Machine Learning Based Climate Projections For Sustainable Potato Production In Prince Edward Island

Download or read book Machine Learning Based Climate Projections for Sustainable Potato Production in Prince Edward Island written by Junaid Maqsood and published by . This book was released on 2021 with total page 0 pages. Available in PDF, EPUB and Kindle. Book excerpt: Prince Edward Island (PEI) is the largest potato-producing province in Canada, and most of its croplands are rainfed. Climate change impacts all fields of life, including agriculture. Thus, there is a need to understand better the historical variations and future projections of climate change and its patterns for PEI. Climate change and its impacts on potato tuber yield have been evaluated in this thesis under three objectives. For the first objective, twenty climate extreme indices were computed with the help of ClimPACT2 software for 30 years (1989-2018) to assess their impacts on the potato tuber yield. The average of daily mean temperature, mean daily minimum temperature (TNm), and the occurrence of continuous dry days (CDD) significantly increased, while daily temperature range (DTR), frost days, cold days, cold nights, and warmest days (TXx) showed decreasing trends during the potato growing seasons (May-October) for the past three decades. The principal component analysis results showed that DTR, TXx, CDD, and TNm were the main indices, defining ~39% variations in tuber yield. However, DTR, TXx, CDD, and TNm individual contributions to the variations in tuber yield were recorded to be 21, 19, 16, and 4%, respectively. For the second objective, the Hargreaves method was used to calculate reference evapotranspiration (ET0) for western, central, and eastern parts of PEI using their two input parameters: daily maximum temperature (Tmax) and daily minimum temperature (Tmin). The Tmax and Tmin from the Canadian Earth System Model Second Generation (CanESM2) were downscaled with the help of statistical downscaling model (SDSM) for three future periods, i.e., the 2020s (2011-2040), 2050s (2041-2070), and 2080s (2071-2100) under three representative concentration pathways (RCP's) including RCP2.6, 4.5, and 8.5. Temporally, there were major changes in Tmax, Tmin, and ET0 for the 2080s under RCP8.5. In the next steps, a one-dimensional convolutional neural network (1D-CNN), long-short term memory (LSTM), and multilayer perceptron (MLP) were used for estimating ET0 for historical and future periods. High coefficient of correlation (r > 0.95) values for both calibration and validation periods showed the potential of the artificial neural networks in ET0 estimation. For the third objective, SDSM, MLP, random forest (RF), and support vector regression (SVR) were used to downscale Tmax, Tmin, and precipitation at eight meteorological stations located in PEI. The comparison results depicted the better performance of MLP to downscale the climatic parameters (Tmax, Tmin, and precipitation). Therefore, the MLP algorithm was used to project the climatic parameters for the future period (2006-2100) under RCP2.6, 4.5, and 8.5. The linear scaling method was used to reduce the biases in the projected data and get real results. The results of the analysis of the data from the annual and the growing season showed that Tmax and Tmin continually increased in the future under all the RCPs, but maximum increment was noticed under RCP8.5. The spatial patterns of average annual precipitation in the growing season showed high, moderate, and low precipitation at the PEI's eastern, central, and western parts for the historical (1976-2003) and future periods. This study will help the decision-makers and farmers to understand better the variations and patterns of the climatic parameters for the historical and future periods in relation to agriculture. The results may also help to develop irrigation scheduling in response to climate change to meet sustainable development goals.