| Title: | Crop and location specific agricultural drought quantification: Part III. Forecasting water stress and yield trends |
| Authors: | McDaniel, R.L., C. Munster and J. Nielsen-Gammon |
| Year: | 2017 |
| Journal: | Transactions of the ASABE |
| Volume (Issue): | 60(3) |
| Pages: | 741-752 |
| Article ID: | |
| DOI: | 10.13031/trans.11651 |
| URL (non-DOI journals): | |
| Model: | SWAT |
| Broad Application Category: | hydrologic only |
| Primary Application Category: | drought assessment |
| Secondary Application Category: | crop, forest and/or vegetation growth/yield and/or parameters |
| Watershed Description: | Upper Colorado River, which drains a portion of southeast New Mexico and west central Texas, US. |
| Calibration Summary: | |
| Validation Summary: | |
| General Comments: | |
| Abstract: | Agriculture is the largest water consumer, with 70% of global water withdrawals being used for irrigation.
Water scarcity issues are being exacerbated by drought and population increases, making efficient water resource management
in agricultural production increasingly important. The objective of this article is to evaluate the use of short-term
weather forecasts for agricultural drought prediction. A crop-specific, linear regression drought analysis technique was
used in this study. This study takes place in the upper Colorado River basin (UCRB) in west Texas. Five variables associated
with agricultural drought (precipitation, temperature, biomass production, soil moisture depletion, and transpiration) were
scaled and used to estimate cotton yields. The yield percentiles were used as a drought index. Precipitation and temperature
were forecasted with a two-week lead time using probable scenarios based on historical data. The other three variables
were estimated using the SWAT model. Forecasts were generated for each week of the growing season from 2010 through
2013. Four statistics were used to evaluate model performance, including the Nash-Sutcliffe coefficient of efficiency (NSE),
the coefficient of determination (R2
), and two error indices, the percent bias (PBIAS) and the RMSE-observations standard
deviation ratio (RSR). Comparing the variables using the forecasted weather data to those using the observed weather data
revealed that four of the five performed satisfactorily. Temperature performed the best statistically, with an NSE of 0.85 and
PBIAS of 9.4%. Precipitation (NSE = 0.51, PBIAS = -34%), cumulative biomass (NSE = 0.69, PBIAS = -38%), and transpiration
(NSE = 0.53, PBIAS = 11%) also performed well. However, the soil moisture depletion forecasts (NSE = 0.28,
PBIAS = 11%) were unsatisfactory. The forecasted cotton yield trends (NSE = 0.72, PBIAS = -12%) and drought index
(NSE = 0.76, PBIAS = -13%) both performed satisfactorily, indicating that this forecasting method may be used for decision
making related to agricultural water management, including irrigation timing. |
| Language: | English |
| Keywords: | Crop modeling, Drought, Drought index, Forecasting, Hydrologic modeling, SWAT, Water conservation, Water management, Water stress. |