Regulatory mechanisms in agroecosystems: A retrospective and forecast of spatial and temporal dynamics of precipitation as a factor of crop yield

The research tested the hypothesis that the climate of the studied area has the property of spatial and temporal regularity, and that this regularity is hierarchically organized, which makes it possible to predict the state of the climate in the coming decades. The practical aspect of the information obtained is the assessment of possible prospects for changes in the yields of the most common crops in the region. The spatial variability of precipitation between the years 1960 and 2023, soil properties and landscape cover structure were investigated within 10 administrative regions of northern and northwestern Ukraine. This region covers the Polissia and Forest-Steppe geographical zones. The MEM spatial variables were able to explain 95.1% of the variability in precipitation. ANOVA revealed that 8 canonical axes were statistically significant. The contribution of the spatial MEM variables to the explanation of the canonical axes is different, which allows us to identify the hierarchical structure of variability of the main spatial precipitation patterns in the region. The RDA1 and RDA2 axes represent the large-scale component of precipitation variability. RDA1 indicates the differentiation of precipitation patterns in the meridional direction with the allocation of the eastern and western sectors of the region. The canonical axes denoting the main spatial patterns of precipitation variability correlated with soil properties and land cover types. RDA1 did not correlate with soil properties, but had a positive correlation with the proportion of broadleaf forests and mosaic of herbaceous cover and shrubs in the landscape cover. This axis had a negative correlation with the proportion of agricultural land. RDA2 was positively correlated with soil organic matter and sand content, but negatively correlated with clay and silt content. This axis increased with an increase in the proportion of broadleaf, coniferous or mixed forests or a mosaic of herbaceous vegetation and shrubs in the landscape cover structure. RDA2 decreased with an increase in the proportion of agricultural crops or sparse vegetation cover. RDA3 was independent of soil organic matter content, but positively correlated with clay and silt content and negatively correlated with sand content. This axis was positively correlated with the proportion of agricultural area, the mosaic of herbaceous vegetation and shrubs, and negatively correlated with the proportion of coniferous or mixed forests. RDA4 was positively correlated with soil organic matter content and negatively correlated with soil silt content. This axis increased with increasing proportions of rainfed crops and sparse vegetation cover, but decreased with increasing proportions of herbaceous cover, coniferous and mixed forests. RDA5 was positively correlated with organic matter and silt content, but negatively correlated with sand content. This axis increased with increasing proportions of mosaic with crops, but decreased with increasing proportions of coniferous and mixed forests. RDA6 was positively correlated with silt content but negatively correlated with sand content. This axis increased with increasing proportions of agricultural crops, but decreased with increasing proportions of broadleaf or mixed forests. RDA7 was positively correlated with silt and clay content, but negatively correlated with organic matter and sand content. This axis was positively correlated with the proportion of agricultural land and negatively correlated with the proportion of broadleaf, coniferous and mixed forests. RDA8 was positively correlated with the silt content of the soil. This axis was positively correlated with the proportion of agricultural land and negatively correlated with the proportion of coniferous and mixed forests. The temporal modelling of precipitation dynamics over more than 60 years can be carried out using eight AEM predictors, which represent temporal patterns of different frequencies and variable amplitudes over time. If we assume that the established oscillatory dynamics will continue in the coming decades, then these AEM predictors can be extended for the time of interest and a regression model can be used to obtain a forecast of precipitation dynamics in the near future. The forecast indicates a downward trend in precipitation, mainly in areas with the most developed agriculture.