Determining the Optimal Planting Time for Agricultural Crops in Northern Kazakhstan

Abstract

Accurate planting-date selection is a critical determinant of crop yield in the semi-arid northern Kazakhstan, where growing-season rainfall is highly variable. This study presents a machine-learning pipeline that recommends optimal sowing dates for six crops grown in the Akmola region-wheat, barley, pea, lentil, canola, and flax-using freely available satellite weather and soil databases. The absence of publicly available field-level rotation records in Kazakhstan necessitated construction of a synthetic training dataset:$\mathbf{5 7, 1 2 4}$growing scenarios generated by pairing 20 published rotation sequences with 35 years of NASA POWER climate data and ISRIC SoilGrids soil profiles, with yield multipliers tuned against official district harvest statistics. Among the eight regression algorithms benchmarked, tree-based models reached$\mathbf{R}^{\mathbf{2}}=0.91-0.92$on the held-out synthetic test set; linear regressors did not exceed 0.85. Random Forest feature weights placed growing-season precipitation first at 49%, with crop identity second (14%) and relative humidity third (8%). Predictions were then tested against 21 years of official Shortandy district wheat yields (2004-2024) under leave-one-year-out crossvalidation, giving$\mathrm{R}^{2}=0.289$(Pearson$\mathrm{r}=0.546, \mathrm{p}=0.010$) with no real yield data seen during training. Drought years 2010 and 2023 and high-yield years 2009, 2018, and 2024 were all ranked correctly. For four of the six crops, the recommended sowing datespea: April 30; barley: May 5; wheat: May 15; flax: May 25-fell within established agronomic windows, providing initial evidence that the optimizer produces agronomically plausible outputs.

More by Nurzhamal Kashkimbayeva

• Evidence-Linked Measurement of Indirect Prompt Injection in Web Security Agents: Paired Trials and Application-Specific Evidence AuditsTechnologies · 2026
• Kazakhstan Agri Recommender — Data Collection PipelineZenodo · 2026
• Replication Data and Code for “Added Predictive Value and Temporal Transferability of MODIS and ERA5-Land Data for District-Level Wheat Yield Forecasting”Harvard Dataverse · 2026
• Multimodal Data Fusion for District-Scale Crop Yield Prediction in Northern KazakhstanIEEE International Conference on Smart Information Systems and Technologies · 2026