Sagebrush ecosystems of the western United States can transition from extended periods of relatively stable conditions to rapid ecological change if acute disturbances occur. Areas dominated by native sagebrush can transition from species-rich native systems to altered states where non-native annual grasses dominate, if resistance to annual grasses is low. The non-native annual grasses provide relatively little value to wildlife, livestock, and humans and function as fuel that increases fire frequency. The more land area covered by annual grasses, the higher the potential for fire, thus reducing the potential for native vegetation to reestablish, even when applying restoration treatments. Mapping areas of stability and areas of change using machine-learning algorithms allows both the identification of dominant abiotic variables that drive ecosystem dynamics and the variables’ important thresholds. We develop a decision-tree model with rulesets that estimate three classes of sagebrush condition (i.e., sagebrush recovery, tipping point [ecosystem degradation], and stable). We find rulesets that primarily drive development of the sagebrush recovery class indicate areas of midelevations (1 602 m), warm 30-yr July temperature maximums (tmax) (30.62°C), and 30-yr March precipitation (ppt) averages equal to 26.26 mm, about 10% of the 30-yr annual ppt values. Tipping point and stable classes occur at elevations that are lower (1 505 m) and higher (1 939 m), respectively, more mesic during March and annually, and experience lower 30-yr July tmax averages. These defined variable averages can be used to understand current dynamics of sagebrush condition and to predict where future transitions may occur under novel conditions.

如果发生严重干扰，美国西部的鼠尾草生态系统可以从较长时期的相对稳定状态过渡到快速的生态变化。如果对一年生草的抵抗力很低，以当地鼠尾草为主的地区可以从物种丰富的原生系统过渡到以非当地一年生草为主导的变种州。非本地一年生草对野生动植物，牲畜和人类的价值相对较小，并且起增加火灾频率的作用。一年生草覆盖的土地面积越多，发生火灾的可能性就越大，因此即使进行恢复处理，也减少了原生植被重建的可能性。使用机器学习算法绘制稳定区域和变化区域的图，既可以识别驱动生态系统动态的主要非生物变量，也可以识别变量的重要阈值。我们用规则集开发了决策树模型，该规则集估计了三类鼠尾草的状况（即鼠尾草的恢复，临界点（生态系统退化）和稳定状态）。我们发现主要驱动鼠尾草恢复类别发展的规则集表明中等海拔地区（1602 m），7月30日温暖的最高温度（tmax）（30.62°C）和30年3月平均降水量（ppt）等于26.26毫米，约为30年ppt年值的10％。临界点和稳定类别分别发生在较低的海拔（1505 m）和较高的海拔（1939 m），3月和每年的中斜率更高，并降低了30年7月的最高吨位平均值。这些定义的变量平均值可用于了解当前的鼠尾草状况动态，并预测在新的状况下将来可能发生的过渡。