Prediction of carbon dioxide (CO2) emissions from agricultural soil is vital for efficient and strategic mitigating practices and achieving climate smart agriculture. This study aimed to evaluate the ability of two machine learning algorithms [gradient boosting regression (GBR), support vector regression (SVR)], and two deep learning algorithms [feedforward neural network (FNN) and convolutional neural network (CNN)] in predicting CO2 emissions from Maize fields in two agroclimatic regions i.e., continental (Debrecen-Hungary), and semi-arid (Karaj-Iran). This research developed three scenarios for predicting CO2. Each scenario is developed by a combination between input variables [i.e., soil temperature (Δ), soil moisture (θ), date of measurement (SD), soil management (SM)] [i.e., SC1: (SM + Δ + θ), SC2: (SM + Δ), SC3: (SM + θ)]. Results showed that the average CO2 emission from Debrecen was 138.78 ± 72.04 ppm (n = 36), while the average from Karaj was 478.98 ± 174.22 ppm (n = 36). Performance evaluation results of train set revealed that high prediction accuracy is achieved by GBR in SC1 with the highest R² = 0.8778, and lowest root mean squared error (RMSE) = 72.05, followed by GBR in SC3. Overall, the performance MDLM is ranked as GBR > FNN > CNN > SVR. In testing phase, the highest prediction accuracy was achieved by FNN in SC1 with R² = 0.918, and RMSE = 67.75, followed by FNN in SC3, and GBR in SC1 (R² = 0.887, RMSE = 79.881). The performance of MDLM ranked as FNN > GRB > CNN > SVR. The findings of the research provide insights into agricultural management strategies, enabling stakeholders to work towards a more sustainable and climate-resilient future in agriculture.

Water availability for agricultural practices is dynamically influenced by climatic variables, particularly droughts. Consequently, the assessment of drought events is directly related to the strategic water management in the agricultural sector. The application of machine learning (ML) algorithms in different scenarios of climatic variables is a new approach that needs to be evaluated. In this context, the current research aims to forecast short-term drought i.e., SPI-3 from different climatic predictors under historical (1901–2020) and future (2021–2100) climatic scenarios employing machine learning (bagging (BG), random forest (RF), decision table (DT), and M5P) algorithms in Hungary, Central Europe. Three meteorological stations namely, Budapest (BD) (central Hungary), Szeged (SZ) (east south Hungary), and Szombathely (SzO) (west Hungary) were selected to forecast short-term agriculture drought i.e., Standardized Precipitation Index (SPI-3) in the long run. For this purpose, the ensemble means of three global circulation models GCMs from CMIP6 are being used to get the projected (2021–2100) time series of climatic indicators (i.e., rainfall R, mean temperature T, maximum temperature Tmax, and minimum temperature Tmin under two scenarios of socioeconomic pathways (SSP2-4.5 and SSP4-6.0). The results of this study revealed more severe to extreme drought events in past decades, which are projected to increase in the near future (2021–2040). Man-Kendall test (Tau) along with Sen's slope (SS) also revealed an increasing trend of SPI-3 drought in the historical period with Tau = −0.2, SS = −0.05, and near future with Tau = −0.12, SS = −0.09 in SSP2-4.5 and Tau = −0.1, SS = −0.08 in SSP4-6.0. Implementation of ML algorithms in three scenarios: SC1 (R + T + Tmax + Tmin), SC2 (R), and SC3 (R + T)) at the BD station revealed RF-SC3 with the lowest RMSE RFSC3-TR = 0.33, and the highest NSE RFSC3-TR = 0.89 performed best for forecasting SPI-3 on historical dataset. Hence, the best selected RF-SC3 was implemented on the remaining two stations (SZ and SzO) to forecast SPI-3 from 1901 to 2100 under SSP2-4.5 and SSP4-6.0. Interestingly, RF-SC3 forecasted the SPI-3 under SSP2-4.5, with the lowest RMSE = 0.34 and NSE = 0.88 at SZ and RMSE = 0.34 and NSE = 0.87 at SzO station for SSP2-4.5. Hence, our research findings recommend using SSP2-4.5, to provide more accurate drought predictions from R + T for future projections. This could foster a gradual shift towards sustainability and improve water management resources. However, concrete strategic plans are still needed to mitigate the negative impacts of the projected extreme drought events in 2028, 2030, 2031, and 2034. Finally, the validation of RF for short-term drought prediction on a large historical dataset makes it significant for use in other drought studies and facilitates decision making for future disaster management strategies.