Autogenic Succession in a Subtropical Savanna: Conversion of Grassland to Thorn Woodland

Dense thorn woodlands occupy what are thought to have been grasslands and savannas prior to settlement of the Rio Grande Plains of Texas. However, the tenet that grasslands have been converted to shrublands and woodlands in recent history is controversial and based largely upon conflicting historical accounts. Our objective was to determine how the presumed physiognomic conversion from grassland or savanna to woodlands might have occurred. Some upland landscapes are dominated by closed—canopy woodlands in southern Texas, whereas others have a two—phase pattern of discrete shrub clusters scattered throughout a grassland. More mesic sites are dominated by losed—canopy woodlands. We hypothesized the two—phase landscapes represented an intermediate stage in the conversion of grassland to woodland. As new shrub clusters were initiated and existing clusters expanded and coalesced, a gradual shift from grassland to savanna to woodland would occur. To address this hypothesis, we inventoried herbaceous interspaces for woody colonizers, quantified the composition and distribution of shrub clusters on upland sites, and compared the structure of clusters to that of adjacent, more mesic areas with continuous woody plant cover. To assess the physiognomic stability of the two—phase landscapes, cluster size, density, and cover were quantified for 1941, 1960, and 1983 from aerial photographs. A lone mesquite (Prosopis glandulosa) plant occurred in >80% of the upland clusters, where it was typically the largest individual in terms of basal area, height, and canopy area. The number of woody species per cluster ranged from 1 to 15 and was strongly related to mesquite basal diameter (R² = 0.86). Cluster diversity, evenness, and size were also significantly correlated with mesquite size. The data suggest that mesquite plants invaded grasslands and served as recruitment foci for bird—disseminated seeds of other woody species previously restricted to other habitats. The result was a landscape composed of discrete chronosequences of woody plant assemblages organized about a mesquite nucleus. Within the two—phase portion of the landscape, 50% of the clusters were within 5 m of another and 95% were within 15 m of another. Analysis of the size class distribution of clusters suggested that most had yet to realize their growth potential. Moreover, the herbaceous clearings between clusters contained high densities of woody seedlings, mostly (>70%) mesquite, which occurred in 85% of the clearings, with a mean density of 350 plants/ha. Coalescence will become increasingly probable if new clusters are initiated and existing clusters expand. This phenomenon appeared to be in progress on one portion of the landscape and had apparently already occurred on others. As clusters developed on the two—phase portion of the landscape, their species composition, dominance, and size class structure became increasingly similar to that of adjacent closed—canopy woodlands on more mesic. Mean cluster size increased from 494 m² in 1941 to 717 m² in 1983. Growth rates of clusters were a function of cluster size and precipitation. During the 1941—1960 period characterized by severe drought, there was a slight decrease in total woody plant cover resulting primarily from the formation of gaps among clusters on the periphery of the site and a 35% decrease in density of clusters <5 m². These cover losses offset the areal expansion of small (<100 m²) clusters which had an average relative growth rate (RGR) of 0.10 m² · m— ² · yr— ¹. Post—drought woody plant cover increased from 8% in 1960 to 36% in 1983 as new clusters were initiated density increased from 16 to 26 clusters/ha), and the RGR of small clusters increased to 0.16 m² · m— ² · yr— ¹. Numerous clusters coalesced during this period. RGRs of clusters >100 m² were an order of magnitude lower than those of clusters <100 m² in each time period. The RGR of large clusters following the drought was not significantly increased, except in the largest size class (clusters >1,000 m²). Our results indicate (1) mesquite invaded grasslands and served as the nucleus of cluster organization on upland sites; (2) woody plant community development has been highly punctuated by variations in precipitation; (3) clusters >5 m² in area are persistent features of the landscape; and (4) the present two—phase pattern is moving toward a monophasic woodland as new clusters are initiated and existing clusters expand and coalesce. As a result, (5) shrub clusters on uplands represent an intermediate stage in the conversion of grassland to woodland, and (6) closed—canopy woodlands on more mesic sites appear to represent portions of the landscape where this has already occurred. Because the conversion of grasslands and savannas to woodlands in the Rio Grande Plains is initiated by mesquite, factors regulating its dispersal, establishment, and role as a facilitator of woody community development are emphasized.