Available online at www.sciencedirect.com
Procedia Environmental Sciences 19 (2013) 534 – 542
Four Decades of Progress in Monitoring and Modeling of Processes in the Soil-PlantAtmosphere System: Applications and Challenges
Improving simulation of soil water balance using lysimeter observations in a semiarid climate M. Soldevilla-Martineza, 5López-Urreab, /Martínez-Molinab, 0Quemadaa -,Lizasoa* a
Department of Crop Production, ETSIA. Technical University of Madrid,, Avenida Complutense s/n. 28040 Madrid,Spain b Water Management Research Unit. ITAP-FUNDESCAM., c/Gregorio Arcos s/n, PO Box 451, 02080, Albacete,Spain
Abstract Water balance simulation in cropping systems is a very useful tool to study how water can be used efficiently. However this requires that models simulate an accurate water balance. Comparing model results with field observations will provide information on the performance of the models. The objective of this study was to test the performance of DSSAT model in simulating the water balance by comparing the simulations with observed measurements. The soil water balance in DSSAT uses a one dimensional “tipping bucket” soil water balance approach where available soil water is determined by the drained upper limit (DUL), lower limit (LL) and saturated water content (SAT). A continuous weighing lysimeter was used to get the observed values of drainage and evapotranspiration (ET). An automated agrometeorological weather station close to the lisymeter was also used to record the climatic data. The model simulated accurately the soil water content after the optimization of the soil parameters. However it was found the inability of the model to capture small changes in daily drainage and ET. For that reason simulated cumulative values had larger errors as the time passed by. These results suggested the need to compare outputs of DSSAT and some hydrological model that simulates soil water movement with a more mechanistic approach. The comparison of the two models will allow us to find which mechanism can be modified or incorporated in DSSAT model to improve the simulations. © 2013 The The Authors. Authors. Published Publishedby byElsevier ElsevierB.V B.V.Open access under CC BY-NC-ND license. © 2013 ofof thethe conference Selection and/or and/or peer-review peer-reviewunder underresponsibility responsibilityofofthe theScientific ScientificCommittee Committee conference. Keywords: Water balance; DSSAT; drainage; evapotranspiration; soil water content; optimization.
* Corresponding author. Tel.: +34 91 452 4900 ext1671. E-mail address:
[email protected] (Jon Lizaso).
1878-0296 © 2013 The Authors. Published by Elsevier B.V Open access under CC BY-NC-ND license. Selection and/or peer-review under responsibility of the Scientific Committee of the Conference doi:10.1016/j.proenv.2013.06.060
M. Soldevilla-Martinez et al. / Procedia Environmental Sciences 19 (2013) 534 – 542
1. Introduction: Water balance simulation in cropping systems, is essential to determine available water for the crop and the possible environmental impact due to the solutes lixiviation. Comparing model results with field observations will provide information on the performance of the model and will reveal strengths and weaknesses. This is essential in selecting appropriate models for practical applications in water resources analysis and/or identifying required model improvements. In this work the water balance of the Decision Support System for Agrotechnology Transfer (DSSAT) (Hoogenboom et al., 2010) was evaluated. DSSAT is a suite of crop models sharing the simulation of common soil processes. In previous work we found some problems in the DSSAT simulation of the soil water balance components . The main issues were related with the simulation of drainage and evapotranspiration (ET). In this experiment, various irrigation cycles were applied to a weighting lysimeter to generate a number of combinations of drainage and ET. The main objective of this study was to test the performance of DSSAT when simulating the water balance components by comparing simulations and observed measurements. Firstly, the soil paremeters were optimized and after the model was tested under different irrigation cycles. Once the water balance simmulation by DSSAT is checked, it could be analyzed the influence of soil water movement on solutes lixiviation in a future study. 1.1. DSSAT model: The soil water balance in DSSAT is based on Ritchie’s model which uses a one dimensional “tipping bucket” soil water balance approach (Ritchie1972; Ritchie 1981a; Ritchie 1981b). Per-layer available soil water is determined by the drained upper limit (DUL), lower limit (LL) and saturated water content (SAT), defined for each layer of the soil profile in the SOIL.SOL file. The water in the upper layer cascades to the lower layers mimicking the process of a series of linear reservoirs. Soil water infiltration is computed by subtracting runoff from rainfall/irrigation. Runoff is calculated with the SCS method (Soil Conservations Service, 1972) based on a curve number defined in the soil profile. Downward saturated flow takes place when a layer water content is above the drained upper limit. Upward flow caused by transpiration and soil evaporation it is calculated within the Soil-Plant-Atmosphere module in DSSAT. Potencial evapotranspiration (ET0), is calculated and partitioned into potential plant evaporation and potential soil evaporation. Then, the actual evapotranspiration is calculated by applying reduction factors, considering the soil moisture conditions. RAINFALL
IRRIGATION
SOIL EVAPORATION
RUNOFF INFILTRATION
SOILWATER MOVEMENT
DRAINAGE
Fig. 1. Soil water balance simulated by DSSAT
535
536
M. Soldevilla-Martinez et al. / Procedia Environmental Sciences 19 (2013) 534 – 542
2. Materials and methods: 2.1. Experimental design: The field experiment was carried out in the experimental lysimeter station “Las Tiesas” (Albacete, Spain) supported by the “Instituto Técnico Agronómico Provincial” (ITAP), during 2011 and 2012. The field is located in Albacete (Spain), 39º 14’ N, 2º 5’ W, and 695 m. A weighting lysimeter on bare soil with continuous electronic data reading devices was used in the experiment. The soil was cultivated previously with sunflower that was harvested and the residues removed before the beginning of the experiment. The dimensions of the lysimeter recipient were 2.3 m x 2.7 m and 1.7 m depth, with approximately 14.5 t total mass. The lysimeter recipient is surrounded by a square protection plot to avoid runoff and is located in the center of a hectare that is cultivated following the same procedures.The essay hosted also a weighing lysimeter cultivated with grass monitoring reference evapotranspiration (ET0). In the bare soil lysimeter, ET was calculated daily based on the registered weight, corrected by drainage. Daily weather and soil parameters were measured at the site. 2.2. Water management Water management was done in three irrigation cycles. First cycle (November 3th until December 15th, 2011) started irrigating the soil until field capacity and leaving it to dry after. In a second cycle (December 15until March 1, 2012) soil was no irrigated. Third cycle (March 1until March 31, 2012) soil was irrigated with 77mm of water and left one month drying. 2.3. Weather: Weather information was collected by a weather station located in the experimental field. The area has a semi-arid, continental climate. The registered weather data was: Relative air humidity and air temperature at 2 m; net short wave radiation at 2 m: net long wave radiation at 2 m, Soil heat flux at 0.05, 0.1, 0.2 and 0.3 m: Atmospheric pressure at 2 m and wind speed and direction, precipitation and evaporation. The mean annual maximum and minimum daily air temperatures for 2011 were 20.9 and 7ºC, respectively. Mean average sunshine was 17.3 MJ/m2 and annual average precipitation 1 L/m2. Rainfall
ET0
Tmaxaverage
Tminaverage
Tmeanaverage
80
20 15
60 50
10
40 30
5
20 10
0
Temperatureª(ºC)
RainfallandET0(mm)
70
0 Ͳ10
Ͳ5
November
December
January
February
March
Fig.2. M onthly rainfall, ET0 measured in the reference lysimeter, average of maximal temperatures, average of minimal temperatures, and average of mean temperatures in Las Tiesas during the first studied period (Nov 2011-Mach 2012).
537
M. Soldevilla-Martinez et al. / Procedia Environmental Sciences 19 (2013) 534 – 542
2.4. Soil: The soil is classified as Petrocalcic Calcixerepts (Soil Survey Staff, 1999). The soil depth of the experimental plot is 170 cm, with a fragmented petrocalcic horizon at 60 cm depth approximately. Texture is silty-clay-loam, with a uniform basic pH across the profile. Additional information is available elsewhere (López-Urrea et al., 2006). Table 1. Physical and chemical properties of the tested soil at different depths. Layer (cm) Property
0-5
5-15
15-63
63-67
67-96
96-170
BD
1.39
1.39
1.49
1.8
1.49
1.7
pH
8.1
8.1
8.1
8.1
8.2
8.2
CEC cmol kg-1
27.8
27.8
17.9
17.9
10.4
10.4
Organic C , %
0.96
0.96
0.46
0.46
0.24
0.23
Total N , %
0.13
0.13
0.08
0.01
0.08
0.01
Coarse fraction
21
21
50
95
60
90
Silt
48.9
48.9
46.4
46.4
50.8
50.8
Clay
37.7
37.7
30.8
30.8
23.2
23.2
Texture , %
Bulk density (BD), coarse fraction, saturated hydraulic conductivity (Ks), residual water content (WCR), and gravimetric and volumetric humidity (VH) were measured at the beginning of the experiment from soil samples extracted from the studied field. The other parameters were taken from doctoral dissertations ((Maturano, 2002; López-Urrea 2004) 2.5. Drainage measurement: Drainage was continuously measured with a tipping bucket rain gauge (HOBO 200, Davis Instruments, Hayward, California, USA) installed at the outlet of the bottom of the lysimeter and connected to a data logger registering the information. The pluviometer was previously calibrated in the laboratory showing a ratio of 6.5 ml/tip 2.6. Soil moisture measurement: The soil water content was monitored hourly using capacitance sensors (10HS ECH2O, Decagon Devices Inc., Pullman, WA) located at 10 and 40 cm depth. The sensors outputs were normalized with a normalization equation based on frequency readings of the sensors exposed to air and water, to determine a scale frecuency (SF). The average SF were transformed into volumetric water content (șv) using a calibration equation that was obtained under laboratory conditions using soil samples from the experimental site according to the procedure described by Gabriel et al. (2010). This calibrated relationship (șv = 1.1052 SF-0.0927) covered a șv range from 0.07 to 0.8 m3 m-3, and had a correlation coefficient r2 = 0.95.
538
M. Soldevilla-Martinez et al. / Procedia Environmental Sciences 19 (2013) 534 – 542
2.7. Model optimization and simulation: In this study we used DSSAT v4.5. Weather inputs were registered by a weather station located in the experimental field. The soil profile was divided into six soil layers, with the upper two layers of 5 and 10 cm to improve the accuracy of the simulation. The soil water content in DSSAT was initialized according to the field measurements. Readings from the capacitance sensors at 10 and 40 cm depth were complemented with gravimetric soil sampling for deeper layers. The methods used in the DSSAT simulations were: FAO-56 (Doorenbos y Pruitt, 1977) for evapotranspiration, Ritchie (Ritchie, 1998) for water balance and infiltration and Suleiman- Ritchie (Suleiman and Ritchie, 2003; Ritchie et al., 2009) for soil evaporation. To reduce the uncertainty associated to soil inputs, the optimization algorithm, Simulated Annealing (SA), as implemented by Goffe et al. (1994), was used. SA found the best collection of soil inputs by minimizing the sum of squares of the difference between predicted and measured outputs (SSE) of soil water content in the upper layers, drainage, and ET. The optimized soil inputs included surface parameters (drainage rate, runoff curve number), and per-layer parameters (LL, DUL, SAT). The optimization started with reasonable ranges of SAT, calculated from the total porosity obtained from field measures of bulk density. DUL and LL were also subsequently optimized. Observed and simulated outputs were normalized using the range of measured values, to provide the same weight to outputs of different magnitudes during the optimization process. 3. Results and discussion: 3.1. Optimization of soil parameters: Table 2 shows the soil parameters before and after the optimization. and Figure 3 depicts the impact of input optimization on the simulated components of the soil water balance . Table 2. Characteristics of the lysimeter soil used in the DSSAT simulations. Soil Layer (cm) Before optimization 0-5 5-15 15-63 63-67 67-96 96-170 After optimization 0-5 5-15 15-63 63-67 67-96 96-170
LL(cm3 cm-3)
DUL(cm3 cm-3)
SAT(cm3 cm-3)
0.254 0.254 0.242 0.120 0.160 0.160
0.374 0.374 0.414 0.414 0.414 0.414
0.449 0.449 0.497 0.497 0.497 0.497
0.185 0.262 0.092 0.230 0.171 0.013
0.195 0.272 0.251 0.274 0.207 0.197
0.340 0.364 0.364 0.305 0.217 0.317
LL: Lower limit, cm3 cm-3 DUL: Upper limit, drained, cm3 cm-3 SAT: Upper limit, saturated, cm3 cm-3
539
M. Soldevilla-Martinez et al. / Procedia Environmental Sciences 19 (2013) 534 – 542
Surface values of drainage rate (DR) and curve number (RO) were also optimized. The original DR was 0.75 and RO was 45. After the optimization DR and RO were 0.57 and 28 respectively. Before optimization
After optimization
0.450
0.320
SW(mm)
0Ͳ10cm
0Ͳ10cm
0.300
0.280
0.150
0.240
Observed
Simulated 0.200
0.000
SW(mm)
0.400
1
21
41
61
10Ͳ40cm
81
101
121
141
1
21
41
61
81
101
121
141
10Ͳ40cm
0.280
0.300
0.200
0.240
0.100
0.000
0.200
Drainage(mm)
80 1
21
41
61
81
Drainage
101
121
141 1
41
61
81
Drainage
101
121
141
80
60
60
40
40
20
20
0
0 4
DailyET(mm/day)
21
1
21
41
61
DailyET
81
101
121
141
1
21
41
61
DailyET
81
101
121
141
3
3
2
2
1
1
0
0 1
CumulativeET(mm)
4
21
41
200
61
ET
81
101
121
141
1
21
41
61
ET
81
101
121
141 200
150
150
100
100
50
50
0
0
1
21
41
61
81
101
121
141 1
21
41
61
81
101
121
141
Days Fig.3. Observed and Simulated components of the soil water balance (soil water content, drainage and daily and cumulative ET) during the optimization period (October 2011-March 2012) before and after optimization.
It is shown in the figures how optimization of parameters greatly improved DSSAT simulations of soil water content reducing the RMSE in 80% and 90% for the 0-10 cm and 10-40 cm respectively. Also the drainage simulation was improved since it was not simulated drainage before the optimization, reducing the drainage RMSE in 30%. However it was found the inability of the model to capture small changes in
540
M. Soldevilla-Martinez et al. / Procedia Environmental Sciences 19 (2013) 534 – 542
daily drainage and ET. For that reason cumulative values have larger errors as the time passed by as it is shown in the figure 3. 3.2. Soil water balance: Once the soil profile was calibrated for the first period of time (11/3/2011-3/31/2012), the second period (11/1/2012- 12/15/2012) of the experiment was simulated with the improved soil inputs. Figure 4 indicates that the soil moisture was simulated quite accurately. Larger daily error was found for drainage and ET, with RMSE between observed and predicted values of 9 and 23 mm respectively. 30
Water(mm)
Rainfall
Irrigation
20 10 0 1
6
11
16
21
26
0Ͳ10cm
0.4 0.3 0.2 0.1
Observed
Simulated
0.0 1
6
11
16
21
26
10Ͳ40cm
0.4 0.3 0.2 0.1 0.0
31
1
5
6
11
16
21
26
31
250
CumulativeET(mm)
DailyET(mm/day)
31
0.5
Soilwater(cm3cmͲ3)
Soilwater(cm3cmͲ3)
0.5
4 3 2 1 0
200 150 100 50 0
1
6
11
16
21
26
31
1
6
21
26
11
16
21
26
31
Drainage(mm)
80 60 40 20 0 1
6
11
16
31
Days
Fig.4. Simulated and observed components of the soil water balance (soil water content, drainage and ET) during the second period (11/1/2012- 12/15/2012) of the experiment.
M. Soldevilla-Martinez et al. / Procedia Environmental Sciences 19 (2013) 534 – 542 Table 3. Total observed and simulated soil water balance with DSSAT for the second experimental period. Water balance
observed
simulated
¨SW
21.82
20.8
Effective Irrigation (I)
0.0
Precipitation (P)
84.76
0.0 79.5
Drainage (D)
62.3
62.3
Runoff (R)
0.0
0.0
Soil Evaporation (P)
47.4
38.0
Final Balance
3.12
0.0
Final balance= (P+I)-(D+R+E) ±¨SW
We were unable to detect any possible error in the simulated daily water balance. Also, the balance for the whole period equaled to zero as shown in Table 3. However, although the global balance is correct, the distribution of water between the components needs to be improved. The soil water content in the first 40 cm of soil was greatly enhanced after the optimization in both studied periods. Drainage and ET simulations however, especially in the first period, were not as accurate as expected. DSSAT drainage simulation seemed unable to reproduce the small drainage occurring over extended time periods. It rather exhibited a stepping curve with strong variations of drainage in a short period of time. ET simulation was consistently underestimated during the first period. Drainage was also underestimated in the first period but overestimated in the second. These results suggested the need to compare outputs of DSSAT and some hydrological model that simulates soil water movement with a more mechanistic approach. The comparison of the two models will allow us to find which mechanism can be modified or incorporated in DSSAT model to improve the simulations.
Acknowledgements This work was funded by Comunidad Autónoma Madrid (AGRISOST, S2009/AGR1630). We would also thank the staff from “Las Tiesas” field station and the financial support of the projects AGL200913124 (Science and Innovation Ministry, Spain) and PPII-0319-8732 (Education and Science Council, JCCM, Spain).
References [1] Hoogenboom G, Jones JW, Wilkens PW, Porter CH, Boote KJ, Hunt LA, Singh U, Lizaso JL, White JW, Uryasev O, Royce FS, Ogoshi R, Gijsman AJ, Tsuji GY. Decision Support System for Agrotechnology Transfer (DSSAT) Version 4.5 [CD ROM].University of Hawaii, Honolulu, Hawaii, 2010. [2] Ritchie JT. A model for predicting evaporation from a row crop with incomplete cover. Water Resources Research 1972;8:12041213. [3] Ritchie JT. Water dynamics in the soil-plant-atmosphere. Plants and Soil 1981a;58:81-96. [4] Ritchie JT. Soil water availability. Plants and Soil 1981b;58:327-338. [5] Soil Conservation Service (SCS) National Engineering Handbook. Hydrology Section 4, Chapters 4/10, 1972. [6] Soil Survey Staff. Soil Taxonomy. A basic system of soil classification for making and interpreting soil surveys. Second Edition. Agriculture handbook n. 466. USDA, Washington, 1999. [7] Lopez-Urrea R, Martin de Santa Olalla F, Fabeiro C, Moratalla A. Testing evapotranspiration equations using lysimeter observations in a semiarid climate. Agricultural Water Management 2006;85(1-2):15-26.
541
542
M. Soldevilla-Martinez et al. / Procedia Environmental Sciences 19 (2013) 534 – 542
[8] Maturano M. Estudio del uso del agua y del nitrógeno dentro del marco de una agricultura sostenible en las regiones maiceras Castellano-Manchega y Argentina. Tesis Doctoral. Departamento de Producción Vegetal y Tecnología Agraria. Universidad de Castilla-La Mancha, 2002. [9] Lopez-Urrea R. Evapotranspiración de referencia: métodos de cálculo y de medición directa en una estación lisimétrica en ambientes semiáridos. Tesis Doctoral. Departamento de Producción Vegetal y Tecnología Agraria. Universidad de Castilla-La Mancha, 2004. [10] Gabriel JL, Lizaso JI, Quemada M. Laboratory versus field calibration of capacitance probes. Soil Sci. Soc. Am. J. 2010;74:593-601. [11] Doorenbos J, Pruitt WD. Guidelines for predicting crop water requirements. FAO, Irrigation and Drainage. Paper No. 24. Rome; 1977. [12] Ritchie JT. Soil water balance and plant water stress.In Understanding options of agricultural production, eds. G.Y. Tsuji, G. Hoogenboom, and P.K. Thornton. Dordrecht, The Netherlands: Kluwer Academic Publishers and International Consortium for Agricultural Systems Applications 1998;p.41-53. [13] Suleiman AA, Ritchie JT. Modeling soil water redistribution during second-stage evaporation. Soil Science Society of America Journal 2003;67(2):377-386. [14] Ritchie JT, Porter CH, Judge J, Jones JW, Suleiman AA. Extension of an existing model for soil water evaporation and redistribution under high water content conditions. Soil Science Society of America Journal 2009;73(3):792-801. [15] Goffe WL, Ferrier GD, Rogers J. Global optimization of statistical functions with simulated annealing. J.Econometrics 1994;60(1):65-99.
