Deborah Lawrence and David Foster

Deborah Lawrence. B.A. Harvard University. Ph.D. Duke University. Address: Department of Environmental Sciences, University of Virginia, P. O. Box 400123, Charlottesville, Virginia 22904-4123, USA. e-mail: lawrence@virginia.edu

David Foster. B.A. Connecticut College. M.S. and Ph.D. University of Minnesota. Address: Harvard Forest, Harvard University, P. O. Box 68, Petersham, Massachusetts 01366, USA.

Se evaluó el impacto relativo de los condicionantes ambientales y los efectos del uso de la tierra a nivel de lotes, sobre procesos del ecosistema en un bosque tropical seco del sur de la Península de Yucatán, a través del muestreo a lo largo de gradientes naturales y hechos por el hombre. Los objetivos fueron 1) describir los gradientes ambientales naturales de precipitación y suelos en escala regional, 2) explorar la respuesta de propiedades clave y procesos del ecosistema a tales gradientes, y 3) comprender los efectos de la edad del bosque sobre lotes dentro de una región dada del gradiente. Se estudiaron la biomasa superficial (viva y detrítica), la producción y química de hojarasca, y los suelos. En cada uno de 3 sitios con un gradiente de lluvia estacional entre 900-1400 mm/año se muestrearon 10 a 13 lotes, incluyendo bosque secundario y maduro. La caída de hojarasca aumentó y su contenido de nutrientes bajó en el pico de la estación seca. En el bosque maduro los factores ambientales asociados al gradiente de precipitación limitaron la producción de hojarasca, biomasa y materia orgánica del suelo. La historia humana del área, incluyendo la tala y decisiones acerca de áreas no afectadas (p. e. vertientes altas), añadió información crítica para comprender diferencias en la biomasa superficial del bosque maduro. Todos los aspectos de estructura y función en bosques secundarios jóvenes (biomasa viva superficial, producción de hojarasca, masa en el piso y propiedades críticas del suelo) están fuertemente influenciadas por la edad del lote. La recuperación de los niveles de bosque maduro se estima conservadoramente en 55-95 años. P y/o N parecen limitar procesos esenciales del ecosistema como producción de hojarasca, descomposición y biomasa superficial. Se sugiere que una vez superada la limitación de agua, el P se hace limitante. La evidencia incluye mayor reabsorción de P de la hojarasca, una fuerte respuesta de la concentración de P, pero no de la de N, al gradiente de precipitación, y un aumento en la eficiencia de uso de P con el aumento en la edad del lote. La presencia de humanos en el sur de Yucatán resulta en mayores cambios del ecosistema que aquellos inducidos por variación natural del ambiente en escala regional. Para comprender mejor estos paisajes y para llevar estudios puntuales a evaluaciones regionales y globales, resulta vital integrar los impactos humanos en los procesos del ecosistema.

Summary

Received: 01/15/2002. Modified: 06/17/2002. Accepted: 07/02/2002

Over the past four decades the southern Yucatan peninsular region has undergone increasing amounts and intensity of land use change, ranging from selective logging to widespread shifting cultivation agriculture to land clearing for intensive agriculture and village establishment. These land uses alter the structure and function of forested lands and often generate new feedbacks in terms of subsequent human use. Consequently, a major goal in assessing regional environmental change is to understand how biogeochemical processes respond to land use change, emphasizing the potential of a human-dominated landscape to sustain continued human use. One of the greatest challenges in these studies is to distinguish the effects of environmental gradients in climate, geology or natural disturbance from the effects of human activity. In the Southern Yucatan Peninsula Region project (SYPR), our approach to this challenge has been to investigate ecosystem processes in several study sites arrayed across the dominant environmental gradients while focusing on the influence of local, human-controlled factors within a given area.

In the southern Yucatan, annual precipitation increases by more than 50% over 120km from the northeastern border of the Calakmul Biosphere Reserve to the Guatemalan Border. Median annual precipitation varies from ca. 900mm in the north of our study area to ca. 1400mm in the south. This range encompasses approximately 50% of the variation in precipitation of dry tropical forests worldwide (Murphy and Lugo, 1986). Thus, the gradient represents a broad sampling of the environment of the southern Yucatan peninsula and provides data relevant to many other dry forests. Rainfall is highly variable on an annual and a monthly basis, but all sites share a distinct seasonal pattern regardless of total annual precipitation. A pronounced dry period (<50 mm/mo) may begin as early as November or December, but often begins in January. It may last from three to seven months, depending on the year and the latitude (Figure 1)

The major stand-level factors determining the rate and dynamics of ecosystem processes are forest age, cultivation history, and management (land use type and landowner practices). Because shifting cultivation of maize (milpa) is the dominant cause of deforestation in the region (Turner et al., 2001), forest recovery following this kind of disturbance was the central focus of investigations. The main objective of this study was to assess how nutrient cycling, productivity and biomass change throughout the course of secondary forest development following (temporary) abandonment of agricultural land. Plot-based studies of forest recovery within stands at a given site were placed into a broader context by comparing data across sites. This allowed us to test for the influence of regional environmental drivers.

The two-tiered conceptual approach, investigating local sites within a regional framework, facilitates an understanding of the multiple constraints on the trajectory of individual patches on the landscape. The landscape under shifting cultivation consists of many such patches at various stages of recovery. The approach employed also makes it possible to characterize the structure and function of the entire landscape at any one point in time, and to project it into the future, through integration with predictive models (Geoghegan et al., 2001). Before making projections, it is necessary to understand current constraints on forest recovery in the SYPR. With that goal in mind, three questions are addressed in this paper:

1) What are the regional environmental influences on forest processes at the scale of 10s to 100s of kilometers? Do precipitation and soil fertility vary significantly at this scale?

2) How do natural environmental gradients in soil and precipitation affect ecosystem processes and characteristics?

3) Within a given region, how do ecosystem processes vary as a function of forest age?

Field work was concentrated in three study sites positioned 60-120km apart in order to sample across the regional precipitation gradient (Figure 2). Thirteen stands were studied in El Refugio (ER), the driest site (ca. 890mm/yr) near the northeastern border of Calakmul Biosphere Reserve in the State of Campeche. At Nicolás Bravo (NB), an area of intermediate rainfall (ca. 1150mm/yr) 60km to the southeast in the State of Quintana Roo, an additional 13 stands were studied. Approximately 120km south of ER on the border of Quintana Roo, Campeche, and Guatemala, 10 stands were studied in the village of Arroyo Negro (AN), the wettest part of the study area (ca. 1400mm/yr). Stands were sampled in both mature forest (2-3 per site) and secondary forest fallows regenerating after shifting cultivation of maize (8-10 per site). The mature forest stands were at least 50 years old (Read and Lawrence, in press) and had not been cleared for agriculture in recent history. None of the fallow stands had received inputs of fertilizer, pesticide, or herbicide. The secondary forest stands ranged in age from 2-25 years since the last harvest (Table I). Despite the absence of stumps or other evidence of past disturbance in the immediate vicinity of our stands, all of the forests may have experienced selective timber extraction in the last 40-100 years (Klepeis and Turner, 2001).

Secondary forest stands were selected from the database compiled by the socio-economic members of our research team (see Geoghegan et al., 2001, Klepeis and Turner, 2001). We used information on the historical details of individual parcels, units of ownership and management by a single household. A parcel may encompass open, cultivated lands and forest stands of different age depending on when they were last used for cultivation. Information on parcel history made it possible to account for confounding factors such as differences in prior land use type, intensity of cultivation, and use of chemical inputs. These factors can modify the trajectory of recovery and should be sampled explicitly rather than inadvertently (e.g. Bushbacher et al., 1988, Fernandes and Sanford, 1995, Hughes et al., 1999). The effect of cultivation history was investigated in a separate paper, we deal primarily with forest age and regional environmental gradients in this study.

In each of the three study sites, several parcels were located containing young, middle-aged, and old fallow stands. Although clearing and cultivation practices may differ by landowner, within a given farmer’s parcel it is possible to minimize management differences while also minimizing local differences in soil texture, inherent soil fertility, and topography. Thus, in establishing chronosequences within a given parcel, we expect to sample the effects of forest age, rather than the effects of contrasting management approaches or subtle edaphic variation. The true sampling unit was, in a sense, the farmer. The various forest stands owned by each farmer represents one trajectory of forest recovery. This approach, using the uniformity of practice within farm ownership is unusual and was facilitated by the cross-disciplinary nature of our project.

One 500m2 circular plot was established in each stand within a given parcel (Figure 3). Litter was collected once a month (Nov. 1998 - Jan. 2000) in four 1m2 litter traps per plot. The plots were located permanently, about 20cm off the ground and 8m from the center along orthogonal axes. Fine litter (leaves, reproductive material, bark and wood <2cm diam.) was separated from coarse litter (bark and wood >2cm in diam.), dried at <70°C and weighed. The nutrient content (C, N, and P) of the fine litter was used to gauge the effects of land use change and forest recovery on nutrient cycling (c.f. Odum, 1969, Vitousek, 1984, Proctor et al., 1983). The dry mass of fine litter was used as an index of productivity. Although biomass accumulation in stems and branches represent a substantial portion of productivity in secondary forests, our analysis suggests that the rate is constant throughout the first 25 years of growth at approximately 3-4 Mg/ha/yr (Read and Lawrence, in press). Furthermore, the rate of litter production was found to be 50-100% higher than total biomass production across the study area.

In each plot, aboveground biomass was sampled, including live vegetation and detritus. To estimate the biomass of detritus on the forest floor, all fine litter was sampled in four 1m2 areas, each 1-2m from an existing litter trap, during Nov. and Dec. 1998. This data yields a low estimate of forest floor biomass as peak litterfall had occurred in all regions and all forest ages eight to nine months prior. All stems >10cm in diameter at breast height (dbh) were recorded for the entire 500m2 plot. A nested design was used to sample smaller stems, with stems 5-10cm dbh measured in a central 100m2 plot, and stems 1-5cm dbh measured in three 10m2 plots (see Figure 3; Read and Lawrence, in press). At 13 systematically arrayed points within the large plot, measurements of height and diameter were taken on the nearest individual in each dbh class. These data (1395 trees from all plots) were used to create regressions to estimate heights for all trees sampled. Then, standing live biomass was estimated from dbh, height and wood density, using established regressions based on studies of dry tropical forest of similar structure (cf. Martínez-Yrizar et al., 1992; see Read and Lawrence, in press for further details).

Figure 4. Seasonal pattern of fine litter production (leaves, twigs, and wood <2cm in diameter). Mean ±SE across all plots within a given region. Variation indicative of the range in productivity as a function of forest age. Region is indicated by symbol as shown in legend.

Results and Discussion

As in the dry forests of Western Mexico and portions of the Northern Yucatan, and in the wetter forests of Guatemala further south, litter production in the southern Yucatan peninsula was strongly tied to the timing of precipitation (Kunkel Westphal and Kunkel, 1979, Martínez-Yrizar and Sarukhan, 1990, Whigham et al., 1990). Both the production and nutrient concentration of litter responded to seasonal drought from Dec. to May. Fine litter production was relatively constant throughout most of the year, at 25-30 g/m2/mo. Several months into the dry season, however, litterfall increased five-fold, apparently due to water stress (Reich and Borchert, 1984, Holbrook et al., 1995; (Figure 4). Thirty percent of total annual litter production fell during the peak, between Mar. and Apr., and roughly 60% fell from Jan. to May.

Litter nutrient concentrations reached their minima during peak litterfall (Figure 5); Read and Lawrence, submitted). This evidence suggests that the trees reabsorbed essential nutrients prior to massive leaf loss (Chapin, 1980, Killingbeck, 1996). Litter P concentrations were reduced 47% from the maximum observed in a given site. In contrast, litter N concentrations were reduced by only 33% (Read and Lawrence, submitted). Lower nutrient concentrations in the dry season have been reported elsewhere as well, in both wetter and drier forests (Swift et al., 1981, Wieder and Wright, 1995, McGrath et al., 2001). Despite a reduction in nutrient concentration, the increase in litterfall during the dry season is likely to result in a pulse of nutrients with the onset of the rains in June (Swift et al., 1981, Singh et al., 1989, Campo et al., 1998, McGrath et al., 2001). The entire SYPR is likely to exhibit a similar pattern of highly seasonal soil nutrient dynamics.

Although the study area is relatively homogeneous in topography and parent material, the physical and chemical properties of mature forest soils did vary (Table II). Some of this variation results from interactions between vegetation and soils through geologic time. It may also be caused by differences in the intensity of historic land use during the past century. It could also result from ancient Mayan activity. An important part of the variation, especially in soil organic matter, is most likely a result of the regional precipitation gradient. In particular, the southern site tended to have a higher organic matter content, as expected if litter inputs were higher due to greater precipitation (c.f. Vitousek, 1984). Other aspects of soil variation, such as the higher percentage of sand-sized particles at AN may relate to the topographic position of the sampled stands. Over time, smaller particles tend to move downslope, leaving a coarser soil texture at the top of a catena (Brady, 1990). Mature sites in AN were somewhat higher and closer to the ridge top. Sites in ER and NB were lower on the slope.

Soil pH was generally quite high across the region (7.4-7.7) as a result of parent material rich in calcium carbonate. However, the Ca content of soils in NB was significantly lower than that of the other two sites. Consequently, soils in NB also had lower pH and lower concentrations of total exchangeable cations. Overall fertility may be lower in NB given lower levels of soluble sulfur and higher concentrations of detrimental and potentially toxic metals (Al, Cu, Fe, and Mn; Table II). This lower fertility could reflect the longer history of agriculture there. ER and AN were established 20-30 years ago, but NB has been settled since the 1950s. NB has large areas of Bracken fern, an invasive species whose dominance may be facilitated by nutrient-poor soils (Suazo, 1998) and/or by repeated fire such as that associated with shifting cultivation.

The simplest explanation for enrichment in metals and loss of Ca may be topography. NB is at a slightly lower elevation, on a flatter part of the study area, and thus it is relatively older geologically than the more hilly areas we sampled to the northwest and the southwest (Grant Goodell, personal communication). The timeframe for weathering is at least modestly longer in NB. The relative abundance of metals can increase during the weathering process through differential loss rates of the elemental constituents of minerals (Brady, 1990). Over time, more Ca could have been lost from the system through weathering followed by leaching. It is unclear whether the difference in geological age is adequate to explain the differences observed. An alternative explanation is that ancient Mayan land use may have disproportionately and negatively affected the soils around NB. Ruins are extremely abundant from NB west to Becan (Turner, 1974). ER is also within the potential realm of influence of major temple sites, and yet its soils were more fertile than those in NB. The intensity of recent historic land use seems to be the best explanation for the difference among sites, but further study is necessary to eliminate the alternative hypotheses.

Differences in total annual precipitation and soil characteristics may contribute to variation in structure and function among the study sites (Figure 6). To isolate the regional influence on forest processes from the effects of forest age and cultivation history, mature forests were examined separately. Forest floor biomass increased with increasing precipitation (Figure 6c). Higher precipitation should enhance decomposition, leading to lower forest-floor biomass if litter production were to remain constant. Thus, the observed increase in forest floor biomass suggests a concomitant increase in litter production. Mean litter production was greater in the south (Figure 6b), but the result was not statistically significant due to small sample size and high variability in AN. One of the two sites demonstrated surprisingly low biomass and production. Martínez-Yrizar and Sarukhan (1990) found higher litter production in the moister of their two study sites in Chamela, Mexico. Their results support our contention that differences in annual precipitation may have led to differences in litter production between the driest and the wettest sites.

Despite significant differences in rainfall and certain soil properties, live aboveground biomass did not vary significantly among the regions (Figure 6a). The highest values came from NB, where soils were apparently less fertile. The lower than expected values in AN may be due to unquantified impacts of recent logging (see Klepeis and Turner, 2001). Mature forests in AN had fewer trees (>10cm dbh) and lower basal area than mature forests in the drier sites, suggesting that some trees may have been extracted in recent history (Read and Lawrence, in press). On the other hand, high biomass in NB may be the result of an over-abundance of lianas, perhaps another legacy of historic disturbance, human or natural (Boose et al, 1994, Gerwing and Lopes Farias, 2000). The mean contribution of lianas to total biomass was 10.2% in NB vs. 5.6% in ER and 4.8% in AN (Read and Lawrence, in press).

Phosphorus, but not nitrogen, dynamics followed trends in annual precipitation. Total annual inputs of P were greater in the wettest region, AN (Figure 7a), as a result of both greater productivity (Figure 6b) and higher P concentrations in the litter (Read and Lawrence, submitted). The exception to this trend was in NB, where intermediate P inputs would be expected. Instead, the low relative soil fertility discussed above seems to have mitigated the effect of added precipitation on P cycling. Nitrogen inputs increased slightly with increasing precipitation, but the trend was not significant (Figure 7b). This modest trend reflects only differences in total productivity, as litter nitrogen concentrations did not vary significantly among regions (Read and Lawrence, submitted). The response of forest P cycling to regional precipitation gradients and the lack of response in the N cycle suggest that forests in the SYPR are more limited by P than by N. When water stress was alleviated, additional P, but not additional N, was introduced into the system. Thus, the manifestation of nutrient limitation depends upon water availability. This may explain why Campo et al. (2001) reported that P did not appear to be limiting in the very dry forests of Chamela, although small amounts of P were retained in the system. The indication of P limitation in this study came out of a regional analysis, where more P, but not more N, was observed to cycle in the wettest region. N cycling appears less sensitive to increased precipitation than P cycling.

As defined by the range of variation induced, forest age had a greater impact on ecosystem properties and function than did regional environmental gradients. Across all regions, live aboveground biomass increased significantly as a function of forest age (Figure 8a). There was a pronounced difference between mature forests and the oldest secondary forests sampled (12-25 y.o.). Although total woody basal area of stems >1cm dbh had recovered to 63% of mature forest levels by 25 years, live biomass reached only 40% of mature forest levels (58 vs. 136 Mg/ha; Read and Lawrence, in press). Earlier analysis, based on tree basal area only, suggested that basal area (>10cm dbh) increased to 80% of mature forest levels within 25 years (Turner et al., 2001). Subsequent analysis of biomass based on all stems (including lianas and small trees <5cm dbh) suggests a slower rate of recovery (Read and Lawrence, in press). Our prior analysis depended on the assumption of a linear increase in biomass beyond 25 years. If biomass growth is not linear, and the biomass increment declines, for example, by 50% after forests achieve a closed canopy, then recovery would be slowed further. If we assume that total canopy closure occurs 25 years following abandonment, instead of reaching mature forest levels in another 15-35 years, it may take 30-70 years. Thus, our most conservative estimate is 55-95 years for recovery to current mature forest levels.

  Figure 9. Total annual phosphorus (a) and nitrogen (b) inputs deposited in fine litter as a function of forest age. Calculated from monthly dry mass and monthly litter nutrient concentrations (Read and Lawrence, submitted). Mean ±SE for all plots within an age class. Statistically significant differences between age classes indicated by different letters.

Both litter production and forest floor biomass increased consistently with age (Figures 8b, 8c). The difference between old secondary and mature forests was not as dramatic as it was for live biomass. With only 15% of the aboveground biomass, the youngest forests produced almost 70% as much fine litter as mature forests. Similar results were reported by Ewel (1976) in Guatemala. The pattern of high productivity in young forests matches that observed in successional forests of the temperate zone as well (Odum, 1969). The increase in forest floor biomass was more rapid than that expected due to the increase in litter production, suggesting that decomposition slows with forest age. Experimental studies confirmed that decomposition is slower in older forests. Lower decomposition rates seemed to occur in response to lower litter quality rather than to changes in microclimate (Xuluc-Tolosa et al., in press). In fact, as an indicator of litter quality, litter P concentration declined as a function of forest age but no significant change in litter N concentration occurred (data not shown; Read and Lawrence, submitted).

Whereas P, but not N, responded to increased precipitation, mean P inputs across all regions did not increase as a function of forest age, whereas N inputs did so (Figure 9). Thus, P cycling changes markedly during the process of forest recovery, with P use efficiency increasing as a function of age as suggested by Brown and Lugo (1990). Older forests produced significantly greater amounts of litter (Figure 8b), while cycling the same amount of P as younger forests (Figure 9a), resulting in a decline in litter P concentration with age (Read and Lawrence, submitted). In contrast, the pool of N cycled through litter increased in step with increasing litter production in older forests. Litter N concentration did not increase significantly, although there was a tendency for slightly higher concentrations in older forests. Taken together, these data suggest that N-use efficiency declined, if it changed at all, as a function of forest age.

Understanding changes in nutrient cycling during forest recovery depends on the nutrient examined. Furthermore, evaluating nutrient constraints on ecosystem processes depends on understanding interactions between water -and nutrient- limitation. When water is broadly limiting (as in the above analysis of nutrient dynamics as a function of forest age, regardless of regional precipitation gradients), productivity seems to be limited by N availability. More N cycling through the system was correlated with higher litter production. All forests cycled similar amounts of P regardless of age. This level may represent the maximum P accessible given the soil-climate regime. P could still be limiting productivity, but we cannot determine the limitation by simply comparing P return with litter production, as a function of forest age. Time alone (as a forest ages) will not allow more P to enter the system, unless fixed biogeochemical constraints are lifted. Although the potential exists for N uptake through biological fixation, there are no additional pools to tap to alleviate P limitation. Increasing production depends, then, on increasing nitrogen availability as the forest and its soils recover from cultivation. When water limitation is alleviated and soil biological activity is enhanced (as it appears to be in AN), certain biogeochemical constraints on P cycling do seem to be lifted. Only then, can additional P enter the system (see Figures 6 and 7). This P apparently yielded gains in litter production in AN, suggesting that P becomes limiting to productivity when water is no longer the primary limiting factor.

Changes in soil properties with forest age suggest that N availability may increase during forest regeneration. Although data on the N and P content of soils are lacking, soil organic matter did increase significantly with age (Table III). In the top 15cm of soil, organic matter increased from 10-11% in young secondary forests to 15% in mature forests. An increase in N availability associated with an increase in organic matter would allow litter concentrations to remain stable while productivity increases. This increase in soil organic matter probably represents a slowing of mineralization rates in the soil along with a demonstrated increase in litter inputs. Likewise, we attributed the increase in forest floor mass with age to added inputs and slowed decomposition. Data on K concentrations support this interpretation: K concentration was lowest in mature forest soils, where turnover rates of organic matter are expected to be slowest due to a decline in litter quality.

The study of nutrient dynamics in the region provided important clues about the regenerative capacity of the forests. Any assessment of land use and land cover change must take into account several critical points. Regional precipitation gradients strongly influenced nutrient cycling through soil organic matter and litter, but had a more limited effect on plant productivity and biomass. Increased precipitation allowed additional phosphorus to enter the system. Although forest age profoundly influenced many aspects of nutrient cycling, its greatest effect was to alter litter production and aboveground biomass. P use efficiency tended to increase and N use efficiency tended to decrease slightly with age. Water limitation seems to be the predominant environmental factor limiting ecosystem processes in these dry forests, and phosphorus may be the most significant limiting nutrient.

Sampling along natural and human-made gradients has allowed the evaluation of broad environmental and stand-level drivers of ecosystem processes. Environmental factors do constrain litter production, forest floor mass, and soil organic matter in mature forest. However, to understand differences in the biomass of mature forests, further information about the human history of the area is critical, including logging history and decisions about what areas have purposefully been left untouched (perhaps the upper slopes). Structure and function in secondary forests (biomass, litter production, forest floor mass and critical soil properties) are strongly influenced by the age of the stand. The presence of humans on the landscape results in changes that are greater than those induced by natural, environmental variation at the regional scale. For the best understanding of these landscapes, and for scaling up from case studies to regional and global assessments, integrating human impacts on ecosystem processes is vital.

Efforts to use remotely sensed data to scale up from point samples to regional values will depend critically on the ability to distinguish young and old secondary forest. Models that predict landscape-scale carbon stocks, for example, will be extremely sensitive to the mapping of field-derived age categories onto image-derived land cover classes (e.g. Cairns et al, 2000). On the other hand, distinguishing older secondary forest and mature forests will be less critical for determining regional productivity, given the similarity in litter production. The amount of classification error that can be safely tolerated thus depends on the goal of the model: carbon stocks or carbon fluxes. Our studies of nutrient cycling suggest that similar considerations must be made when trying to model other biogeochemical cycles at the regional scale.

The authors thank the farmers of El Refugio, Nicolás Bravo, and Arroyo Negro who worked with them; Juan Méndez-Díaz for essential assistance and leadership in the field; numerous students for excellent help in the field and the lab: Fausto Bolom, Jakara Hubbard, Larissa Read, Jessica Sisco, Heidi Wasson, and Tana Wood; and Pedro Macario Mendoza of ECOSUR for logistical support. Funding was provided by NSF, NASA, The Mellon Foundation, Carnegie Mellon University, and the University of Virginia.