No abstract

Shortlisting ecologically adaptable plant species can be a starting point for agricultural diversification projects. We propose a rapid assessment framework based on an ecological model that can accelerate the evaluation of options for sustainable crop diversification. To test the new model, expert-defined and widely available crop requirement data were combined with more than 100,000 occurrence data for 40 crops of different types (cereals, legumes, vegetables, fruits, and tubers/roots). Soil pH, texture, and depth to bedrock data were obtained and harmonised based on the optimal rooting depths of each crop. Global baseline temperature and rainfall data were used to extract averages at each location. To evaluate the ability of the method to capture intraspecies variation, a test was performed using more than 1000 accession records of bambara groundnut (Vigna subterranea (L.) Verdc.) as an exemplar underutilised crop. Results showed that a suitability index based on soil pH and an index that combines the thermal suitability moderated by the soil pH, texture, and depth suitability have the potential to predict crop adaptability. We show that the proposed method can be combined with traditional land use and crop models to evaluate diversification options for sustainable land and agrobiodiversity resources management.

Evidence based crop diversification requires modelling for crops that are currently neglected or underutilised. Crop model calibration is a lengthy and resource consuming effort that is typically done for a particular variety or a set of varieties of a crop. Whilst calibration data are widely available for major crops, such data are rarely available for underutilised crops due to limited funding for detailed field data collection and model calibration. Subsequently, the lack of evidence on their performance will lead to the lack of interest from the policy and regulatory communities to include these crops in the agricultural development plans. In order to motivate further research into the use of state of the art techniques in modelling for less known crops, we have developed and validated an ideotyping technique that approximates the crop modelling parameters based on already calibrated crops of different lineage. The method has been successfully tested for hemp ( Cannabis sativa L.) based on a well-known crop model. In this paper we present the method and provide an impetus on the way forward to further develop such methods for modelling the performance of minor crops and their varieties. The approach works based on modelling the performance of hemp using the knowledge from an existing model that was developed for sugar cane. The customisation uses one of the most prominent models (AquaCrop) to approximate growth coefficients for hemp ( Cannabis sativa L.). A sequential procedure was used to approximate the phenological stages in the growth model that performs well in the calibration and validation steps.

Data from global soil databases are increasingly used for crop modelling, but the impact of such data on simulated crop yield has not been not extensively studied. Accurate yield estimation is particularly useful for yield mapping and crop diversification planning. In this article, available soil profile data across Sri Lanka were harmonised and compared with the data from two global soil databases (Soilgrids and Openlandmap). Their impact on simulated crop (rice) yield was studied using a pre-calibrated Agricultural Production Systems Simulator (APSIM) as an exemplar model. To identify the most sensitive soil parameters, a global sensitivity analysis was performed for all parameters across three datasets. Different soil parameters in both global datasets showed significantly (p < 0.05) lower and higher values than observed values. However, simulated rice yields using global data were significantly (p < 0.05) higher than from observed soil. Due to the relatively lower sensitivity to the yield, all parameters except soil texture and bulk density can still be supplied from global databases when observed data are not available. To facilitate the wider application of digital soil data for yield simulations, particularly for neglected and underutilised crops, nation-wide soil maps for 9 parameters up to 100 cm depth were generated and made available online.