Conservation tillage (CT) is of primary importance in food security, soil conservation, and sustainable development, even though its comprehensive effects on runoff (RO) and soil loss (SL) are still not fully understood. In 2004, a field-scale study was launched in southwest Hungary to investigate the long-term (16 years) effects of CT on RO, SL and soil, under a warm-summer humid continental climate. Four, especially large, 1200 m² plots (2 ploughing tillage (PT) and 2 CT) were established, using a special, two-channel collection system. By the end of the study period, significantly higher water-stable aggregates (PT: 20.0 %, CT: 30.4 %), higher soil organic matter (PT: 1.4 %, CT: 1.9 %), greater earthworm abundance (4.9 times that in PT plots) was recorded on the CT plots. Conservation tillage decreased surface RO by 75 % and SL by 95 %. The difference between PT and CT was significant for mean annual soil erosion, with values of 2.8 t ha⁻¹ and 0.2 t ha^-1, respectively. The exceedance of extreme precipitation events was <2%, but their impact on soil erosion was extraordinarily high. Runoff and SL were predicted for the whole dataset, and for the sub-dataset of maize culture, in four separate Random Forest (RF) model developments. The often used linear models are not suitable for predicting soil erosion, hence a more robust, non-parametric, advanced method of classification tree analysis was used. The RF classification method was able to predict erosion risk. For the maize sub-dataset, the RF model best predicted the extreme events, followed by the no-runoff category. The sensitivity of the groups with the highest and lowest risk all exceeded 82 % for SL and 64 % for RO. Tillage type was the most important factor. This long-term study demonstrated that the use of CT enabled the maintenance of a major fraction of precipitation on arable land, and consequently, soil loss remained an order of magnitude lower than its tolerable value. The RF method is suitable for modelling RO and SL. In future, the integration of more datasets in modelling would considerably improve the precision and accuracy of prediction of RO and SL.

尽管耕作耕作对径流（RO）和土壤流失（SL）的综合影响尚不完全清楚，但它在粮食安全，土壤保护和可持续发展中至关重要。2004年，在匈牙利西南部展开了一项实地研究，以研究在温暖夏季湿润的大陆性气候下CT对RO，SL和土壤的长期影响（16年）。四个，尤其大，1200 m ²使用特殊的两通道采集系统，建立样地（2耕作（PT）和2 CT）。到研究期末，水稳性团聚体（PT：20.0％，CT：30.4％），土壤有机质（PT：1.4％，CT：1.9％）显着提高，（丰度更高（（为4.9倍）。 PT图）记录在CT图上。保护性耕作使表面RO降低了75％，SL降低了95％。PT和CT之间的差异对于年均土壤侵蚀具有显着性，其值分别为2.8 t ha ^-1和0.2 t ha ^-1， 分别。极端降水事件的发生率小于2％，但它们对土壤侵蚀的影响却异常高。在四个单独的随机森林（RF）模型开发中，对整个数据集和玉米栽培的子数据集预测了径流和SL。经常使用的线性模型不适用于预测土壤侵蚀，因此使用了更可靠，非参数的高级分类树分析方法。RF分类方法能够预测腐蚀风险。对于玉米子数据集，RF模型最好地预测了极端事件，其次是无径流类别。最高和最低风险组的敏感性对于SL和RO均超过82％。耕作类型是最重要的因素。这项长期研究表明，CT的使用可以维持耕地上的大部分降水，因此，土壤流失仍比其可承受的数值低一个数量级。RF方法适用于模拟RO和SL。将来，在建模中集成更多数据集将大大提高RO和SL预测的准确性和准确性。