Scientific Registration n o : 1359 Symposium n o : 03 Presentation : poster Use of pedotransfer functions in crop growth simulation Utilisation de fonctions de pédotransfert dans la simulation de la croissance des cultures SINGH Anil Kumar Water Technology Centre, Indian Agricultural Research Institute, New Delhi-110012, India ABSTRACT Under water limiting situations, estimates of four critical parameters are required as inputs for integral simulation models, for predicting crop growth and yield. These are the soil water content values at saturation, field capacity, wilting point and air dryness. For many locations actual measurements are generally not available. Soil survey data routinely provides layerwise distribution of the textural components, organic carbon content and bulk density values. Simple relationships have been developed between the easily available or measurable soil properties routinely reported in literature and the four critical soil water constants. The data was collected from various sources representing the various agro-ecological regions of the country. Soil water content at saturation and wilting point were significantly correlated and satisfactorily estimated using clay content. Field capacity was 2negatively correlated and satisfactorily estimated using the sand content. The difference in R using all the variables viz., sand, silt, clay, organic carbon content and bulk density, together and the properties taken individually, was generally less than two per cent. Hygroscopic coefficient values were assumed to represent the air dry moisture and could be estimated sufficiently accurately by using the clay content. The predicted values were very close to the estimated with the correlation coefficient (r) significant at P=0.01 for individual parameters. It was shown that the four soil water constants estimated by these simple relationships 2 for 20 randomly selected soils were also very close to the observed values (R =0.9472). The saturated hydraulic conductivity was negatively but significantly correlated with clay content but not as satisfactorily as the other soil water constants. 1
INTRODUCTION Irrigation and/or rainfall, run-off, infiltration, deep percolation etc. are components of soil water balance which are controlled by the hydrophysical properties of the soil and are directly or indirectly responsible or its availability to plants. A realistic soil water balance is, therefore, an integral component of crop growth models. Models that simulate movement and uptake of water are generally categorized into capacity/integral (simple) and mechanistic models and the input data are dependent on the level of detail required. In the simplest of approaches for predicting crop growth and yield under water limiting situations like SAHEL (Penning de Vries et al., 1989), estimates of four critical parameters are required as inputs. These are the soil water content values at saturation, field capacity, wilting point and air dryness. The CERES crop growth models (Tsuji et al., 1994) need saturated hydraulic conductivity as an additional input. Actual measurements of these properties are generally not available for many locations when the model has to be used for regional applications. However, soil survey data routinely provide layerwise information of textural components, organic carbon content and occasionally bulk density values. Attempts have, therefore, been made by many workers to develop models for soil water retention characteristics e.g. Clapp and Hornberger (1978), Gupta and Larson (1979) and Rawls and Brakensiek (1982), and transmission properties e.g. Vereecken et al. (1990). In this paper, the possibility of developing simple models which use easily measurable or available properties as inputs, has been explored. The effect of the estimated values on the yield of wheat using a model WTGROWS (Aggarwal et al., 1994) was also examined. DATABASE In capacity type models like SAHEL, instead of the complete soil water retention curve the input is confined to four constants only. These are soil water constants at saturation (θsat), field capacity (θfc), wilting point (θwp) and air dryness (θad). θsat represents the total porosity of soil or water held by soil at 0.0 MPa matric potential. θfc is the water held by the soil at -0.033 MPa matric potential and is considered as the upper limit of available water for crops other than rice. θwp is the water held by the soil at -1.50 MPa matric potential and assumed to be the lower limit for available water for non-rice crops. θad is the lower limit of water content up to which the soil could lose water through evaporation and it was taken to be the soil water content corresponding to the hygroscopic coefficient value (-3.10 MPa matric potential). Published literature was scanned for data on actual measurements of soil water retention values corresponding to the values stated above. The other properties of the soil included were sand, silt, clay, organic carbon and bulk density values. Only those data sets which were complete in all respects were retained for analyses. The data collected from various 2
sources(ali et al., 1966; Gupta et al., 1984; Yadav et al., 1995) were representative of the major agro-ecological zones of India and covered soils of alluvial, black, laterite and lateritic, red and desert regions of the country. The data sets are summarized in Table 1. Table 1 : Summary of soil properties used in the database (Number of data sets = 256) Range Variables Mean Minimum Maximum Sand(%) 54.27 0.30 87.80 Silt(%) 18.99 0.50 68.00 Clay(%) 26.01 4.00 79.80 Organic Carbon(%) 0.65 0.02 4.90 Bulk density(g/cm3) 1.47 1.01 1.82 θsat (cm3/cm3)@ 0.58 0.35 0.89 θfc (cm3/cm3) 0.29 0.09 0.60 θwp (cm3/cm3) 0.13 0.02 0.35 θad (cm3/cm3)* 0.06 0.01 0.19 Ksat (cm/day)# 50.93 2.22 320.00 @based on a subset of 100 observations; *based on 75 observations; #based on 58 observations RESULTS AND DISCUSSION As a first step, the correlation coefficients between all the variables were calculated. All variables were significantly correlated. The dependence of soil water retention characteristics on the soil physical and related properties has been reported by many workers (Ali et al., 1966; Biswas and Ali, 1967; Velayutham and Raj, 1971; Gupta and Larson, 1979; and others). It indicated the possibility of developing simple relationships with a satisfactory level of accuracy. Multiple and step wise regression analyses were carried out with θsat, θfc, θwp and θad as dependent variables and the remaining as independent variables to develop pedotransfer functions. When all five independent variables were taken together the R 2 ranged from 65 to 87 per cent (significant at P=0.01 level). However, when step wise regression analysis was carried out, the number of variables was reduced from five to two only with a sacrifice in overall R 2 by around 0.5 per cent. When only one independent variable, i.e. the variable that entered at step I during the step wise regression analysis was considered, the R 2 value decreased generally by less than two per cent compared to the multiple regression using all the five variables. On this basis, four simple regression models have been proposed for estimating the four critical soil water 3
contents. They are given below and the closeness of fit between the measured and estimated values as evident from the 1:1 line shown in fig. 1. θsat = 0.3773-0.0069 * Clay(%) r = 0.8002 (0.8141)..(1) θfc = 0.5758-0.0050 * Sand(%) r = 0.8806 (0.8971)..(2) θwp = 0.0237 + 0.0043 * Clay(%) r = 0.9015 (0.9042)..(3) θad = -0.0118 + 0.0021 * Clay(%) r = 0.8569 (0.8908)..(4) The figures in parenthesis are r values with all the independent variables taken together. Ksat correlated with bulk density in addition to clay content and was statistically significant at 1% level, the coefficient of determination was low (Fig. 2). This is expected in a property like Ksat which is highly spatially variable. The four equations given above were then used for predicting the values of twenty soils randomly selected from the data base. The main criterion for selection was choosing soils having as wide a range as possible in sand, silt, clay and organic carbon contents. The magnitude of these parameters for the soils selected ranged from 4.0 to 73.3 for sand, 4.0 to 38.0 for silt, and 11.8 to 69.9 for clay. The values predicted by equations 1 to 4 and the corresponding observed values are shown in fig. 3. The values lie close to the 1:1 line with R 2 value of 0.9472, indicating a very close agreement between the observed and predicted values. The magnitude of the intercept was 0.0167 with slope of 0.9448. When WTGROWS was run for simulating crop growth, water use and yield using five years weather data with measured and predicted values of soil water constants as inputs, both yield and evapotranspiration were unaffected. CONCLUSION The above results indicate that simple models based on very routinely measured and easily available soil physical properties can be used for generating the four critical soil water contents required as input in crop growth models having a water balance component like SAHEL/CERES without affecting the performance of the model in predicting crop growth and yield under water limiting conditions. Some approximate estimates of Ksat can also be obtained based on the clay content. 4
BIBLIOGRAPHY Aggarwal, P.K.,Kalra, N., Singh,A.K. and Sinha,S.K. (1994). Analyzing the limitations set by climatic factors, genotype, water and nitrogen availability on productivity of wheat. I. The model description, parametrization and validation. Field Crops Res. 38: 73-91.. Ali,M.H., Chatterjee,R.K. and Biswas,T.D. (1966). Soil moisture tension relationships of some Indian soils. J. Indian Soc. Soil Sci. 14 : 51-62. Biswas,T.D. and Ali,M.H. (1967). Influence of organic carbon and clay content of the soils on the permanent wilting percentage. Indian J. Agric. Sci. 37 : 322-331. Clapp,R.B. and Hornberger,G.M. (1978). Empirical equations for some soil hydraulic properties. Water Resour. Res. 14 : 601-604. Gupta,R.P., Kumar,S. and Singh,T. (1984). Soil management to Increase Crop Production. Consolidated Report 1967-82 of the AICRP on Improvement of Soil Physical conditions to Increase Agricultural Production of Problematic Areas, ICAR, p. 104. Gupta,S.C. and Larson,W.E. (1979). Estimating soil water retention characteristics from particle size distribution, organic matter percent and bulk density. Water Resour. Res. 15 : 1633-1635. Penning devries, F.W.T., Jansen, D.M., ten Berge, H.F.M. and Bakema,A. (1989). Simulation of ecophysiological processes of growth in several annual crops. Simulation Monographs 29, Pudoc, Wageningen, p. 271. Rawls,W.J. and Brakensiek, D.L. (1982). Estimating soil water retention from soil properties. J. Irrig. Drain. Div. ASCE 108 : 166-171. Tsuji,G.Y., Uehara,G. and Balas,S. (1994). DSSAT v3. University of Hawaii, Honolulu, Hawaii, p. 164. Velayutham, M.and Raj, D. (1971). Interrelationship between soil separates and properties of the soils of Tamil Nadu. J. Indian Soc. Soil Sci. 19 : 353-361. Vereecken,H., Maes,J. and Feyen,J.(1990). Estimating unsaturated hydraulic conductivity from easily measured soil properties. Soil Sci. 149 : 12-32. 5
Yadav, B.S., Verma, B.L. and Rama Deo (1995). Water retention and transmission characteristics of soils in command area of North- Western Rajasthan. J. Indian Soc. Soil Sci. 43: 1-5. Keywords : Pedotransfer functions, soil properties, crop simulation models, crop yield, water balance Mots clés : fonction de pédotransfert, propriétés des sols, modèles de culture, rendement, bilan hydrique 6