Soil temperature is an important factor within the climate system. Changes of trends in soil temperature and analysis of vulnerability due to heat stress can provide useful information on climate change. In this paper, the soil temperature regime was analyzed on seasonal and annual scales at depths of 2, 5, 10, 20, 30, and 50 cm at 26 sites in Croatia. Trends of maximal, mean, and minimal soil temperatures were analyzed in the periods 1961-2010 and 1981-2010. Duration of extreme soil temperatures and vulnerability due to high or low soil temperatures in the recent standard period 1981-2010 was compared with the reference climate period 1961-1990. The results show a general warming in all seasons and depths for maximal and mean temperatures in both observed periods, while only at some locations for minimal soil temperature. Warming is more pronounced in the eastern and coastal parts of Croatia in the surface layers, especially in the spring and summer season in the second period. Significant trends of maximal, minimal, and mean soil temperature in both observed periods range from 2.3 to 6.6°C/decade, from −1.0 to 1.3°C/decade, and from 0.1 to 2.5°C/decade, respectively. The highest vulnerability due to heat stress at 35°C is noted in the upper soil layers of the coastal area in both observed periods. The mountainous and northwestern parts of Croatia at surface soil layers are the most vulnerable due to low soil temperature below 0°C. Vulnerability due to high or low soil temperature decreases with depth.

<p dir="ltr">Seasonal climate predictions are becoming important to increase preparedness and adapt agricultural decision making before and within the growing season. Evapotranspiration represents a major component of the water cycle and agricultural water balance, and is thus of key relevance also for assessment of crop growth and irrigation water needs. This study investigates the comparison of different empirical methods for reference evapotranspiration calculation with the reference Penman Monteith method, implemented in the AquaCrop model, over different climate zones of Croatia using national high resolution high quality meteorological dataset in comparison with EOBS and AgERA5 datasets. This research aims to address the appropriateness of using the method with limited data to reduce uncertainty in crop yield modelling with seasonal climate forecasts. </p> <p dir="ltr">The motivation for the research stems from the need to use a reduced number of parameters for ETo calculation, which as such enters the crop yield model. Estimation of seasonal predictions of ETo relies on the skill of prediction of multiple meteorological variables; in the case of Penman Monteith, the estimates would rely on the skill of temperature, vapour pressure, winds speed and global solar radiation prediction. Reducing the number of input variables for estimating the reference evapotranspiration might reduce the overall reliance on the limited skill of seasonal predictions of above-mentioned meteorological variables. </p> <p dir="ltr">In testing the usefulness of the methods, the interpolated field of the national data network, EOBS dataset, and AgERA5 for the period from 1981 to 2020 are used. Preliminary results show that for the continental area of Croatia the Trajkovic method is satisfactory for ETo calculation, and the EOBS dataset provides better results than AgERA5 when compared to high resolution national dataset. In addition, we analyse the ETo seasonal prediction skill by comparing both methods, original Penman Monteith and Trajkovic method with reduced number of input variables. This will give us an important insight on the relevance of the meteorological dataset choice as well as the ET0 method selection for seasonal prediction purposes. Since there is no comprehensive research on the reference crop evapotranspiration in Croatia, this research fills the gap of knowledge at the local level, but also contributes to the global picture</p>