UUnniivveerrssiittyy ooff NNeebbrraasskkaa -- LLiinnccoollnn DDiiggiittaallCCoommmmoonnss@@UUnniivveerrssiittyy ooff NNeebbrraasskkaa -- LLiinnccoollnn U.S. Department of Agriculture: Agricultural Publications from USDA-ARS / UNL Faculty Research Service, Lincoln, Nebraska 2015 IIDDEENNTTIIFFYYIINNGG AANNDD EEVVAALLUUAATTIINNGG AA SSUUIITTAABBLLEE IINNDDEEXX FFOORR AAGGRRIICCUULLTTUURRAALL DDRROOUUGGHHTT MMOONNIITTOORRIINNGG IINN TTHHEE TTEEXXAASS HHIIGGHH PPLLAAIINNSS Jerry E. Moorhead Department of Agriculture, West Texas A&M University, [email protected] Prasanna H. Gowda [email protected] Vijay P. Singh Conservation & Production Research Laboratory, USDA-ARS Dana O. Porter Texas A&M University Thomas H. Marek Department of Biological and Agricultural Engineering, Texas A&M AgriLife Research and Extension Service Follow this and additional works at: https://digitalcommons.unl.edu/usdaarsfacpub See next page for additional authors Part of the Environmental Indicators and Impact Assessment Commons, and the Environmental Monitoring Commons Moorhead, Jerry E.; Gowda, Prasanna H.; Singh, Vijay P.; Porter, Dana O.; Marek, Thomas H.; Howell, Terry A.; and Stewart, B A., "IDENTIFYING AND EVALUATING A SUITABLE INDEX FOR AGRICULTURAL DROUGHT MONITORING IN THE TEXAS HIGH PLAINS" (2015). Publications from USDA-ARS / UNL Faculty. 1506. https://digitalcommons.unl.edu/usdaarsfacpub/1506 This Article is brought to you for free and open access by the U.S. Department of Agriculture: Agricultural Research Service, Lincoln, Nebraska at DigitalCommons@University of Nebraska - Lincoln. It has been accepted for inclusion in Publications from USDA-ARS / UNL Faculty by an authorized administrator of DigitalCommons@University of Nebraska - Lincoln. AAuutthhoorrss Jerry E. Moorhead, Prasanna H. Gowda, Vijay P. Singh, Dana O. Porter, Thomas H. Marek, Terry A. Howell, and B A. Stewart This article is available at DigitalCommons@University of Nebraska - Lincoln: https://digitalcommons.unl.edu/ usdaarsfacpub/1506 JOURNAL OF THE AMERICAN WATER RESOURCES ASSOCIATION AMERICANWATERRESOURCESASSOCIATION IDENTIFYING AND EVALUATING A SUITABLE INDEX FOR AGRICULTURAL DROUGHT MONITORING IN THE TEXAS HIGH PLAINS1 Jerry E. Moorhead, Prasanna H. Gowda, Vijay P. Singh, Dana O. Porter, Thomas H. Marek, Terry A. Howell, and B.A. Stewart2 ABSTRACT: Drought is a complex and highly destructive natural phenomenon that affects portions of the United States almost every year, and severe water deficiencies can often become catastrophic for agricultural production. Evapotranspiration (ET) by crops is an important component in the agricultural water budget; thus, it is advantageous to include ET in agricultural drought monitoring. The main objectives of this study were to (1) conduct a literature review of drought indices with a focus to identify a simple but simultaneously adequate drought index for monitoring agricultural drought in a semiarid region and (2) using the identified drought index method, develop and evaluate time series of that drought index for the Texas High Plains. Based on the literature review, the Standardized Precipitation-Evapotranspiration Index (SPEI) was found to satisfy identi- fied constraints for assessing agricultural drought. However, the SPEI was revised by replacing reference ET with potential crop ET to better represent actual water demand. Data from the Texas High Plains Evapotranspi- ration network was used to calculate SPEIs for the major irrigated crops. Trends and magnitudes of crop- specific, time-series SPEIs followed crop water demand patterns for summer crops. Such an observation suggests that a modified SPEI is an appropriate index to monitor agricultural drought for summer crops, but it was found to not account for soil water stored during the summer fallow period for winter wheat. (KEY TERMS: evapotranspiration; Standardized Precipitation Index; Standardized Precipitation-Evapotranspi- ration Index; semiarid regions.) Moorhead, Jerry E., Prasanna H. Gowda, Vijay P. Singh, Dana O. Porter, Thomas H. Marek, Terry A. Howell, and B.A. Stewart, 2015. Identifying and Evaluating a Suitable Index for Agricultural Drought Monitoring in the Texas High Plains. Journal of the American Water Resources Association (JAWRA) 1-14. DOI: 10.1111/jawr. 12275 INTRODUCTION agricultural regions, losses can come in the form of reduced crop yield and reduced forage and even crop failure or livestock death. Crop losses can be some- Drought is a complex natural phenomenon that what mitigated with timely irrigation. When drought can cause devastating losses across large regions. In conditions occur, precipitation is typically scarce and 1Paper No. JAWRA-14-0113-P of the Journal of the American Water Resources Association (JAWRA). Received May 8, 2014; accepted October29,2014.©2015AmericanWaterResourcesAssociation.Discussionsareopenuntilsixmonthsfromprintpublication. 2Biological Science Technician (Moorhead), Research Agricultural Engineer (Gowda), and Retired Laboratory Director and Research Leader (Agricultural Engineer) (Howell), Conservation & Production Research Laboratory, USDA-ARS, PO Drawer 10, Bushland, Texas 79012; Distinguished Chair and Distinguished Professor (Singh), Department of Biological and Agricultural Engineering, Texas A&M Uni- versity,CollegeStation,Texas77843;AssociateProfessorandExtensionAgriculturalEngineeringSpecialist(Porter),DepartmentofBiologi- cal and Agricultural Engineering, Texas A&M AgriLife Research and Extension Service, Lubbock, Texas 79403; Research Agricultural Engineer (Marek), Department of Biological and Agricultural Engineering, Texas A&M AgriLife Research, Amarillo, Texas 79106; Director, Dryland Agriculture Institute and Distinguished Professor (Stewart), Department of Agriculture, West Texas A&M University, Canyon, Texas79016(E-Mail/Moorhead:[email protected]). JOURNALOFTHE AMERICAN WATERRESOURCESASSOCIATION 1 JAWRA MOORHEAD, GOWDA, SINGH, PORTER, MAREK, HOWELL, AND STEWART more water evaporates to the environment, which in more consecutive days with less than 6.35 mm turn demands relatively more irrigation to meet the (0.25 in.) of precipitation in 24 h. Marcovitch (1930) crop water needs, and the amount of irrigation developed an index based on the climatic require- applied is increased (if possible) to sustain produc- ments for the bean beetle in the eastern United tion. Increased irrigation application, however, places States (U.S.). This index incorporated both tempera- greater demand on available water resources that are ture and precipitation using the total number of two likely to be subject to additional demands from the or more consecutive days above 32°C (90°F) and the drought conditions while simultaneously increased total summer rainfall for the same period. Blumen- irrigation costs reduce net profits. stock (1942) used probability theory to conduct a Agricultural drought can be best monitored using a climate study on drought frequencies, and Blumen- droughtindexthatsimultaneouslyusesbothprecipita- stock used the length of drought in days where a tion and evapotranspiration (ET), two major compo- drought was terminated by at least 2.54 mm nents of the water budget. ET represents the loss of (0.10 in.) of precipitation in 48 h or less. McGuire waterfrom thesoil throughsoil evaporation and plant and Palmer (1957) developed an index that used transpiration, whereas precipitation represents addi- potential ET (PET) termed as the Moisture Adequacy tion of water to the soil. The difference between the Index (MAI). This index compared the water need to two components represents irrigation demand in arid the rainfall and stored soil water for a given location. and semiarid regions where inadequate precipitation The MAI is expressed as a percentage of the actual is expected during the growing season and is supple- water supply compared to the water need. mentedbyirrigation.Therefore,thesetwocomponents After extreme droughts occurred over very large are essential parts of monitoring agricultural drought regions of the U.S. during the 1930s and again in the conditions and for determining irrigation require- 1950s, more drought research was conducted and new ments. The main objectives of this study were to (1) droughtindicesweredevelopedinanattempttobetter conduct a detailed review of drought indices available quantify and study drought. Both instances affected intheliteraturetodetermineasuitabledroughtindex much of the Texas High Plains. Interest in drought for monitoring agricultural drought and (2) calculate research has further increased because of the ongoing and evaluate time series of the selected drought index extreme drought in the region (Baumhardt et al., fortheTexasHighPlains. 2014).Alistofcommondrought indices thatarefound intheliteratureispresentedinTable 1. LITERATURE REVIEW Palmer Drought Severity Index The Palmer Drought Severity Index (PDSI) Numerous drought indices have been developed (Palmer, 1965) was derived to provide a methodology over time and are variously applied throughout the for calculating an index for evaluating the meteoro- literature (Heim, 2002; Mishra and Singh, 2010; logical anomaly characterized by a prolonged and Sivakumar et al., 2011). Munger (1916) reported that abnormal water deficiency for a variety of time scales. the intensity of droughts were more harmful than Palmer’s objective was to use a specific definition of their length. Therefore, Munger used the number of drought and create a measurement technique that consecutive days where 24-h rainfall was less than would allow for comparison of drought events. Palmer 1.27 mm (0.05 in.) to develop a drought index for noticed that a variety of drought definitions existed, studying fire hazard in the Pacific Northwest. and to define what drought means precisely for all Munger (1916) then developed a graphical technique invested stakeholders would be extremely difficult. to represent the intensity of the drought using the Palmer concluded that all drought indices dealt with area of a right triangle whose height and base were variations in water deficiency and generalized the proportional to the duration of drought. The equation drought definition to “a prolonged and abnormal for Munger’s index was the square of the length of moisture deficiency.” Use of a generalized drought the drought in days divided by two. definition allowed Palmer to derive an index that Kincer (1919) prepared a series of maps and charts could be used with any definition of drought. showing seasonal distribution of precipitation and Palmer (1965) attempted to develop a procedure to distribution of the average annual number of days compute the amount of precipitation that would be with precipitation of various intensities. Kincer’s considered climatically normal for a given area. One work was the first work of this type that had been problem with using derived water deficiencies and performed. Kincer used a drought definition of 30 or excesses for a number of time periods, however, is JAWRA 2 JOURNALOFTHE AMERICAN WATERRESOURCESASSOCIATION IDENTIFYING AND EVALUATING A SUITABLE INDEX FOR AGRICULTURAL DROUGHT MONITORING IN THE TEXAS HIGH PLAINS TABLE1. CommonlyUsedDroughtIndices. Index Classification InputParameters PalmerDroughtSeverityIndex(PDSI—Palmer,1965) Meteorological Soilcharacteristics, referenceET,precipitation CropMoistureIndex(CMI—Palmer,1968) Agricultural SameasPDSI SurfaceWaterSupplyIndex(SWSI—ShaferandDezman,1982) Hydrological Soilwater,streamflow PalmerHydrologicalDroughtIndex(PHDI—Karl,1986) Hydrological SameasPDSI StandardizedPrecipitationIndex(SPI—McKeeetal.,1993) Meteorological Precipitation ReclamationDroughtIndex(Hayes,2006) Hydrological Temperature,precipitation,streamflow VegetationConditionIndex(VCI—Kogan,1995) Agricultural Satelliteimages U.S.DroughtMonitor(NDMC,2013) Meteorological Precipitation,variousdroughtindices, groundwaterlevel,reservoirstorage StandardizedVegetationIndex(SVI—Petersetal.,2002) Agricultural Remotesensingdata ReconnaissanceDroughtIndex(RDI—TsakirisandVangelis,2005) Agricultural ActualET,PET SoilMoistureDeficitIndex(SMDI—NarasimhanandSrinivasan,2005) Agricultural Soilwater EvapotranspirationDeficitIndex(ETDI—Narasimhan Agricultural ActualET,PET andSrinivasan,2005) StandardizedPrecipitation-EvapotranspirationIndex Meteorological Precipitation,referenceET (SPEI—Vicente-Serranoetal.,2010) AccumulatedDroughtIndex(ADI—CIIAGRO,2012) Agricultural Precipitation,referenceET RelativeWaterDeficit(RWD—Sivakumaretal.,2011) Agricultural ActualET,PET Note: PET,potentialevapotranspiration;ET,evapotranspiration. that it does not take drought duration into account. of water capacity in the surface layer is minor com- In addition, departures from normal vary from one pared to the underlying layer, which results in water location to another. To overcome these issues, a balance methods being insensitive to the inclusion of weighting scheme was developed to transform the the surface layer. Furthermore, the runoff estimate departures in accordance with their significance to in Palmer’s method can be a source of deficiency in the climate of the location being considered. In the the computations. Palmer uses a “threshold-type” end, all these needed to be combined into an index of model in assuming runoff which assumes that runoff abnormality for an extended duration of drought, and does not occur until the available water capacity of systematic procedures derived for delineating the both layers of the soil is filled (Alley, 1984) — this abnormal periods. As a result, Palmer (1965) com- may not always be a valid assumption. puted water excess or deficiency using a water Karl (1986) studied the sensitivity of PDSI to its balance or hydrologic accounting approach. Such an assumptions and parameters. Because of the com- approach considers water inputs and losses to pro- plexity of the water budget system, relatively few duce a current state of water at site or regional studies have investigated this aspect. Karl observed scales. The primary water loss in Palmer’s approach that the Z-index, calculated in the PDSI procedure as is PET. In Palmer’s process, PET is calculated using a single point indicator of drought or wetness, was Thornthwaite’s formula (Thornthwaite, 1948). The likely to be more adequate for agricultural water calculation methodology for the PDSI is available in shortages than would the PDSI. The observation was Palmer (1965). attributed to the fact that the Z-index is more respon- The PDSI provides the capability for comparisons sive to short-term soil water changes than PDSI. The across regions and time scales; however, limitations PDSI is more suited as a meteorological index with a to PDSI exist. In the hydrological accounting proce- longer time scale. The Z-index is calculated before dure, Palmer assumes that ET takes place at the duration is taken into account, which allows for a potential rate when precipitation is greater than shorter time scale. The Z-index still behaves as a PET. Because of the nature of the occurrence of pre- meteorological drought index where the first month cipitation, this may not be true for all of the time the weather goes from dry to normal, or vice versa, periods used (Alley, 1984). In some instances, sub- the drought or wet spell ends even though the soil stantial precipitation may fall in the beginning or water may still be above or below their normal levels. end of the time period. This inherently leads to either over- or underestimation of ET. Alley (1984) noted that in Palmer’s two-layer soil Crop Moisture Index profile, the water capacity of the surface layer of the soil is 25 mm, which can physically and numerically The Crop Moisture Index (CMI) was developed by go from full to empty in a single month. The 25 mm Palmer (1968) to provide a drought monitoring tool JOURNALOFTHE AMERICAN WATERRESOURCESASSOCIATION 3 JAWRA MOORHEAD, GOWDA, SINGH, PORTER, MAREK, HOWELL, AND STEWART for agricultural regions. Palmer noted the difficulty Standardized Precipitation Index in using a meteorological approach to monitor agri- cultural drought because of variations in soils, crop The Standardized Precipitation Index (SPI) was types, and precipitation amounts. However, Palmer developed by McKee et al. (1993) in response to defi- (1968) did see the need for a broad scale general ciencies seen in the PDSI. They described the need drought monitoring method, such as monitoring the for a probability based drought monitoring method crop water situation in a soybean producing region. and noted that the use of the PDSI is limited, in part Using the output from the PDSI, Palmer combined because of its undefined yet inherently built-in time the weekly ET deficit into an index of the ET anom- scale. Those authors also pointed out the differences aly, which is a measure of cumulative agricultural in the various drought definitions, as outlined by drought. To account for wet weather, where excessive Dracup et al. (1980) and Wilhite and Glantz (1985) water can damage crops, Palmer combined the and the effect these differences had on drought moni- recharge and runoff components of the PDSI into a toring. It was noted that all drought definitions wetness index. The ET index is usually negative classify drought as a condition of insufficient water while the wetness index is positive. The final CMI is because of a lack of precipitation. To provide an index the sum of these two values. The CMI is near zero and drought definition, the SPI was created. The SPI under normal conditions, positive for wet conditions, defines drought based on standardized precipitation. and negative for dry conditions. The CMI provides a It is simple to calculate and can be used for any sin- shorter time scale than the PDSI, which is more gle parameter and time scale desired. To calculate the suited for agricultural drought. The CMI has the SPI, the precipitation data must first be fitted to same limitations and complexity as the PDSI. a probability distribution which best fits or ade- quately mimics the data. For example, the normal distribution is not always appropriate. The step of Surface Water Supply Index distribution choice and parameter estimation is often overlooked by some investigators (Guttman, 1999). The Surface Water Supply Index (SWSI) was devel- After a distribution is fit to the data, the SPI is calcu- oped by Shafer and Dezman (1982) to improve limita- lated by taking the difference of precipitation from tions from the PDSI. The SWSI is an indicator of the mean for a time period divided by the standard hydrological drought and is based on nonexceedance deviation. probabilities of reservoir storage, stream flow, snow- The SPI requires only one input parameter so it pack, and precipitation (Mishra and Singh, 2010). The can be used for precipitation as well as snowpack, SWSI is designed to monitor surface water supply stream flow, reservoir storage, soil water, or ground- sources, which limits its usefulness for agricultural water. This feature allows it to be used for any droughtmonitoring. drought definition. In addition, the SPIs that are based on different parameters can be compared and contrasted. Mishra and Singh (2010) pointed out that Palmer Hydrological Drought Index the greatest strength of the SPI is its ability to be used on multiple time scales. This ability makes the The Palmer Hydrological Drought Index (PHDI) SPI useful in monitoring drought according to multi- refers to the PDSI when calculated using real time ple definitions. A disadvantage of the SPI is that data (Karl, 1986). The main difference between the using precipitation (or any single parameter) alone PDSI and the PHDI is the treatment of the beginning does not describe all factors that influence drought. and ending times of droughts or wet periods. The For example, a period with average precipitation beginning and ending of a drought is determined by should indicate that there are no drought conditions theratioofmoisturereceived tothe moisturerequired present, even though the evaporative demands could todefinitivelyendadroughtandisexpressedasaper- be well above average, creating a water shortage. centage(termedP ).ForthePDSI,adroughtisconsid- Also, the SPI is sensitive to the length of the pre- e eredtohaveendedwhenP becomesgreaterthanzero, cipitation record for the area of interest where the e and continues to be greater than zero until reaching SPI values may differ when computed from different 100%. In real time, it cannot be known if a period of lengths of record. The SPI values can exhibit similar time with P greater than zero demarks a temporary results when different precipitation records have e interruption of the current drought (Karl, 1986). For similar gamma distributions (Mishra and Singh, thePHDI,thedroughtisnotconsideredtohaveended 2010). Because of the nature of the variability of pre- until P reaches 100%. Thus, the PDSI and PHDI are cipitation, it can be difficult to fit the precipitation e different only when P is greater than zero and less data to conventional probability distributions. Several e than100%. different distributions, such as normal, gamma, and JAWRA 4 JOURNALOFTHE AMERICAN WATERRESOURCESASSOCIATION IDENTIFYING AND EVALUATING A SUITABLE INDEX FOR AGRICULTURAL DROUGHT MONITORING IN THE TEXAS HIGH PLAINS Pearson III, are widely used with each distribution data. The SVI is similar to the VCI and has similar resulting in differing nuances to SPI interpretation. limitations for agricultural drought monitoring. It is observed by the authors that the drought indices identified herein do well for monitoring meteo- Reclamation Drought Index rological drought, but can present challenges in moni- toring agricultural drought. ET and soil water storage The Reclamation Drought Index was developed in are the two major components in the agricultural response to the Reclamation States Drought Assis- waterbudget.ETrepresentsthebasicwaterneedfora tance Act of 1988, which allowed states to seek assis- crop, thus is very beneficial in the calculation of an tance from the Bureau of Reclamation to financially agricultural drought index, and soil moisture storage mitigate for the effects of drought. The Reclamation indicates the basic water availability. Reference ET Drought Index is similar to the SWSI in that it is cal- can be estimated for most regions in the U.S. as culated at a river basin level. The Reclamation weatherdataarereadilyavailable.However,soilwater Drought Index uses precipitation, snowpack, stream valuesandsoilcharacteristicscanbedifficulttoobtain. flow, and reservoir levels (Hayes, 2006). The large Withoutaccuratedata,indices,includingsoilmoisture, spatial scale used in the Reclamation Drought Index may be difficult to compute on some scales. Because of limits its application for agricultural drought because this, a drought index is desired that includes ET, but of the variations in agricultural practices over large doesnotutilizesoilwaterorsoilcharacteristics. regions. Five drought indices that meet these constraints are: SPEI, ETDI, RDI, ADI, and RWD. These indices either use the difference in precipitation and PET of a Vegetation Condition Index reference crop or the ratio of actual ET to PET to quantify water availability to plants. Precipitation is The Vegetation Condition Index (VCI) was devel- considered as addition of water whereas ET is the loss oped by Kogan (1995). It uses the Normalized Differ- of water. The difference will illustrate the change in ence Vegetation Index (NDVI) obtained from satellite stored soil water. PET is the maximum rate at which data. The VCI is a normalization of the NDVI that water will be utilized by the plant. When a water allowsforrelativeassessmentsofchangesintheNDVI stress is imposed, the plants do not take up water at (Quiring, 2009). One issue with NDVI-based drought the potential rate; therefore, the ratio of actual ET to monitoringisthatwhereasNDVIdoesprovideanindi- PET will indicate water stress conditions. However, cation of plant health, it is difficult to distinguish the term PET refers to the maximum ET rate for a between crop types, and it is difficult to distinguish specific crop, which is difficult to calculate directly. whether plant stress is attributable to drought alone Thus, the PET for a reference crop of known height, orotherstressors(disease,pests,nutrients). under well-watered conditions, is often what is calcu- lated. This provides the PET for a reference crop or reference ET (ET ). For the indices that use the ratio o U.S. Drought Monitor of actual ET to PET, this ratio is only suitable if the actual ET and PET are for the same crop. The U.S. Drought Monitor combines many drought indices into a single graphical representation that provides a summary of drought conditions for the Evapotranspiration Deficit Index and Soil Moisture U.S. The U.S. Drought Monitor also uses input from Deficit Index over 270 multiagency experts across the U.S. (NDMC, 2013). This drought map is designed to be simulta- The Evapotranspiration Deficit Index (ETDI) and neously informative to the general public as well as Soil Moisture Deficit Index (SMDI) were developed by the scientific community. Because the U.S. Drought Narasimhan and Srinivasan (2005) in an attempt to Monitor was designed to be a graphical portrayal of provide a better index for monitoring agricultural drought conditions, its usefulness as a real-time agri- drought. Because of the SMDI using soil water as cultural drought monitoring tool is limited. input and the difficulty in acquiring soil water data, the SMDI is excluded from further discussion. More information on the SMDI can be found in Narasimhan Standardized Vegetation Index and Srinivasan (2005). After noticing the deficiencies in PDSI and SPI, Narasimhan and Srinivasan sought The Standardized Vegetation Index (SVI), devel- to develop an agricultural drought index that could be oped by Peters et al. (2002), uses a standardization produced at a much finer resolution than that of PDSI procedure on NDVI that is obtained from satellite or SPI. Because of the variability of precipitation and JOURNALOFTHE AMERICAN WATERRESOURCESASSOCIATION 5 JAWRA MOORHEAD, GOWDA, SINGH, PORTER, MAREK, HOWELL, AND STEWART soil characteristics, finer resolution is preferred for has been used globally for some time with the aridity droughtindicesthatutilizethesefactors.Thefirststep index developed by the United Nations Environmen- in calculating the ETDI is to calculate the weekly tal Program (UNEP, 1992). The aridity index is the waterstressratio(WS): long-term mean of annual mean precipitation divided by the annual mean ET and is defined as o PET(cid:2)AET WS¼ ð1Þ P PET j¼kP a ¼P j¼1 j ð2Þ k j¼kET where AET is the actual ET. The WS value has a j¼1 oj range from 0 to 1 with 0 indicating actual ET at the potential rate and 1 indicating no actual ET. After The aridity index indicates that both precipitation calculating the WS, the maximum, minimum, and and PET are needed to measure dryness (Tsakiris median water stress are used to calculate the weekly and Vangelis, 2005). Although the aridity index was water stress anomaly (WSA) as outlined in Narasim- intended to categorize regional climates, the parame- han and Srinivasan (2005). The WSA values will ters could be used to assess drought. The RDI uses range from (cid:2)100 to +100 with negative values indi- the ratio of cumulative precipitation and cumulative cating dry conditions and positive values indicating ET for a given period. This ratio can then be stan- o wet conditions. Seasonality is inherently removed dardized or normalized, as outlined in Tsakiris and from the WSA; therefore, it can be compared across Vangelis (2005), who note that drought cannot be seasons. A detailed explanation of the procedure can monitored using only water inputs (precipitation). be found in Narasimhan and Srinivasan (2005). Using the output (ET) provides a more complete When water is limited, plants cannot transpire at assessment of the status of water. the potential rate. Because of this fact, the ratio of The RDI can also be calculated at any time scale, actual ET to PET closely reflects the stress on the which adds versatility. In addition, it allows for com- plant. This ratio can bemore indicative ofagricultural parisontootherdroughtindicesatvarioustimescales. drought than the indices that use precipitation. Even Although ET can be difficult to calculate, there are though a shortage in precipitation is likely to cause simpler methods that require minimal data inputs. water stress on a plant, it is not a direct indication. With a simplistic ET calculation method, the RDI can Thelimitationtothisdroughtindexisthatactualcrop becalculatedwithoutrequiringcomplexdatasets.Con- ET data are difficult to obtain. In addition, the PET versely, whereas precipitation and ET are important must be for the same crop as the actual ET. Reference factors to consider in drought assessment, the ratio of ET is often used as the PET; however, reference ET is precipitation to ET is not a very satisfactory relation- only valid for the reference crop. Different crops will ship between the two parameters. Being that precipi- havedifferentrelationshipstoreferenceET;therefore, tation is a water input and ET is a water output, it is reference ET alone would be insufficient. It is possible morelogicaltousethedifferenceratherthantheratio. to estimate potential crop ET using reference ET and Also, even though there are simplistic methods avail- crop coefficients. This estimate would provide an ET ableforcalculatingET,themethodsthatrequiremore value based on well-watered conditions. To use an data inputs are found to produce a more accurate ET appropriate ETratio, itwould bemore accuratetouti- estimate. For example, the ASCE-EWRI Standardized lize actual ET, obtained from tools such as lysimeters Reference ET Equation (Allen et al., 2005) is widely or thermal remote sensing techniques (Gowda et al., used throughout the U.S. and around the world and it 2008), and estimated crop PET, obtained from refer- is generally considered more accurate than simpler enceETandcropcoefficients.Doingthiswouldprovide methodsthatarebasedonfewerdataparameters. a measure of the stress under given conditions com- pared to estimated, nonstressed conditions, but does includecomplicateddataneeds. Standardized Precipitation Evapotranspiration Index The Standardized Precipitation Evapotranspiration Reconnaissance Drought Index Index (SPEI) was developed by Vicente-Serrano et al. (2010) with the intention of defining a drought index The Reconnaissance Drought Index (RDI) was thatwouldbesensitivetoclimatechange.Vicente-Ser- developed by Tsakiris and Vangelis (2005) to develop rano et al. (2010) noted that the main factor influenc- a drought index that accounts for ET. They focused ing drought is precipitation; although other factors only on the natural processes that affect drought, such as air temperature, ET, wind speed, and soil such as precipitation, ET , and soil and vegetation waterholdingcapacitycanalsoinfluencedrought.The o cover characteristics. The ratio of precipitation to ET SPEI procedure is similar to that for SPI, but rather o JAWRA 6 JOURNALOFTHE AMERICAN WATERRESOURCESASSOCIATION IDENTIFYING AND EVALUATING A SUITABLE INDEX FOR AGRICULTURAL DROUGHT MONITORING IN THE TEXAS HIGH PLAINS than using only precipitation, it uses the difference time periods, and N is the number of periods with between precipitation and ET . In order to calculate less than 10 mm of precipitation. The full ADI calcu- o theSPEI,ET mustfirstbeestimated.Vicente-Serrano lation is provided in Sivakumar et al. (2011). o et al. (2010) used the Thornthwaite (1948) method but Whereas the ADI does use precipitation and ET, the notedthatothermethodsareavailableandacceptable. relationship used in the DI calculation is somewhat To calculate the SPEI, the difference (D) between the subjective. Using the precipitation to ET relationship i precipitationandET iscalculatedby is a valid method for monitoring drought; however, o the calculation should be based on a physical rela- Di ¼Pi(cid:2)EToi ð3Þ tionship. In addition, there is limited information available in the literature regarding the establish- The D values are then standardized which allows ment or verification of the ADI. i for comparison for different regions and time scales. Vicente-Serrano et al. (2010) provided equations for aggregating the D values for computing the SPEI at Relative Water Deficit i longer time scales. To standardize the D values, a i three-parameter log-logistic distribution is used. The The Relative Water Deficit (RWD) uses the ratio of three-parameter log-logistic distribution was selected actual ET to PET (Sivakumar et al., 2011) as: over a two-parameter log-logistic distribution to allow (cid:2) (cid:3) a range of values that can include negative numbers. AET RWD¼ 1(cid:2) 100 ð5Þ First, the probability-weighted moments (PWMs), or PET equivalently the L-moments, are calculated to esti- mate the values for a, b, and c parameters of the Using the ratio of actual to PET is an accurate log-logistic distribution and thus fit the distribution measure of water stress. In RWD, if ET is occurring to the data. The data are then standardized using an at the potential rate, the RWD is equal to zero, which equation provided in Vicente-Serrano et al. (2010). indicates no stress. The RWD represents the dimin- Similar to the SPI, the SPEI can be used across ished ET as a percentage that allows for comparisons different drought definitions as it can be calculated at across differing regions and time scales. In addition, multiple time scales. In addition, because SPEI uses the RWD can be calculated for any time scale. The ETo, it will inherently account for changes in wind main disadvantage to the RWD is that, whereas speed, temperature, and other parameters that affect using actual ET can provide a good indication of drought. One issue, however, is that using ETo in the water stress, it can be difficult to obtain. This can calculation can be misleading in characterizing agri- create problems when attempting to calculate the cultural drought as now explained. In agriculture, RWD for multiple sites or over a large geographical ETo is usually higher than PET of most crops and area. In many cases, a region will have varying crops PET of crops is zero during nongrowing season while in production. Using only one crop may not provide ETo may not be zero. Moreover, the PET of a particu- an accurate representation of multiple crops within lar crop also depends on the crop stages whereas ETo the region. is calculated for standard crop conditions. Therefore, Based on the literature review, it is concluded the use of ETo may not adequately express the water that the SPEI is the best method for agricultural demands of a diversified crop environment. drought monitoring as it uses both precipitation and ET as inputs and is relatively simple to use. How- o ever, because of a large variability in water require- Accumulated Drought Index ments of different crops grown in a region, using ET for calculating SPEI alone may not be suffi- o The Accumulated Drought Index (ADI) was devel- cient. Because crop ET is available for the Texas oped by the Integrated Center of Agrometeorological High Plains through the Texas High Plains ET Information in Brazil (CIIAGRO, 2012; Sivakumar (TXHPET) network, it would be advantageous to et al., 2011). This index uses precipitation and ET as calculate a drought index using crop ET to derive inputs. This index uses a relation between precipita- crop-specific drought indices that could provide tion and ET: options for utilizing the most appropriate index or a X crop-weighted SPEI. Therefore, in this study, the ADI¼ DI=ð3nNÞ ð4Þ SPEI was revised by replacing reference ET with potential crop ET to enhance representation of where DI is determined based on the relationship actual agricultural water demand and be applicable between precipitation and ET, n is the number of to the Texas High Plains. JOURNALOFTHE AMERICAN WATERRESOURCESASSOCIATION 7 JAWRA MOORHEAD, GOWDA, SINGH, PORTER, MAREK, HOWELL, AND STEWART MATERIALS AND METHODS In the northern Texas High Plains, approximately 55% of cropland is irrigated and uses about 1.76 Gm3 (1.43 million ac-ft) of water per year for irrigation Study Area (Marek et al., 2009). Irrigated winter wheat, corn, cotton, and grain sorghum are the predominant The Texas High Plains is a significant portion of crops, comprising 30, 26, 23, and 10% of the total irri- the Ogallala Aquifer region (Figure 1). Agriculture is gated area respectively (Colaizzi et al., 2008). Corn is a large portion of the land use and irrigated land a high water use crop requiring about 835 mm of ET accounts for the majority of agricultural production. from planting to maturity in the Texas High Plains Irrigation in this region uses about 89% of the total (New and Dusek, 2005). Because of inadequate pre- freshwater consumed, in contrast with about 60% for cipitation for corn production, an annual average of the state of Texas (Marek et al., 2010). The Texas over 480 mm (19 in.) of irrigation is required in this High Plains is a major corn and cotton producing region (Marek et al., 2009). Therefore, nearly all of region in Texas. The vast majority of irrigation water the corn produced in this region is irrigated. Cur- is withdrawn from the underlying Ogallala Aquifer. rently forage crops and silage feedstock are minor Under modern climate and geologic circumstances, crops in the region; however, there has been major the Ogallala Aquifer in the region receives little to no expansion in regional dairy production (Guerrero recharge, and is essentially being mined. Conse- et al., 2012) and forages may soon become major quently, conservation is an integral part of regional irrigated crops. water plans (Marek et al., 2009). The northern and In the southern Texas High Plains, cotton is the southern parts of the Texas High Plains are similar predominant crop comprising 65% of the total irri- in size; however, the northern Texas High Plains has gated area (Colaizzi et al., 2008). The popularity of about 1.1 million ha under irrigation while the south- cotton in this area is a reflection of the water ern Texas High Plains has about 760,000 ha under resource limitations where the saturated thickness of irrigation (Colaizzi et al., 2008). In both northern and the Ogallala Aquifer decreases near the southern southern regions, irrigated crop yields are at least boundary. Cotton requires less irrigation water than double neighboring dryland yields. the other predominant summer crops (Marek et al., FIGURE1. Locationsof19TexasHighPlainsEvapotranspirationWeatherStationsintheTexasHighPlains. JAWRA 8 JOURNALOFTHE AMERICAN WATERRESOURCESASSOCIATION