Impacts of climate change on the potato (Solanum Tuberosum L.) productivity in Tasmania, Australia and Kenya

This study assessed the potential impact of climate change on potato production both in Australia (Tasmania) and Kenya. Potato is an important commodity in both regions but there is little information about how this crop will respond to projected changes in climate compared to other regions. Previous to this doctoral study, APSIM-potato had only been tested and calibrated with a small number of datasets from a long-term experiment conducted in Lincoln, New Zealand with 'Russet Burbank' and its application to productivity modelling required further parameterisation and evaluation. A first step in this study was therefore to parameterise and evaluate the Agricultural Production System sIMulator (APSIM-potato) model under both Tasmanian and Kenyan potato growing conditions. Throughout this thesis the word parameterisation refers to the process of determining a set of parameter values deemed suitable for model use in a specific study area, and evaluation as the process of assessing the level of precision and accuracy of a model in reproducing observed data using performance measures and statistical values. Four on-farm monitoring plots located on different farms were established in North-West Tasmania within well‚Äö-managed potato fields grown during the 2012/13 cropping season. 'Russet Burbank' cultivar was planted at two sites and 'Moonlight' at the other two sites. In Kenya, experiments were conducted at Kabete, Kiambu County during the short rains (SR2013) and in the long rains (LR2014). The design for the SR2013 experiment was a split-plot with two water levels (supplementary irrigation and rain-fed) as the main plot factor and three genotypes as the sub-plot factor, with four replications. A randomized complete block design was used in the LR2014 experiment, with three nitrogen levels (23, 63 and 104 kg N/ha, hereafter referred to as N23, N63 and N104 treatment levels) and four replicates. Measured soil, weather and crop datasets for 'Russet Burbank' and 'Moonlight' in Tasmania, and for 'Unica', CIP 300046.22 and 'Shangi' in Kenya were used to parameterise and evaluate the model. In both Tasmania and Kenya, the model adequately captured the phenology and the partitioning of assimilates to the tuber state variable over time, with a good index of agreement using a Normalized Root Mean Squared Error (N-RMSE) and Modelling Efficiency (EF). In Tasmania, measured mean Tuber Dry Matter (TDM) was 17 t ha\\(^{-1}\\) for 'Russet Burbank' compared to a simulated value of 20 t ha\\(^{-1}\\). N-RMSE values between observed and simulated TDM ranged between 10 to 20%, with a mean of 16.3% for 'Russet Burbank and 14.5% for 'Moonlight', and a mean EF of 1.0 for both cultivars. For 'Moonlight' the mean simulated TDM value was 16.0 t ha\\(^{-1}\\) compared to the measured value of 15.1 t ha\\(^{-1}\\). Similarly, prediction of phenology and tuber N-uptake was good: respectively a mean N-RMSE value of 25.7% and 20.9% for 'Russet Burbank', 24.2% and 32.7% for 'Moonlight'. However, prediction of other parameters (leaf and stem dry biomass and LAI) were poor with N-RMSE values ranging from 27.6 to 40.8% for 'Russet Burbank', and 20.7 to 48.2% for 'Moonlight'. In Kenya, the model predicted TDM yield with good precision, providing a mean N-RMSE of 18.4% for SR2013 and 28.7% for LR2014, and a mean EF of 0.9 for both seasons. Similarly, prediction of phenology was good, the model providing a mean N-RMSE value of 20.8% and an EF value of 0.8 for both SR2013 and LR2014. In the SR2013 experiment, the measured TDM across the three cultivars under rain-fed conditions was 3.8 ¬¨¬±0.2 t ha\\(^{-1}\\) compared to simulated value of 4.4 t ha\\(^{-1}\\). With supplementary irrigation, the observed value was 6.2 ¬¨¬±0.2 t ha\\(^{-1}\\), close to the simulated value of 6.3 t ha\\(^{-1}\\). In the LR2014 experiment, when pooled across nitrogen levels, the model underestimated TDM, providing a mean simulated TDM at 6.6 t ha\\(^{-1}\\) against the measured value of 7.7 ¬¨¬±0.4 t ha\\(^{-1}\\). In contrast, the index of agreement between simulated and observed above ground biomass was generally low (a mean N-RMSE value of 37.2% in the LR2014 experiment and 47.5% in the SR2013 experiment, EF values ranging from -0.3 to 0.3 in the SR2013 and -0.5 to 0.4 for the long rains). The simulation results provide a database for further testing of the model and this work provides future users with a foundation to further improve the model. While the model accurately predicts plant phenology and TDM, modification of other key crop specific parameters are still needed to improve its accuracy when simulating the development of other plant organs. Further refinement of the model will require collection of long-term field crop data. The model's ability to realistically simulate potato phenology and TDM provided a sound basis to investigate the potential impacts of climate change on potato productivity with confidence. The calibrated APSIM-potato model was used to quantify the potential impact of future climate scenarios on potato productivity in the two contrasting environments of Tasmania and Kenya. Data used in the model to simulate future climates included dynamically downscaled bias-corrected climate projections for Tasmania (Climate Futures Tasmania), and an ensemble of climate projections under the CORDEX‚Äö-Africa initiative for Kenya. Across the three potato growing sites studied temperature projections indicate a 1.2 ¬¨‚àûC increase at each site for maximum temperature and 1.3 ¬¨‚àûC for minimum temperature by 2050. By 2085, a 2.4 ¬¨‚àûC increase is projected for maximum temperature and for minimum temperature; a 2.6 ¬¨‚àûC increase is projected. Annual rainfall is projected to increase across the study sites relative to the baseline period by 6.1%, 4.5%, and 9.0% at Cressy, Forthside and Scottsdale respectively by 2085. Similarly, the coefficient of variation (CV) of annual and seasonal rainfall is projected to increase by 2% at both Scottsdale and Forthside above the baseline value of 13% and 14% respectively and by 3% at Cressy above the baseline value of 14%. Annual and seasonal rainfall intensity is projected to increase from the baseline to 2085. Whilst temperature is projected to increase in the Tasmanian potato growing regions, the duration at which the crop is exposed to temperatures outside the crops optimal range is negligible. Consequently, climate change will have little influence on projected future multi-model ensemble median (MME) tuber yield under current farmer practice (based on the last five years 2012 to 2016). There is a steady increase in the rate of growing degree day (GDD) accumulation from planting to harvest; 4.8% by 2050 and 12.3% by 2085 relative to the baseline period of 1981-2010 across the three sites. This increased rate in GDD accumulation is projected to shorten the time to crop maturity against the baseline period by 10 days in 2050, and 15 days by 2085. Shortening of the duration to crop maturity could potentially translate to savings in irrigation and the amounts of pesticide used. However Tasmania is free of some of the major potato pests and diseases. This situation could change as shifts in climate encourage the establishment of pests and diseases or if strict biosecurity regulations are not maintained. Extreme events leading to waterlogging could destroy crops. Thus it is difficult to predict what a potential shortening of 10 - 15 days will imply for potato production in Tasmania. In Kenya, marked inter-annual and inter-seasonal rainfall variability is projected for the two sites investigated, with no clear trend throughout the 21st century. Annual rainfall is projected to reduce at Bomet by 46.1% by 2050 and 42.3% by 2085 while the opposite is projected at Kabete with a 1.6% increase by 2050 and by 15.9% by 2085. In both sites, a reduction in annual rainfall is projected for LR and an increase is projected for SR. At Bomet rainfall intensity and number of rain days are projected to decrease and the opposite is projected at Kabete. Mean maximum and minimum temperature are projected to increase across the potato growing areas investigated. The projected increase is higher at Kabete, the lower altitude region with a mean increase of a 2.4 ¬¨‚àûC (Tmax) and 2.6 (Tmin) by mid-century and 4.1 ¬¨‚àûC (Tmax) and 4.4 ¬¨‚àûC (Tmin) increase by 2085 compared to 0.6 ¬¨‚àûC (Tmax) and 1.3 ¬¨‚àûC (Tmin) by mid-century and 2.3 ¬¨‚àûC (Tmax) and 3.1 ¬¨‚àûC (Tmin) in Bomet. In terms of generating future climatic data, refinement of the projected data and bias-adjustment is recommended for Kenya as there were large disagreements among the projected datasets generated by the different GCMs used in the study to generate future climate data for the two study sites. The modelling predicted that Kabete will experience an increasing number of hot days (maximum daily temperature ‚Äöv¢‚Ä¢ 24 ¬¨‚àûC but <34 ¬¨‚àûC) as the century progresses, thus future potato production may be less viable here than in Bomet where mean daily temperatures are within the ideal range throughout the century (mean of 17 ¬¨‚àûC, 18 ¬¨‚àûC and 19 ¬¨‚àûC for the baseline, 2050 and 2085 compared to 21 ¬¨‚àûC, 24 ¬¨‚àûC and 26 ¬¨‚àûC for Kabete). At Bomet, simulated potato yields were less variable than at Kabete though simulation results indicate an increase throughout the century in both sites. As temperatures in Bomet are likely to be within the optimal range for potatoes tuber yield will most likely be driven by rainfall amount, and more importantly, distribution. In Kabete, where temperatures are predicted to be intermittently above the optimal range, temperature and rainfall are the key drivers in both determining and reducing potato tuber yield. Importantly, benefits of increasing CO\\(_2\\) concentrations counteracted the negative impacts of elevated temperatures and contributed to the projected positive impacts of tuber yield in a warmer future climate pa...