Abstract. We consider Gromov's homological higher convexity for complements of tropical varieties, establishing it for complements of tropical hypersurfaces and curves, and for nonarchimedean amoebas of varieties that are complete intersections over the field of complex Puiseux series. Based on these results, we conjecture that the complement of a tropical variety has this higher convexity, and prove a weak form of this conjecture for the nonarchimedean amoeba of any variety over the complex Puiseux field. One of our main tools is Jonsson's limit theorem for tropical varieties.A tropical hypersurface is a polyhedral complex in R n of pure dimension n−1 that is dual to a regular subdivision of a finite set of integer vectors. This implies that every connected component of its complement is convex. A classical (archimedean) amoeba of a complex hypersurface also has the property that every connected component of its complement is convex [2, Ch. 6, Cor. 1.6].Gromov [3, § ] introduced higher convexity. A subset X ⊂ R n is k-convex if for all affine planes L of dimension k+1, the natural map on kth reduced homologyis an injection. Connected and 0-convex is ordinary convexity. Henriques [4] rediscovered this notion and conjectured that the complement of an amoeba of a variety of codimension k+1 in (C × ) n is k-convex, and established a weak form of this conjecture: the map ι k sends no positive class to zero [4]. Bushueva and Tsikh [1] used complex analysis to prove Henriques' conjecture when the variety is a complete intersection. Other than these cases, Henriques' conjecture remains open.An amoeba is the image in R n of a subvariety V of the torus (C × ) n under the coordinatewise map z → log |z|. Similarly, the coamoeba is the image in (S 1 ) n of V under the coordinatewise argument map. Lifting to the universal cover and taking closure gives the lifted coamoeba in R n . There is also a nonarchimedean coamoeba and a lifted nonarchimedean coamoeba [11]. The complement of either type of lifted coamoeba of a variety of codimension k+1 is k-convex [10], which was proven using tropical geometry.We investigate Gromov's higher convexity for complements of tropical varieties. We show that the complement of a tropical curve in R n is (n−2)-convex. Both this and the convexity of tropical hypersurface complements rely only on some properties of tropical