal variation in hydrological regime response like total streamflow and potential evapotranspiration will lead to improving capabilities for water management in various water use systems, and the results provide water resource planners at various sectors a sign of which catchment landscape-climate descriptors may aid in describing the hydrological regime response spatiotemporal variation. This information should aid in the modeling and management of total stream flow and potential evapotranspiration. Establishment of more flow gauging stations at different reaches of the river and with more distributed meteorological stations, a much better understanding of the spatial and temporal variability of the hydrology of the catchment could be achieved. However, with the existing limited data, the research demonstrated the dominant role of rainfall, temperature, and NDVI on hydrological system response specifically on the total stream flow and potential evapotranspiration of the catchment.

Availability of data and materials

The datasets used and/or analyzed during the current study are available from the corresponding author upon reasonable request.

Abbreviations

PET:

Potential evapotranspiration (mm)

NDVI:

Normalized difference vegetation Index

BFI:

Baseflow index

SWC:

Soil water content (mm), streamflow (mm), rainfall (mm)

Tmean:

Long year-based mean annual temperature (°c)

VIF:

Variance inflation factor

\(\mathrm{Q}\) :

Total stream flow (mm)

OLS:

Ordinary list square

SWAT + :

Soil water assessment tool plus

References

Download references

Acknowledgements

The observed climate and stream flow datasets used for further applications were obtained from the National Metrology Institute of Ethiopia (NMIE), and the Ministry of Water and Energy (MoWE), respectively. We greatly appreciate the Awash Basin Administrative office, Melka Worer, Ethiopia for offering the vehicle for fieldwork.

Funding

The research was financed through the Ph.D. scholarship (Grant No. ACEWM/GSR/4334/11) awarded to the first author by the Africa Center of Excellence for Water Management, Addis Ababa University, Ethiopia.

Author information

Authors and Affiliations

Authors

Contributions

HG conceptualization of the research, develop the methodology, and field data collected and analyzed, and wrote the original draft and final manuscript. TA, TA, and MD substantively edited the manuscript. All authors read and approved the final manuscript.

Corresponding author

Correspondence to Hirpo Gudeta Bati.

Ethics declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors have declared that no competing interests exist.

Additional information

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Bati, H.G., Agumassie, T.A., Tegaye, T.A. et al. Effects of landscape attributes and climate variables on catchment hydrology. Environ Syst Res 12, 9 (2023). https://doi.org/10.1186/s40068-023-00290-y

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: https://doi.org/10.1186/s40068-023-00290-y

Keywords