Mountainous headwater streams represent a substantial proportion of the global stream network. These small streams may flow episodically, seasonally, or perennially, providing diverse values and services. Given their broad importance and growing pressures on terrestrial and aquatic resources, we must improve our understanding of the drivers of flow permanence to facilitate informed land and water management decisions. We used field observations from >10 cross-sections in each of 101 nonfish bearing, headwater streams across four geomorphic provinces in Northern California to quantify flow permanence and network connectivity during the summer low flow period in 2018. At each stream cross-section, we noted the presence or absence of streamflow and used this information to classify streams as perennial (continuous streamflow in all cross-sections) or non-perennial and connected (surface water in the most downstream cross-section) or disconnected. At each cross-section, we also quantified channel size (width and depth) and grain size. We coupled field observations with geospatial data of catchment physiography, hydrology, and climate in random forest models to investigate controls of flow permanence and network connectivity. Potential drivers of flow permanence or network connectivity included in our models were channel geometry, grain size, slope, aspect, elevation, annual and seasonal precipitation, air temperature, and topographic wetness index. We found more perennial streams in the Klamath Mountains and Sierra Nevada than in the Cascades and N. Coast regions. Streams in the Klamath were the most connected followed by streams in the N. Coast, Sierra Nevada, and Cascades. The most important variables for predicting flow permanence were channel grain size, winter 2018 precipitation, and drainage area. Comparatively, the most important variables for predicting network connectivity were winter and spring 2018 precipitation, grain size, and bankfull depth. Our study illustrated the complexity of the processes that drive flow permanence and highlighted the uncertainty in projecting the precense of water in streams across diverse regions.