Combining Crop and Climate Models

The spatial and temporal scales of numerical climate and crop simulation models are not the same. General circulation models operate on grid sizes of about 200 km, but crop simulation models are designed to use information on climate, soil parameters and management practices at the scale of a field. Crop simulation models operate on daily time-steps and use some seasonal information. GCMs include processes resolved at a range of time-steps, but daily output is not always archived. This mismatch in the spatial and temporal scales of climate model output and crop model input needs to be resolved in order to reduce the uncertainties of seasonal crop forecasts.

A number of approaches to improving the skill of seasonal crop forecasts have been proposed (see review by Hansen et al. 2006). One of these has involved the development of a combined crop and climate forecasting system (Challinor et al. 2003). A number of discrete stages in the development of a combined crop and climate forecasting system have been defined (Fig. 3.1). The first is the definition of the spatial scale of relationships between crop productivity and climate using observations. Crop models that use climate information as input implicitly assume that there is a strong relationship between climate and crop growth, development and yield. Variability in large-scale climate processes such as the Southern Oscillation has been correlated with yields of four crops in Australia (Nicholls 1985) and with maize in Zimbabwe (Cane et al. 1994). Analyses of historical crop data show that the variability in yields due to climate differs with location, from a small climate signal in the temperate UK wheat

Fig. 3.1. The stages of development of a combined crop and climate forecasting systems (redrawn from Challinor et al. 2003)

Fig. 3.1. The stages of development of a combined crop and climate forecasting systems (redrawn from Challinor et al. 2003)

crop (Landau et al. 1998) to more than half the variability of the major crops growing in India attributed to monsoon rainfall (Krishna Kumar 2004). Challinor et al. (2003) examined the spatial scale of the relationship between crops and climate. They found a coherent spatial and temporal pattern between the yield of groundnuts and seasonal rainfall across India for 1966-1990 on the scale of subdivisions (irregular polygons of about 130-480 km). Furthermore, this pattern was closely correlated with the smaller-scale pattern of crop yields at a district level (an average linear scale of 98 km), and with the 850 hPa large-scale circulation pattern. These large-scale correlations between crop productivity and climate therefore established the basis for combining GCM output directly with crop model output in that region (Challinor et al. 2003).

The simulation of crop productivity over a large area needs some simplification of the crop simulation process. A complete set of field-scale inputs will not be readily available over areas of countries and regions, and the grid size will encompass spatial heterogeneity in parameters that describe soils, crop genotype and management practices. Reduced-form crop models (for example, Brooks et al. 2001) and statistical models (such as Landau et al. 2000; Baez-Gonzalez et al. 2005) have been developed, and shown to have predictive skill over large areas. Challinor et al. (2004) sought to maintain a process-based approach in a large area crop model in order to simulate the effects on the crop of short time-step events such as intra-seasonal variability in rainfall, and high temperatures. They proposed a general large area model (GLAM) for annual crops and demonstrated good forecasting skill of the model in a hindcast of the groundnut crop aggregated to all India for 1966-1990.

There is increasing evidence from crop experiments that short-term climate events of only a few days duration can severely impact crop productivity if they coincide with a sensitive phase of crop growth. One example is the occurrence of high temperatures near to the time of crop flowering (Wheeler et al. 2000; Fig. 3.2). The nature of crop response when these climate thresholds are exceeded will be a vital part of the impact of climate change on crop productivity in some regions (Wheeler et al. 1996; Vara Prasad et al. 2002). Therefore, successful prediction of crop productivity by large area crop models on both seasonal and decadal timescales needs robust simulations of the

Fig. 3.2. The effect of a 1-day high temperature event on the fruit/seed set of groundnut plants grown in controlled environments (redrawn from Vara Prasad et al. 2001)

Fig. 3.2. The effect of a 1-day high temperature event on the fruit/seed set of groundnut plants grown in controlled environments (redrawn from Vara Prasad et al. 2001)

effects of short-term variability in climate on the crop. The high temperature response shown in Fig. 3.2 has been incorporated into GLAM to give a GLAM-HTS (high temperature stress) model version (Challinor et al. 2005c). A similar response of rice to high temperature is also represented in the ORYZA crop model (Matthews et al.1995). Such models open up opportunities to examine how short-term, sub-seasonal variability in temperature will affect crop productivity.

Rainfall distribution within a growing season can affect crop productivity independent of the seasonal mean. For example, two years with similar rainfall amounts of 394 mm (1975) and 389 mm (1981) during the growing season of groundnut crops in Gujarat, India, are shown in Fig. 3.3. Rainfall in 1975 was evenly spread throughout the season and a yield of 1360 kg ha-1 was attained (Fig. 3.3a). In 1981, there was a break in the monsoon rains during part of the period of grain filling of the crop (55-80 days after planting, Fig. 3.3b). Observed yields were reduced by 34% to 901 kg ha-1. The GLAM crop model simulated a 20% drop in yield in 1981 compared with 1975 (Challinor et al. 2004). Thus, the impact of sub-seasonal variation in rainfall on crop yield was reproduced by the large area crop model.

Was this article helpful?

0 0

Post a comment