Streamflow drought interpreted using SWAT model simulations of past and future hydrologic scenarios: application to Neches and Trinity River Basins, Texas

Authors: Sohoulande, Clement

Submitted to: Journal Hydrologic Engineering
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 4/17/2019
Publication Date: 6/27/2019
Citation: Sohoulande Djebou, D.C. 2019. Streamflow drought interpreted using SWAT model simulations of past and future hydrologic scenarios: application to Neches and Trinity River Basins, Texas. Journal Hydrologic Engineering. 24(9). https://doi.org/10.1061/(ASCE)HE.1943-5584.0001827.
DOI: https://doi.org/10.1061/(ASCE)HE.1943-5584.0001827

Interpretive Summary: In water resources management, hydrologic indices are useful as decision support tools because they can be easily interpreted. This is true with streamflow drought index (SDI) which can be used to assess the availability of water resources at the watershed level. However, they are not enough studies on how to realistically project SDI for future periods even though a such projection could help to better plan water resources use. In response to this gap, this study uses a process-based watershed modeling approach to propose a framework for SDI projection. Especially, the soil water assessment tool (SWAT) model is employed to simulate distinctly two watersheds located in the State of Texas (i.e. the Trinity and the Neches river basins). The SWAT model was successfully calibrated and validated with monthly streamflow data (calibration period 1990 to 1995, validation period 1996 to 2015). The calibrated model was considered to simulate runoff for the future period 2041 to 2070 using future climate scenario inputs. Probability analysis was conducted on streamflow data and SDI values were computed for the past period (1996 to 2015) and the future period (2041 to 2070). During the period 1996 to 2015, the SDI values recovered from the SWAT simulations, match closely with the ones derived directly from the observed discharge data. This result portends the capacity of the analytical procedure to capture and project realistically SDI signals. However, a comparison of SDI patterns between the past and future periods does not reveal any significant difference for the studied watersheds.

Technical Abstract: In water resources and environmental management, hydrologic indices are often valued as decision support tools because of their practical interpretability. This is true with streamflow drought index (SDI) which is considered as a relevant tool for assessing the availability of water resources at the watershed level. Hence, the future of freshwater resources at the watershed scale could be better understood by achieving a realistic projection of SDI. This study uses a process-based watershed modeling approach then describes a framework for SDI projection. Especially, the soil water assessment tool (SWAT) model is employed to simulate distinctly two watersheds located in the State of Texas (i.e. the Trinity and the Neches river basins). The SWAT model is calibrated with monthly streamflow of the period 1990 to 1995. The model is later validated with two decades discharge data (i.e. period 1996 to 2015). The evaluation of the SWAT performance during the calibration and validation stages, reveals acceptable values of efficiency criteria for both watersheds (i.e. Nash-Sutcliffe's Efficiency ranging from 0.56 to 0.65; Index of agreement from 0.79 to 0.92). Thus, the calibrated model is considered to simulate runoff for the future period 2041 to 2070 using inputs retrieved from a future climate scenario. However, the SDI calculation requires the knowledge of the probability distribution of cumulative discharge data. A Kolmogorov-Smirnov’s goodness-of-fit analysis was conducted for both observed and simulated cumulative discharges. A lognormal distribution is henceforth considered for estimating time series of SDI. During the period 1996 to 2015, the SDI values recovered from the SWAT simulations, match closely with the ones derived directly from the observed discharge data (0.52=R2=0.91 for Neches river, 0.79= R2 = 0.89 for Trinity river). This result portends the capacity of the analytical procedure to capture and project realistically SDI signals. However, a statistic’s analysis of the SDI patterns during the past and the future periods does not reveal any significant difference.