Comparison of Water Balance Predictions Made with HYDRUS-2D and Field Data From the Alternative Cover Assessment Program (ACAP)

Size: px

Start display at page:

Download "Comparison of Water Balance Predictions Made with HYDRUS-2D and Field Data From the Alternative Cover Assessment Program (ACAP)"

Frederick George
6 years ago
Views:

Albright2 1 University of Wisonsin papiwantrago@wis.edu benson@engr. wis.edu Madison W USA 20esert Researh nstitute bi/l@dri.

1 MODFlOW and More 2003: UndefStaning hrough Modeling - Conferene ProeedingS PoeW Zheng H. & Qd'8rty -_.mines.eduligwml Comparison of Water Balane Preditions Made with HYDRUS-2D and Field Data From the Alternative Cover Assessment Program (ACAP) Preeha Apiwantragoon 1 Craig H. Benson 1 and William H. Albright2 1 University of Wisonsin papiwantrago@wis.edu benson@engr. wis.edu Madison W USA 20esert Researh nstitute bi/l@dri.edu Reno NV USA ABSTRACT A omparison is made between water balane preditions obtained from HYDRUS-2D and field measurements from a test setion simulating a apillary barrier. The preditions. from HYDRUS-2D deviate appreiably from the field data. Differenes between the measurements and preditions are related to errors in prediting surfae runoff (SRO) and evapotranspiration (ET) and unertainties in the hydrologi properties of the soils. Greater differenes between the measured and predited SRO were obtained when hourly meteorologial data were used as input instead of daily data. nreasing the saturated hydrauli ondutivity until the measured and predited surfae runoff were essentially equal allowed more water to enter the soil profile and resulted in more realisti hanges in soil water storage (SWS) during the wetter portions of the reord. However alibration to the SRO data also resulted in large over-preditions of ET and muh lower SWS than was measured in the field. Adjusting the vegetation parameters had little effet on the preditions. Zero perolation was predited by the model for all ases. n ontrast approximately 50 mm of perolation was measured in the field. NTRODUCTON Alternative overs employing a water balane approah are being onsidered for apping waste ontainment failities loated in semi-arid and arid limates in lieu of onventional overs employing barriers with low saturated hydrauli ondutivity (e.g. lay layers and/or geomembranes). Alternative overs are being used beause they are often less ostly to onstrut and are believed to be less suseptible to damage aused by weathering. Design of alternative overs generally inludes simulating the hydrology of the over with an unsaturated flow model that aounts for interations between the soil plants and atmosphere. One of the models used by designers is HYDRUS-2D (Simunek et al. 1999) whih uses the finite element method to solve a modified form of Rihards' equation: a -8 :1) n Eq. 1 0 is volumetri water ontent t is time X is a spatial oordinate (i = 1 2) K is the unsaturated hydrauli ondutivity and K: are anisotropy fators and S is a sink term for plant water uptake. Although HYDRUS-2D inludes rigorous algorithms to simulate the proesses ontrolling unsaturated flow limited study has been undertaken to evaluate the auray of HYDRUS-2D for alternative over design (e.g. see Sanlon et al. 2002). Model auray S partiularly important in alternative over design and assessment because alternative overs generally are required to meet strit perolation riteria «1-3 mm/yr). The US Environmental Protetion Ageny (USEPA) urrently is onduting a long-term field study referred to as the Alternative Cover Assessment Program (ACAP). One of the purposes of ACAP is to evaluate and improve tools used to design and assess alternative overs suh as HYDRUS-2D. Data from 24 large-sale test setions being monitored by ACAP are available for this purpose (Albright and Benson 2002). A detailed desription of the test setions an be found in Roesler et al. (2002). 751

M9O'J;.OW8Rd More 2003: Understanding thrqlgh Modell!lg - Conferene PRleedings Poeter Zheng Hi. & Doherty - _.mines.

nput to the model was seleted to be as representative of on-site onditions as pratial.

TEST SECTON The field data are from an ACAP test setion bated in Marina CA. Marina has a semi-arid to sub-humid oastal limate with 460 mm of preipitation eah year on average.

A root barrier was plaed between the sandy lay and the sand.

The hydrology of the over is monitored using a large-sale (10m by 20 m) ysimeter that permits diret measurement of all water balane quantities exept evapotranspiration (ET) whih is omputed as the

Surfae runoff and perolation are monitored by measuring the rate at whih water aumulates in olletion basins.

Soil water storage is omputed by integrating the water ontent profiles. Meteorologial data are obtained from a weather station mounted on the test setion.

One-dimensional sjmulations were onduted vapor flow and heat transfer were assumed to be negligible both soil layers.

2 M9O'J;.OW8Rd More 2003: Understanding thrqlgh Modell!lg - Conferene PRleedings Poeter Zheng Hi. & Doherty - _.mines.edu/igwmd This paper desribes a omparison made between preditions from HYDRUS-2D and field data olleted from an ACAP test setion. nput to the model was seleted to be as representative of on-site onditions as pratial. Field meteorologial data and laboratory-measured hydrologi properties of the over soils were used as input whenever possible. TEST SECTON The field data are from an ACAP test setion bated in Marina CA. Marina has a semi-arid to sub-humid oastal limate with 460 mm of preipitation eah year on average. The over profile onsists of a m thik layer of sandy lay used for water storage over a 300-mm layer of poorly graded sand that forms a apillary break. A root barrier was plaed between the sandy lay and the sand. The test setion is vegetated with a mixture of grasses native to the Marina area (blue wild rye California brome reeping wild rye and paifi hairgrass). The hydrology of the over is monitored using a large-sale (10m by 20 m) ysimeter that permits diret measurement of all water balane quantities exept evapotranspiration (ET) whih is omputed as the residual of the water balane. Benson et al. (2001) provides a detailed desription of the ysimeter and Roesler et al. (2002) desribes the water balane omputations. Surfae runoff and perolation are monitored by measuring the rate at whih water aumulates in olletion basins. Soil water ontent is measured using low frequeny (40 MHz) time domain refletometry (TOR) probes and soil water potential is measured with thermal dissipation sensors. Soil water storage is omputed by integrating the water ontent profiles. Meteorologial data are obtained from a weather station mounted on the test setion. MODEL SET-UP HYDRU5-2D was used to simulate the hydrology of the apillary barrier at Marina. One-dimensional sjmulations were onduted vapor flow and heat transfer were assumed to be negligible both soil layers.in the over were assumed to be uniform homogeneous and isotropi and hysteresis was ignored. Unsaturated hydrologi properties of the over soils were desribed using the van Genuhten and van Genuhten-Mualem equations (van Genuhten 1980). Hydrologi properties used as input for many of the simulations are summarized in Table 1. These parameters are based on laboratory tests onduted on 15 undisturbed samples olleted during onstrution of the test setion (Gurdal et al. 2003). Geometri means were used for the saturated hydrauli ondutivity (K) and 11. Arithmeti means were used for Or 0. and n. The pore interation fator was assumed to be 0.5 for both soils. Saturated Hydrauli a The vegetation input to HYDRUS-2D Soil 8r 8. n onsisted of potential transpiration Condutivity (m/s) '' demand a root density funtion and _? 6.8x10'" a water stress funtion. Potential Sand 3.6x transpiration (PT) was omputed from potential evapotranspiration Table 1. Hydrologi properties of the over solis (PET) using the Ankeny-Rithie- Burnett equation (Fayer 2000) whih is based on leaf area index (LA). PET was omputed using the modified Doorenbos-Pruitt equation as implemented in UNSA T -H (Fayer 2000). The root density funtion was derived from a database ompiled by Winkler (1999). The water stress funtion in Feddeset al. (1978) was used with the following matri sutions: anaeorbiosis point = 3.6 m optimal point = 3.6 m limiting point = 26.9 m and wilting point = 450 m. The wilting point was estimated from the lowest water ontents measured in the root zone during the growing season and the water retention urve for the sandy lay. A root growth rate of 7.5 mm/d was assumed and the root depth was limited to the interfae between the sandy lay and sand layers. nitial onditions were assigned using water ontents measured in the field (0.23 for sandy lay 0.18 for sand). An atmospheri boundary ondition was applied at the surfae. Meteorologial data used for the atmospheri boundary are summarized in Roesler et al. (2002). A seepage fae boundary was applied at the base of the profile as reommended in Sanlon et aj. (2002) for modeling ysimeters. Mesh size was 752

MOOflC1N - More 2003:..16..:"-;::i--.g tn)ugh - C : ;-.w;. Po.-r. Zheng.. & Di8fty --.n.s. seleted iteratively so that the umulative mass balane remained less than 1 %. The vertial spaing was <0.

The minimum time step was set at 10-10 d and the water ontent and pressure head toleranes were set at 10' 6 and 10-3 m.

Results of these simulations are referred to herein as 'base ase' preditions. Water balane quantities predited with HYORUS-20 and measured n the field for the base ase are shown in Fig.

(2002) ndiate that runoff predited with HYDRUS-20 (and similar models) s sensitive to the temporal disretization of the nput.

Hour1y input resulted in larger over-preditions of SRO beause preipitation s applied with greater intensity when hour1y data are used (preipitation is uniformly distributed throughout the day with

3 MOOflC1N - More 2003:..16..:"-;::i--.g tn)ugh - C : ;-.w;. Po.-r. Zheng.. & Di8fty --.n.s. seleted iteratively so that the umulative mass balane remained less than 1 %. The vertial spaing was <0.1 mm at the boundaries and layer interfaes and was as large as 50 mm at interior points. The minimum time step was set at d and the water ontent and pressure head toleranes were set at 10' 6 and 10-3 m. BASE CASE PREDCTONS Simulations were onduted using hourly and daily meteorologial nput along with the other aforementioned input parameters. Results of these simulations are referred to herein as 'base ase' preditions. Water balane quantities predited with HYORUS-20 and measured n the field for the base ase are shown in Fig. 1 (preditions with HYORUS-20 are marked as H2O). Both hourly and daily data were used beause Sanlon et al. (2002) ndiate that runoff predited with HYDRUS-20 (and similar models) s sensitive to the temporal disretization of the nput. Surfae runoff (SRO) was over-predited appreiably for both hourly and daily input mainly during intense preipitation events. Hour1y input resulted in larger over-preditions of SRO beause preipitation s applied with greater intensity when hour1y data are used (preipitation is uniformly distributed throughout the day with daily input resulting in lower intensity). Sanlon et al. (2002) also found that SRO inreased in some ases when hour1y input was used instead of daily input. Evapotranspiration (ET) generally was underpredited espeially during wetter periods. ET predited using dally input was loser to that measured in the field. Daily input results in more infiltration (less SRO) making more water available for ET. Similar patterns of soil water storage (SWS) were predited with daily and hourly input and they generally follow the seasonal variations observed in the field (Fig. 1). However the hanges in soil water storage are smaller than those observed in the field largely due to the under-preditions of infiltration and ET. The relatively large derease in SWS predited early in the reord using dally input is due to a short-term over-predition of ET at the beginning of the reord. No peraation was predited using hourly or daily input whereas 49.8 mm of perolation was measured in the field. The most likely ause of the under-predition of perolation is due to the over-predition of SRO whih prevented the SWS from reahing a point that would result in appreiable drainage. Khire et al. (1997) also report that errors in SRO affet preditions of perolation as well as. all other subsurfae proesses. 753

MOOFLOWandMore 2003: Underslanding hh Modeq. Conferene Proeedk'9S. Poeter Zheng Hil&Doherty. _.mines.

Mathing the field-measured SRO required inreasing Ks of the sandy lay from 6.7x10-8m/s to 4.0x10-4m/s (i.e. nearly four orders of magnitude).

using hourly meteorologial input and the alibrated K. are shown in Fig. 2.

also allowed water to be removed by ET more readily (i.e. water ould move to the surfae more readily with higher Ks). However.

The under-predition of SWS is largely due to the over-predition of ET. The hanges in SWS are an exeption.

4 MOOFLOWandMore 2003: Underslanding hh Modeq. Conferene Proeedk'9S. Poeter Zheng Hil&Doherty. _.mines.edu/igwm/ Attempts were made to improve the preditions by adjusting Ks of the sandy lay until the predited SRO resembled the field-measured SRO (whih was nil). Mathing the field-measured SRO required inreasing Ks of the sandy lay from 6.7x10-8m/s to 4.0x10-4m/s (i.e. nearly four orders of magnitude). All other parameters were the same as those in the base ase simulations. Water balane preditions made with HYORUS- 20 using hourly meteorologial input and the alibrated K. are shown in Fig. 2. Calibrating the model resulted in larger hanges in SWS and higher ET largely due to additional water entering the soil profile (Fig. 2). Higher K. also allowed water to be removed by ET more readily (i.e. water ould move to the surfae more readily with higher Ks). However. the preditions made with the alibrated model generally are poorer than those using base-ase input. ET tends to be grossly over-predited and SWS is grossly under-predited. The under-predition of SWS is largely due to the over-predition of ET. The hanges in SWS are an exeption. The hanges in SWS predited with the alibrated model during the wetter months are loser to the hanges in SWS observed in the field. Despite the higher Ke the alibrated model predited no perolation whih is likely due to the large under-predition in SWS. That is the soil never beame wet enough to result in appreiable drainage. SURFACE RUNOFF CALBRATON 400 E g "Q. 200 C m '5 0- m.:t E.s. 5 o ' &. w 100 J...:: r ET H2D. _ -' Feld ET r Preitation 0 V -- Field & Field SWS -J H2O SWS.. t" Fl81d Perolation :. J Perolation PROPERTES OF VEGETATON #PetionH2D Attempts were made to redue the over-predition 0 0 of ET obtained early in the reord by adjusting the 5/1/00 8/1/00 11/1/002/1/01 5/1/01 8/1/01 "tj CD (") 400 OJ "2: o' C- OO =R wilting point and maximum root depth of the vegetation both of whih are not known with Figure 2. Field data and water balane ertainty. The wilting point was varied from 450 m preditions from HYDRUS-2D alibrated by (base ase) to 150 m and the maximum root depth adjusting K. until the predited and measured was redued from 1.22 m to 0.75 m. n addition a SRO mathed. simulation was onduted where transpiration was set to zero during the first growing season (May- De. 2000). Adjusting these parameters had negligible effet on the water balane preditions. CONCLUSONS Water balane preditions made with HYDRUS-2D and field measurements from a test setion simulating a apillary barrier over have been ompared in this paper. Preditions made with HYDRUS-2D deviated appreiably from the field data. The differenes appear to be related to errors in prediting surfae runoff (SRO) and evapotranspiration (ET) and unertainties in the hydrologi properties of the soils. Greater SRO "0 g 3" :?- "U [ '0 G 0- j G j Co Ch G 2 ;u -n.9 '3.! 754

Adjustments were made to the model by inreasing the saturated hydrauli ondutivity (K.) until the measured and predited SRO mathed and by varying the parameters desribing the vegetation. nreasing K.

5 MODFlOW n More 2003:.it6.-':'- fn)ugh - -.:. Poe8r... & Doherty -_.rit88.edufgwmo' differenes between the measured and predited SRO were obtained when hourly meteorologial data were used as input instead of daily data. Adjustments were made to the model by inreasing the saturated hydrauli ondutivity (K.) until the measured and predited SRO mathed and by varying the parameters desribing the vegetation. nreasing K. allowed more water to enter the soli profile and resulted in more realisti hanges in soil water storage (SWS) during the wetter portions of the reord. However inreasing K. also resulted in large over-preditions of ET. and muh lower SWS than was measuted in the field. Adjusting the vegetation parameters had little effet on the preditions. Zero perolation was predited by the model for all ases despite the adjustment of Ks. n ontrast approximately 50 mm of perolation was measured in the field. This disrepany suggests that perolation rates predited by HYDRUS-2D (and other similar models) an only be onsidered as estimates and should be interpreted with aution. ACKNOWLEDGEMENT Funding for ACAP has been provided by the United States Environmental Protetion Ageny (USEPA) through the Superfund nnovative Tehnology Evaluation (STE) program. Steve RoCk of USEPA is the projet manager. Support for onstrution and instrumentation of the ACAP test setions was also provided by site owners. The findings and opinions in this artile are solely those of the authors. Endorsement or approval by USEPA or the site owners is not implied and should not be assumed. REFERENCES Albright W. Benson C Alternative Cover Assessment Program 2002 Annual Report. Publiation No Desert Researh nstitute Reno NV USA. Fayer M Unsaturated Soil Water and Heat Flow Model: v 3.0 Theory User Manual and Examples Report No Paifi Northwest National Laboratory Rihland WA USA Feddes R.A. Kowalik. P.J. Zaradny H Simulation offield water use and rop yield John Wiley & Sons New York. Gurdal T. Benson C. Albright W Hydrologi Properties of Final Cover Soils from the Alternative Cover Assessment Program Geo-Engineertng Report No University of Wisonsin Madison W USA. Khire M. Benson C. Bossher P Water balane modeling of earthen landfill overs J. of Geotehnial and Geoenvironmental Eng. 123(8) Roesler A. Benson C. Albright W Field Hydrology and Model Preditions for Final Covers in the Alternative Cover Assessment Program Geo-Engineertng Report No University of Wisonsin Madison W USA. Sanlon B. Christman M. Reedy R. Porro. Simunek J. Flerhinger G nterode omparisons for simulating water balane of surfiial sediments in semi-arid regions Water Resour. Res. 38(12) Simunek J. Sejna M. van Genuhten M The HYDRUS-2D Software Pakage for Simulating the Two-Dimensional Movement of Water Heat and Multiple solutes in Variably-Saturated Media v. 2.0 US Salinity Lab. Agr. Res. Serv US. Dept. of Agri. Riverside CA USA. van Genuhten M A dosed-form equation for prediting the hydrauli ondutivity of unsaturated soils Soil $d. So. Am. J Winkler W Thikness of Monolithi Covers in Arid and Semi-arid Climates MS Thesis University of Wisonsin Madison W USA. 755

6 q 'l Volume (pages ) y a( y H=u - YY L (5..-1X5..- k. k1 kz W ).lvz J 'l!; v- = + w E!; s Organizing Committee Eileen Poeter GWMC Colorado Shool of Mines Chunmiao Zheng University of Alabama Mary Hill U.S. Geologial Survey John Doherty Watermark Computing Australia NTERNATONAL GROUND WATER MODELNG CENTER (GWMC) COLORADO SCHOOL OF MNES

GROUND WATER/SURFACE WATER INTERACTIONS AWRA SUMMER SPECULLTP CONF~NCR CZ-I PDG- ANALYSIS OF INFILTRATED STREAM WATER TO A PUMPING WELL

GROUND WATER/SURFACE WATER INTERACTIONS AWRA SUMMER SPECULLTP CONF~NCR CZ-I PDG- ANALYSIS OF INFILTRATED STREAM WATER TO A PUMPING WELL JULY 1-3 GROUND WATER/SURFACE WATER NTERACTONS AWRA SUMMER SPECULLTP CONF~NCR 22 32 CZ- PDG- ANALYSS OF NFLTRATED STREAM WATER TO A PUMPNG WELL Xunhong Chen' and Xi Chen' ABSTRACT: Pumping of groundwater