The implication of spatial interpolated climate data on biophysical modelling in agricultural systems

Abstract

Spatial modelling of agricultural production has been more frequently analysed to assist with regional and long‐term planning. Typically, point‐scale crop models are used to construct these analyses with researchers using various sources of input data, including observed and interpolated gridded climate data. Understanding the implication of data choice on production estimates is crucial to appreciate the consequences of selecting methodological approaches for agricultural model outputs. In this study, we compared site observed climate data and gridded data sets interpolated from site observations that are commonly used the Australian climate data set. We assessed the consequences of the differences in climate variables (biases) to the modelled outputs. Our results showed that the major differences between gridded and observed climate data sets were for rainfall variables. Interpolated gridded data tended to have larger rainfall frequency and smaller rainfall intensity, leading to lower surface runoff and higher soil evaporation that caused less plant water uses and less nitrogen uptake, and ultimately resulted in biases in simulated crop biomass and yield. In addition, the results indicated that agricultural models implemented with gridded data could produce similar overall model outputs as those driven by observed climate data, but can result in more uncertainties in simulated spatial outputs. This is particularly evident in regions with high rainfall. Our results showed that applying agricultural models with observed data and then interpolating these results spatially might be the optimal approach to minimize the biases in production modelling outputs and reduce computing time and storage.

Characterizing spatiotemporal rainfall changes in 1960–2019 for continental Australia

“…Additionally, trends in more frequent and less intense rainfall can impact winter crop growth. For example, storage of sub‐soil moisture over summer may be reduced as frequent, ineffective small rainfall events were lost through evaporation rather than filling the soil profile for subsequent crop use (Liu et al ., 2019; Liu et al ., 2020). In southern NSW and Victoria, winter cropping was the dominant agricultural system but with reduced growing season rainfall and less soil moisture storage these systems will be challenged to maintain future productivity.…”

“…However, using station data instead of gridded data was a deliberate choice as interpolation typically introduced additional uncertainty in rainfall. The gridded data can result in larger rainfall frequency and smaller rainfall intensity when compared with station data (Liu et al ., 2020), which would also have implications for examination of rainfall trends. The high‐quality stations with a high proportion of observed data minimized the data‐source bias interfering to the analysis.…”

A comprehensive, multisource database for hydrometeorological modeling of 14,425 North American watersheds

“…Complexifying the issue, there are now several available data sources for meteorological forcing data 4,5 . Recent work 6,7 has shown that impact model results are influenced by the choice of forcing data. Providing several relevant forcing datasets should therefore be considered important to better understand how uncertainties propagate in the water resources management chain.…”

Silver lining to a climate crisis in multiple prospects for alleviating crop waterlogging under future climates

“…A key reason for using the data from 27 GCMs was to better quantify the spatiotemporal distribution of and variability in precipitation during the growing season. We downscaled the GCM datasets using the NASA/POWER gridded historical weather database 64 . However, previous work has shown that interpolated gridded data tends to be conducive to producing rainfall events that are smaller in quantum but more frequent, which can lead to lower surface runoff and higher soil evaporation 64 .…”