Uncertainties in the appraisal of water availability and consequences for simulated sugarcane yield potentials in São Paulo State, Brazil

Abstract

Summary crop growth models are promising tools for assessing crop yield potentials, but it is necessary to indicate uncertainties in model calculations that result from simplifying assumptions and errors/uncertainties in model input data and parameter values. The present study attempts to do this for soil factors that affect the water balance of land-use systems with sugarcane ( Saccharum spp. L.) on strongly weathered soils (mainly Ferralsols) in São Paulo State, Brazil. Propagation of uncertainties were assessed for: (1) water retention characteristics of the soil; (2) effective rooting depth as limited by acid layers with large Al saturation (percentage of effective cation exchange capacity occupied by Al 3+) and (3) possible errors originating from the methods used to calculate water withdrawal by roots. Probability density functions were constructed for the maximum effective rooting depth (RDM, cm), estimated on the basis of the soils’ Al saturation at 60–80 cm depth; and for ‘available’ water retained 2 days after field saturation ( θ 2d−1.5 MPa , cm 3 cm −3). These functions were used in a Monte Carlo analysis. For each of 81 sugarcane crops, 100 independent drawings of θ 2d−1.5 MPa and RDM were input to a summary crop growth model (PS123) which was adapted to sugarcane and available soil data. Within-harvest variation of calculated yield potentials was compared with variance of historical yield records and that of yield potentials calculated with average values of θ 2d−1.5 MPa or RDM. Two methods were used to calculate soil water uptake by roots: (1) with water uptake conditioned by the water content of the soil layer(s) where it is most readily available and (2) with water uptake determined by the average soil water content throughout the rooting depth. Results show that the pooled (within-harvest) standard deviation of calculated sugarcane yields caused by the two ‘errors’ (up to 4436 kg ha −1) is large in comparison to the pooled standard deviation of recorded yields from neighbouring plots within a field (2226 kg ha −1) as well as to the root mean square of residuals from regression between simulated yield potentials and actual yield records (4449 kg ha −1). Simulated yield variances among the 100 drawings are strongly non-uniform. Uncertainty in Al-limited RDM caused greater variation in calculated yields than uncertainties in θ 2d−1.5 MPa Yield estimates calculated with method 2 for root water uptake were systematically smaller (average: 1500 kg ha −1) than with method 1. It is concluded that different appraisals and common errors/uncertainties in input data for water balance calculations of crop models may cause substantial deviation in computed yield potentials. Results suggest that future research on crop water relations in south-east Brazil should give special attention to in situ determination of water withdrawal by roots as related to soil Al concentration.