A discussion of the content of this report requires it to be divided into two component parts: (i) methodological issues; and (ii) the results of the application of the methodology to the case studies described in the Appendix.
Methodological issues
The design and formulation of the methodology to meet the project objectives and its application to real case studies on the ground brought a number of methodological issues to light.
First, the methodology is not intended to be ecosystem specific. It is not oriented specifically to the measurement of forests, grassland or cropland. Rather, it addresses the land use in a geographical area of interest regardless of the ecosystems that may be present.
In the methodology, LUC is considered the pivotal point for addressing synergistically the concerns of enhancing carbon sequestration, promoting biodiversity conservation and preventing land degradation, while deriving yields and produce to address food security and rural income. In this sense, “win-win” strategic LUC should emerge from the application of the proposed methods in terms of both the synergies expressed above and ecological benefits derived from such synergies.
The methods and procedures in this report aim to determine first the status of carbon stock, biodiversity and land degradation under current land use and management, and then to identify promising land-use patterns. That is to say, it seeks to identify LUTs and the crops, trees and shrubs within them that are biophysically and socio-economically suitable for each zone or ecological condition in the area, and with enhanced CSP for generating potential scenarios of LUCs. Finally, it aims to optimize the multiple objectives of stakeholders in order to select the potential land-use pattern that optimizes such multiple objectives simultaneously. In the process, stakeholder participation is elicited through the articulation of preferences for scenarios, which become consensual priorities. These priorities guide the sequence in the optimization process.
Definition of study area
The watershed or subwatershed was found to be a convenient geographical framework, facilitating the identification and mapping of land cover classes, the identification and delineation of ecological units or ecozones, the design and deployment of a sampling scheme on the ground, and the compilation of data and information for data processing and analysis.
Methods used
The methods proposed here are a methodological framework. They allow for a certain degree of flexibility so they can be modified to suit the particular circumstances of implementation. They rely on accessible technologies and on the likely existence of databases, which should be available in many developing countries. Thus, it is assumed that the databases in the area of application hold necessary and sufficient attribute and spatial data, and that if some data need to be collected, this can be done with ease and at low cost. This is a major obstacle to implementing modelling and data analysis techniques in some extreme conditions in the developing world. Hence, every attempt has been made to simplify the steps to achieve results in a typical environment in the developing world, e.g. Latin America and the Caribbean region. The ability to obtain results in remote areas of Mexico (the southern part of the Yucatan Peninsula) or the Rio Cauto watershed in the eastern part of Cuba, by working with existing data sets complemented with the field procedures described here, attests to the feasibility of the methods proposed.
From a methodological perspective, the main objective functions in the methodology, i.e. the maximizing of carbon sequestration, biodiversity conservation and minimizing of land degradation while maximizing production and income, should be seen as the main pillars of the methodology. The structure of the report follows these objectives.
Field sampling and upscaling
The issue of upscaling, i.e. spatial interpolation or “spatialization”, of site estimates of either biomass, carbon stock, biodiversity indices, land degradation indicators and any other variables measured or calculated at specific sites in the landscape, is a generic, important technical issue. It may mean significant differences in overall stock over an entire area. Therefore, it is a technical issue that practitioners may not want to underestimate and to which they may want to devote considerable attention and effort.
In situations such as the Bacalar case study, the band ratio image calculated is a useful mechanism for “spatialization” of estimates of aboveground biomass, rendering spatial interpolation unnecessary.
On the other hand, the case studies demonstrated that, in order to achieve efficiencies and economies of effort, time and cost, the field sampling procedures should be multipurpose in the sense of performing simultaneous measurements for biomass, biodiversity and land degradation assessment. Such data sets should be collected in the field by a multidisciplinary team. Experience through the case studies has shown that a multidisciplinary team of 7-9 technicians can complete the ground measurement and data recording tasks efficiently. The sampling requires variable-size nested sampling units (i.e. quadrats). The number of sampling sites required to provide a statistically representative sample of conditions in the study area can be achieved through a multistage stratified random sampling. At each stage of sampling, the variability of key measured variables (biodiversity indices, biomass estimates, etc.) can be assessed iteratively (to judge whether or not further sampling is required) or through the estimation of the a priori variance of the variable of concern from pre-survey sampling. When further samples do not contribute significantly to the variance (e.g. asymptotic behaviour of diversity indices curves with respect to the axis depicting the number of samples), further sampling is redundant and wasteful in terms of cost and effort. The use of pre-designed field forms became quite useful in data recording and in achieving time and effort efficiencies. Such field forms can be modified to suit the local conditions of each area of study. Also of great importance is the geopositioning or accurate georeferencing of the sampling quadrat sites. GPS instruments with high accuracies are preferred in order to achieve less than 5-m accuracy in positioning.
Wherever possible, the judicious inclusion of forest inventory data may achieve considerable savings in field sampling efforts, while achieving a representative sample with fewer sampling sites. In order to minimize sampling effort, much existing data on local soil inventories, vegetation classes, land cover types, topography and characteristics of the cropping, pasture and agroforestry systems in the area of study can be organized and incorporated ahead of the field sampling campaign. In some instances, it became necessary to collect plant samples for botanical identification. The inclusion in the field team of a botanist (equipped with a kit for collecting herbaria for later identification in the laboratory) may save considerable time and effort in the field.
Estimation of aboveground biomass
In terms of the estimation of aboveground biomass, all land cover types and ecosystems need to be assessed for accounting purposes. In this respect, the methods proposed make use of standard biomass inventory methods. Detailed estimates need to be produced for all land cover types including all levels of canopy, shrubs and herbaceous vegetation. The need to account for all strata is what makes the field surveys complex and somewhat labour intensive.
The methodological framework makes a commitment to direct carbon accounting instead of to sophisticated methods that are more technology-driven, such as monitoring C through CO2 fluxes (e.g. eddy covariance methods). In this accounting approach, the partition into pools of aboveground and belowground biomass, whether live or dead, saplings, debris and litter, presented no real logistical or technical problems in any of the case studies. This was because the measurement and monitoring methods at the field plots were standard and well known.
On the other hand, the use of remote-sensing products and techniques (multispectral satellite images and air-photographs) proved to be very useful when backed up by quadrat sampling measurements on the ground. The land cover classes generated from a supervised classification, after correction and validation with the quadrat samples, were crucial in the case of densely vegetated areas, such as the tropical forests of Bacalar, in providing the spatial framework for the upscaling of aboveground estimates. The land cover classes also allowed exploration of empirical relationships between the various pixel reflectance values and their corresponding biomass estimates from sampling sites on the ground. In this sense, the satellite images lent themselves (as part of the interpolation and upscaling mechanisms) as a good predictive regression equation of biomass as a function of the digital numbers of a band ratio index (GVI) was developed. As this relationship is site dependent, it was applicable only to the Bacalar study area. However, it indicates that the development of empirical regression equations to estimate biomass as a function of canopy reflectance, or the digital numbers of a band ratio index calculated from the original multispectral images (NDVI, GVI, etc.), may prove a profitable line of enquiry in other study areas where estimation may be required.
The use of standard regression equations of aboveground biomass as a function of measurements of trunk diameter, tree height, crown dimensions and other characteristics of the canopy proved quite advantageous in obtaining estimates, particularly as they incorporated other variables difficult to obtain from field measurements such as wood density. The key issue in the use of standard regression equations to estimate biomass is the careful selection of the equation that applies to the area of concern. This should be made according to the similarity in ecological conditions and forest type where the equation was developed. For routine monitoring programmes, calibration of such equations with locally measured data would be ideal.
An important issue in volumetric and biomass estimates in forested quadrats is the assessment of tree crown biomass. Estimation of canopy density and its contribution to tree biomass may be subjective and prone to contribute to the variance of estimates. The consistent use of specialized instruments to measure tree crown density and its contribution to tree biomass may help in alleviating this problem. A related issue is the estimation of litter and debris on the ground. The estimates of litter and debris biomass within the 1 × 1 m quadrat required extrapolation over the entire area of the 10 × 10 m quadrat (100 m2). This assumes that litter and debris ground cover is uniform and isotropic with the same thickness over the area of the larger quadrat. This issue may only be relevant in relatively undisturbed forested areas where the contribution of litter and debris to biomass and C may be significant.
A conservative approach was adopted when estimating root biomass. This may result in underestimates of total live belowground biomass. However, the trade-off is preferable to performing destructive sampling, which is ecologically, economically and even logistically prohibitive. In estimating root biomass by an indirect method, e.g. by a standard regression equation, caution must be exercised in the selection of the equation and in its applicability. The Winrock method (MacDicken, 1997) and the equation by Santantonio, Hermann and Overton (1997) provided useful and satisfactory results. The coefficients used to derive estimates of belowground biomass as a function of aboveground biomass made sense when distinguishing between coniferous and broadleaf trees.
The mapping of biomass relied on the existence of maps and the presence of polygons representing vegetation classes or land cover classes. This could have also been achieved through the use of spatial interpolation techniques. In this sense, geostatistical theory and techniques have a major role to play in the upscaling of estimates to the entire area of study (i.e. the watershed) starting from the point data represented by the sampling quadrat sites. A major constraint on the use of geostatistical techniques and in particular of kriging is that the their applicability depends strongly on the so-called “size of support” or number of point samples that support an interpolation. These techniques become reliable with an increasing number of samples.
Mapping the stock of carbon
Mapping the stock of C as biomass required a conversion factor only from biomass to C. This could be improved if tables containing the variations of carbon content in biomass by plant species were made available publicly. A similar issue is the estimation of carbon content in SOM. For the purpose of the case studies, a standard and to some extent arbitrary coefficient was used (derived from literature) to estimate carbon content as a function of SOM. SOM values are readily available from soil survey reports, making it possible to estimate C in soil from SOM at sites where soil profile analysis were conducted during soil survey.
Methods for estimating and measuring changes and fluxes of C in soil are direct and indirect. An indirect method for carbon estimation based on carbon accounting was preferred as direct methods are not feasible or affordable for routine assessment and monitoring programmes in the developing world.
The assessment of carbon stock and sequestration in the case studies of this report was carried out with data at each quadrat cell measured on the ground. Therefore, the upscaling of estimates is a crucial problem in determining the final accuracy of stocks for an area, watershed, district, state, etc. Central to this problem is the number of samples required in order to calculate a reliable semi-variogram or other form of auto-correlation function. Such a function could indicate the structure of spatial variability of the stocks in the area, and make it possible to use an optimal interpolator such as kriging with confidence. Considerations of cost, time and effort necessary for a relatively large number of samples to achieve a reliable semi-variogram and optimal interpolation must enter into a trade-off with considerations of accuracy of estimates, precision, ecosystem variability and area coverage (i.e. the size of the area that can be assessed and monitored). However, it is not possible to recommend a practical rule to achieve an optimal “compromise” solution to this problem.
Simulation models of soil organic matter turnover
Concerning the use of simulation models of SOM turnover, in order to determine the partitioning of added plant residues into the pools of SOM and contributions to them, a precondition is the presence of minimum data sets in the databases of the study area. The parameterization of the models demands the availability of certain data from the databases. Depending on the model used, parameterization could be quite demanding and even challenging and time consuming. However, experience has shown that there is an apparent trade-off between the degree of sophistication (and therefore of expected accuracy of a given model) and the difficulty in its parameterization. No model is simple, their use as a powerful predictive tool requires a learning curve, training, economic resources and even experimental fieldwork for their calibration and validation.
The parameterization of models also requires the existence of data on the parameters for each one of the pixels or soil polygons present in the study area. In order to achieve such fairly large databases, it may be pertinent to undertake a comprehensive exercise in agro-ecological zoning in the area in order to generate the pixel-based databases (i.e. PCCs) on all the parameters demanded by the model. Geostatistical interpolation techniques have a very important role to play here in terms of enabling the creation of complete “coverages” or data layers of multiattribute data over the entire study area.
Experience from the case studies indicates that the selection of a carbon dynamics model is a tradeoff between the model sophistication, the degree of “compartmentalization” of the recognizable carbon pools it simulates, the accuracy of simulated results and data requirements. All of these are set against intuitiveness, access and ease of manipulation of the model. It is difficult to derive a generic rule for model selection as many factors may intervene in the decision. Readers are advised to consult reliable “online” sources of model information, such as SOMNET, where all the characteristics of each model can be examined.
The models selected for the case studies (CENTURY and RothC-26.3) represent extremes in the spectrum of complexity, ease of access and manipulation, parameterization and degree of compartmentalization. These, among other particular reasons, are why these models were selected. Moreover, both models had been adapted to simulate the turnover of SOM in tropical and subtropical conditions.
Model parameterization remains the major obstacle to the use of these types of models on a routine basis in assessment and monitoring programmes. The demands placed on databases by the parameterization of the models make it necessary to split spatial variability of soil and climate into relatively uniform ecological zones, each of which requires its independent set of model parameters. Spatial interpolation and GIS modelling techniques play an important role in the zoning process and the generation of PCCs as part of the biophysical characterization of the study area. This step in data preparation is as important as the parameterization and running of the model itself and its relevance in the methodology cannot be underestimated.
Another important step in data preparation for running model scenarios is a set of ad-hoc calculations necessary to estimate the monthly contributions of organic matter by the cropping/land use occurring in a particular location. This report offers a procedure for the rapid calculation of such contributions, beginning from crop yields and estimates of contributions from aboveground biomass.
In order to achieve realistic results through modelling, the calibration of simulation models is of extreme importance. Calibration should be performed using sites independent to those in the study, with reliable laboratory analyses of SOC, which in turn should be predicted with satisfactory accuracy by the model. Input interfaces could be a problem for nonspecialist users of the simulation models. Both RothC-26.3 and CENTURY require interfaces. The current MS-DOS interface of CENTURY is particularly cumbersome and difficult. The model itself, when used in all its parameters, may require almost 600 variables. As RothC-26.3 already had a primitive graphic user interface (GUI), no further development was needed on this model. On the other hand, a GUI was developed for CENTURY. This was achieved by customizing the input module of the model through developing software code through Visual Basic. Through the customization effort, it became possible to reduce the number of necessary input variables to a handful. Assumptions were made in order to achieve this. Such assumptions are not considered unreasonable as parameters suggested by the model developers were taken into consideration and then built into the customization. Therefore, the model customization is considered sufficiently robust.
Linking the SOM turnover output of simulation models to GIS was a rather laborious task in terms of the number of linking operations required to complete the transfer and formatting of files. From this perspective, customization of CENTURY represented a considerable advantage in simplifying such model-GIS linkage. Further customization work is still necessary to make the link seamless and transparent to the non-specialist user.
Suitability assessment of LUTs with potential for carbon sequestration
In determining the PLUTs to evaluate as candidates for an LUC in the study area, it is necessary to consider criteria such as the potential for: increasing carbon sequestration, increasing crop yields, enhancing biodiversity, and decreasing potential impacts on land degradation. These and a high physical and economic suitability to the ecological and economic conditions of the area must all be applied as rigorously as possible in the selection process. A full land suitability assessment, including the opinions of farmers and local experts in a participatory fashion, is fundamental to achieving realistic results. C4 crops (photosynthetic pathway with assimilation rates of 70-100 mg CO2/dm2/h; crop groups III and IV) and C3 crops (photosynthetic pathway with assimilation rates of 40-50 mg CO2/dm2/h; crop groups II and V) are to be preferred in that order. Another key factor in determining carbon sequestration is the crop and land management activities that are part of the LUT selected. Land management practices are as important as, and could potentially offset the gains from, a careful selection of crop mixes in a new land-use plan. The mapping of suitability classes of PLUTs follows a standard procedure of allocation of attributes to land mapping units mapped as either classes of pixels or polygon class entities. These procedures are well known in GIS operations and should represent no technical problems for assessors implementing the methodology.
Regarding the PLUTs, where possible, the biomass of the cropping pattern should be predicted from well-calibrated crop growth models that estimate total biomass of crops and crop yields. The alternative to crop growth models is to estimate biomass and yield production from standard phenology-based equations. Such equations depend on crop species, simple climate variables and soil parameters. Estimates from the AEZ approach on expected constraint-free and constrained yields can be derived from suitability tables, where there are results from the application of an AEZ exercise in the area of concern (e.g. through the application of AEZWIN tools). Another option is to resort to records on average attainable yields in the area of study from experimental plots established in the area, where experimental information is available, or from official records or from farmers directly. From these data on biomass and yields, crop residue fractions, which are additions as inputs to local soils, should be estimated and verified against data from local farmer informants. The information about the contributions of organic matter from crops and vegetation to the soil is deemed crucial in the parameterization of the SOM turnover simulation models and the projected scenarios of SOC sequestration. Therefore, the estimation of crop residues and other organic inputs to soils should be verified carefully.
The scenarios of carbon stock and sequestration generated for the PLUTs required predictions of attainable biomass and yields and the predictions, through simulation, of the fate of organic C once in the soil. It is of considerable importance to verify such estimates in reality. This could be achieved by setting up validation sites in areas such as those of the case studies reported here, or in areas where the methodology had been applied. It is of extreme importance that follow up on the validation of the methods here proposed takes place in order to implement a set of test sites and studies, which could shed light on the accuracy of the methods proposed.
Estimation of biodiversity
Regarding biodiversity, it was found that concentrating on plant diversity was much more straightforward and brought faster assessment results than focusing on total biodiversity, including soil biodiversity with its micro and macro flora and fauna components. Therefore, plant diversity was considered a “proxy” for the other components of biodiversity. Later refinement of the methodology should examine the possibilities of including rapid assessments of other components of biodiversity, such as fauna and soil biodiversity, which is considered a key indicator and expected to be highly correlated to the accumulation of organic matter and C in the soil.
The variables used for assessing plant diversity were found appropriate to provide a picture of the status of plant diversity in the area. The indices used in the assessment (number of species, species richness, and species abundance) provided reasonably good information on the components of plant diversity. However, a fundamental problem was the in-situ identification of plant species. It is for this reason that a multidisciplinary team that includes a botanist is of key importance in the operational stages of the assessment. The use of local or indigenous knowledge concerning plant species and their distribution proved to be of key importance in this phase of the assessment. It is highly recommended that these resources be tapped into while implementing the assessment in the field.
Sampling by stages proved to be a good strategy, when observing the behaviour of the diversity indices’ curves as plotted against the number of samples (i.e. quadrat sites) already achieved. Increasing the number of samples on the ground until the curves became asymptotic to axis of the number of samples proved to be an efficient means of determining the optimal number of sampling sites on the ground. The customization of a database software program to allow for automatic entry of data and calculation of indices proved to be extremely effective in monitoring and achieving the optimal number of sampling sites.
Satellite image classification provided the appropriate geospatial framework (i.e. land cover classes) for interpolating and upscaling the calculated plant diversity indices at the quadrat sampling sites.
Follow-up work on this particular aspect of the methodology should incorporate rapid measurements or indicators of faunal diversity, and particularly of soil biodiversity, and their projected variations after LUCs.
Assessment of land degradation
The assessment of land degradation, carried out at the same quadrat site locations and during the field sampling and assessment of biomass and biodiversity, was based on a parametric semi-quantitative approach. This approach was found to be applicable and relatively straightforward as it uses data already recorded by inventory programmes of agencies in the developing world (e.g. climate and soil parameters). However, it was noted that, typically, there may be some data gaps that would require parameter estimation, particularly in the pre-field or initial stages. Nevertheless, these data gaps are relatively easy to fill as the variables used in the computation of the indices are commonly observed and recorded by national agencies in charge of natural resources.
The degradation assessment relied on four groups of individual indicators: soils, climate, topography and human activity. These were rated (weights) in their importance for determining land degradation and used to judge the status of each of physical, chemical and biological degradation of the land. This multivariate and multicriteria process was aided by the application of a synthesis process based on the “maximum limitation method”. This approach helped in calculating compound indices for each of the three types of degradation. In turn, these ratings were grouped into classes that reflected the degree of affectation of the land, and could be mapped as ratios attributed to land polygons in the legend of an overall land degradation map. Creating a map for each indicator, indicator group and type of degradation could be overwhelming and could create greater confusion as eventually a synthesis process is required to sum up the assessments of the different types. Thus, the derivation of ratios was found rather useful in the synthesis process. The diagnosis of the status of land degradation thus remained as quantitative as possible. The rating system provided by FAO (1978a) and the general framework of land degradation assessment proved extremely useful in the design of the approach and methods presented in this report.
The design and later use of pre-survey field forms facilitated considerably systematic assessments and data and observation gathering in the field. These forms also enabled more rapid data capture (digitally) and the computation of indices and indicators in the field and on the computer and with GIS.
The problem of upscaling or interpolating spatially these assessments made at the quadrat site locations is an important one. Techniques such as disjunctive kriging are worth exploring as they allow for the treatment of qualitative and semi-quantitative regionalized variables. This problem of “spatialization” of computations and estimates made at the site of the field quadrat was common to the three assessments made using this methodology. It remains an important problem that warrants further investigation in follow-up activities to this project.
Multicriteria decision-making
In terms of decision-making regarding the scenarios for LUC, the application of multicriteria techniques, particularly the AHP, proved useful in helping structure the decision-maker’s problem and in providing a mechanism for establishing priorities for the optimization of land-use scenarios. The AHP sequence for establishing priorities about land-use scenarios, followed by optimization with goal programming, appears to provide the tools and sequence of operations necessary to optimize the scenarios for LUC. Concerning the optimization process, the FAO approach using tools, such as AEZWIN, may prove fruitful in situations where specialized personnel, familiar with such a powerful program, can manipulate the software and data with relative ease. However, its use in the optimization of scenarios was found cumbersome for the nonspecialist user.
Case study results
In order to derive the full advantage possible from this discussion, the reader is advised to read the Appendix containing the case studies referred to in this section. They represent the practical experience of applying the suggested methods and procedures described in Chapters 2-8. This section concentrates on discussing the particular details of the suitability of the methods proposed to conditions of the sites studied separately.
The Texcoco River Watershed,Mexico
The Texcoco River Watershed is a narrow subwatershed, densely populated, with more than 120 000 habitants in an area of about 50 km2 (a narrow strip of 2.5 km × 25 km). Hence, the watershed has endured major human impacts on its resources. It has been occupied by human groups for millennia. In pre-Hispanic times, indigenous groups (e.g. Coluhas, Chichimecs and Aztecs) grew crops, harvested the forest and hunted on its lands. Thus, agriculture has been practised in some of these soils for millennia. The Spanish colonization saw a dramatic increase in resource exploitation: cattle ranching in the lowlands; crops on the lowlands and gentle slopes; and harvesting of the forests. The dramatic population increase in the last century has resulted in considerable degradation of resources, in particular: uncontrolled deforestation, soil erosion, and soil fertility depletion through intensive tillage, cultivation and mismanagement of soils. Urban encroachment on agricultural land and pastures and, in turn, of these on forest lands, has accelerated resource degradation. Related problems of water scarcity and pollution are also important in the watershed. Conversions of land to urban land use and rough terrain have further reduced the agricultural land to a narrow strip in the middle portion of the watershed and to the flat lowlands. A receding forest area is still of some importance at the higher elevations and steep slopes of the watershed in the sierra, towards headwaters. The issue of urban encroachment on agriculture and subsequently of this on forested land, and the impacts of human settlements on resources are perhaps the most crucial problems determining the fate of agricultural and forest land use in this watershed.
Maintaining soil fertility and SOM is a serious problem. Organic matter and plant nutrients in the soil are being “mined” out of the agro-ecosystems in the watershed. Crop residues are used as animal feed and not returned back to the soil. The shortage of pastures and fodder for animals is critical in the area. Thus, crop residues are much sought after and valued by farmers. This makes incorporation of crop residues into the soil an almost impossible practice as the market value of crop residues is relatively high.
The results obtained by running the SOM simulation models indicate that only when sufficient amounts of organic residues as inputs are present in the soil, together with sufficient moisture (most probably from irrigation or rainwater harvesting), does carbon sequestration occur in this type of ecological and cultural condition. At present, no land and crop management practices conducive to the accrual of C in these soils are applied in the watershed. Consequently, the scenarios computed make assumptions of incorporation of organic residues to the soil. Overall, only in a few LUTs could simulated scenarios show that by 2012 there would be a demonstrated sink of C in the soil.
Biomass estimates
As far as C in biomass is concerned, forest biomass has been seriously depleted in the watershed since the later part of last century. Forests are communal lands in the watershed. Therefore, forest management is inadequate with loose regulatory and enforcement mechanisms. Extractions by local farmers for fuelwood, fencing and other domestic uses (so-called “leakages”) occur regularly and uncontrollably. The estimates of forest biomass indicate that the forest is sparse in some areas, with relatively low density of biomass as trees. Moreover, forest extraction rates, which the communities in the watershed give in concession to a private forest products company for a fee, tend to be such that current forest management practices are unsustainable. The incidence of other compounding factors and random disturbances, such as forest fires, exacerbate the problem of biomass depletion.
Litter and wood debris are a relatively large component of the carbon sinks in the forested lands of the watershed. The C in these materials could be lost from these ecosystems by erosion or by degradation of the organic matter on exposure to agriculture or other forms of intensive resource exploitation.
It was found that there were no standard procedures for special situations of biomass measurement such as the dimensioning of cacti plants, e.g. Opuntia spp., known locally as “nopal”. These are very abundant and used as plot barriers in the agricultural area of the middle zone of the watershed. An in-situ procedure was developed to estimate the biomass of these plant species of rare intricate growth habit. This involved obtaining the average weight of morphological components of such species, and then counting the components present in each cactus plant to be measured.
The estimates of biomass in Texcoco indicate that the regression equation proposed by FAO (1999) was that best suited to the conditions of the watershed. Such an equation was found to be superior to the Winrock and Luckman equations in that it provided the closest estimates to those derived from morphometric measurements on the ground for the same sites. A formal error assessment, in the statistical sense, was not possible as no independent sites could be spared for use as “check sites”. For the analysis of the accuracy of estimates, only informal procedures based on rapid comparisons with ground morphometric measurements (based on volume measurements) were made. Therefore, this poses a fundamental question that needs to be tackled in the follow up on this project. A full study on error assessment of estimates for field validation and verification of carbon stock is the necessary and logical next step in refining the methodology proposed here. The conversion of biomass to carbon stock was automatic, by using a conversion coefficient. Another important observation concerning aboveground and belowground biomass estimation is that standing biomass estimates were only calculated for those mapping units with perennial vegetation, or with vegetation present at the time of measurement. This is reflected in the distribution of biomass in the maps produced as little or no biomass could be measured for most agricultural land owing to the fact that the fields were in fallow. Therefore, crop biomass was simulated for PLUT. Estimates of crop residues were used as inputs to models of SOM turnover.
Land suitability for PLUT for carbon sequestration
The land suitability assessment in Texcoco indicated that the staple crops, i.e. maize (cereal), beans and squash, are only marginally suitable in most of the area of the watershed. This lends support to the evidence of only marginal yields being obtained in the area. However, in the flat lowlands of the watershed these crops have moderate suitability owing to the presence of deeper soils and of irrigation with groundwater. Other crops not currently grown in the watershed, such as sunflower, and a mix of horticultural crops (carrots, onion, lettuce, etc.) were found to be moderately suitable. A common limiting factor in the most suitable soils in the lowlands was the level of salinity and sodicity as the soils were closer to the dry lake bed. However, the most suitable crop pattern consisted of maize, barley, beans, onions and other horticultural crops. These were part of the scenarios of carbon sequestration developed through simulation.
SOM simulations with CENTURY
As far as the SOC pool is concerned, the scenarios of SOC dynamics, generated through simulation with the crops selected from suitability analysis with the CENTURY model, allowed for the elucidation of patterns of variation over space and time in the watershed under “business-as-usual” management. These scenarios show that there is a trend in terms of the spatial distribution of SOC dynamics in the watershed. Carbon losses to the atmosphere tend to increase with slope, altitude and terrain roughness, that is, with marginal conditions, in the agricultural portion of the watershed, and for the staple crops. The soils in the upper agricultural areas with moderate to strong slope, near the edge of the forested area, are experiencing the highest carbon losses (in a period of 12 years). These findings are somewhat intuitive as the SOM of formerly forested areas is degraded rapidly by the incorporation of relatively new agricultural areas to tillage. On average, these soils would lose up to 1 060 tonnes/ha of C over the 12-year period for maize, and 1 151 tonnes/ha for the same period for beans, assuming that the same crop and land management (“business as usual”) prevails in the same land in the 12-year period. This is so in spite of terracing and other soil erosion control measures present in the same area of the watershed, and of chemical fertilizer applications. The soils are thin, already affected by erosion, and depleted of SOM by tillage, and above all, by the lack of applications of crop residues to the soil. The high radiation and dry conditions, and the export of organic matter from such ecosystems as crop residues for animal feed, create the conditions for important carbon losses both to the atmosphere and as carbon “mining”. According to the results of the modelling, the carbon losses in such agro-ecosystems occur mainly from the “slow” and “active” pools of SOM. The “passive” pool of SOM remains almost unaffected. Except for the redistribution of a small fraction of SOM among other pools, most carbon losses are to the atmosphere as CO2.
In the middle portion of the watershed (rainfed agriculture on gentle slopes with ravines and hill slopes), similar carbon losses (averaging 1 160 and 1 163 tonnes/ha in a 12-year period for maize and beans, respectively) occur under “business-as-usual” management. The only exception to this trend for this part of the watershed is alfalfa cropping under limited irrigation. This scenario would require the application of limited amounts of irrigation and incorporation of realistic amounts of crop residues to this LUT. Carbon sequestration in soils of this middle part of the watershed under alfalfa cropping for the 12-year period occurs to a total amount of 3 381 tonnes/ha (scenario “Sn08Af”). A plausible explanation for these results lies in accounting for the C and N interactions, which the CENTURY model simulates well. The same also holds for the role of legume plants such as alfalfa in symbiotic Rhizobium fixation of atmospheric N, thus enhancing carbon sequestration. Hence, an increase in the area grown with alfalfa or similar legumes, whether in association or in rotation with cereals and fruit trees, would tend to enhance conditions for carbon sequestration in these soils of the watershed. There is also reason to believe that such carbon sequestration could be enhanced by conservation agriculture strategies, such as no tillage, mulching with crop residues, legumes and green manures, and conservation of soil moisture.
In the flat lands at the lowest portion of the watershed, near the town of Texcoco and towards the edge of the agricultural area with the dry lake bed, the CENTURY-simulated scenarios showed a more positive trend towards carbon sequestration. These are deep, fertile alluvial and former lacustrine soils with moderate SOM content. The SOM is of pedogenetic origin (i.e. the lake bed) and also the result of contributions of nutrients from sewer, sludge and other residual waters from the upper portions of the watershed. These lands are irrigated by groundwater and by a mix of sewer and storm waters coming down from the communities at higher altitudes in the watershed.
The staple crops show only moderate to low carbon losses in the 12-year period under conventional management (averaging 191 and 1 021 tonnes/ha for maize and beans, respectively). In contrast, alfalfa (and possibly other legumes) together with vegetable crops, show a higher CSP. In the 12-year period, alfalfa in these flat lowlands could sequester an average of between 3 132 and 4 775 tonnes/ha with enhanced management (i.e. moderate irrigation, incorporation of crop residues and animal manures, and minimum tillage). The sequestration potential for horticultural cash crops, particularly legumes, in rotation or association with other crops could be as much as half the amount of carbon sequestration achieved with alfalfa under irrigation. The major increases occur in the “slow” (SOM2C) and “active” (SOM1C) pools of SOM, with moderate increases in the “passive” pool of SOM (SOM3C), to the rate of 40-70 tonnes/ha/year.
Cultivated grasslands, modelled by CENTURY on this portion of the watershed, could not yield comparable results to those positive carbon sequestration results obtained for alfalfa.
The soils in the primarily agricultural zone in the middle portion of the watershed are Phaeozems (Haplic) and Cambisols (Distric and Eutric) with significant inclusions of Lithosols and Regosols. These soils are thin, with the presence of a hardpan layer of 10-40 cm in depth (“tepetate”), and have low organic matter content. The intensive use to which they have been subjected has led to important erosive processes consisting mainly of water and wind erosion. Only the Mollic Andosols in the forested areas of the upper sierra zone of the watershed are of moderate to high fertility with sufficient SOM. Other soils of the flat lowlands, downstream near the dry lake, are affected by salinity and sodicity.
The climate in the watershed does not create favourable conditions for carbon sequestration in soils. The evapotranspirative demand exceeds precipitation for more than seven months of the year. The precipitation is not well distributed throughout the year. Therefore, soil moisture protection from radiation and the accumulation of SOM is not possible. Thus, management practices that aim to protect SOM (e.g. mulching and no tillage) are to be encouraged.
It was possible to generate a scenario of carbon sequestration under staple crops (maize and beans) and alfalfa (already cropped in the area) in the soils of the middle gentle slopes and lower plains of the watershed. In this scenario, maize achieves 40-50 tonnes/ha of C as SOM for 12years for soils in the middle portion of the watershed under rainfed conditions, provided that 20-30 tonnes/ha of organic inputs (crop residues, manures, etc.) are applied yearly to these soils. For the Vertisols and alluvial soils in the lowland plains, 50-101 tonnes/ha of SOM can be sequestered with the same inputs of organic materials, but with the added organic inputs from residual (sewer and storm) waters as irrigation sources. In such a scenario, alfalfa accrues 47-80 tonnes/ha of C as SOM but with minimal amounts of organic inputs as crop residues and simply with the organic additions from sewer water irrigation.
According to the CENTURY simulations, these scenarios indicate that it is possible to achieve carbon sequestration in soils. However, considerable amounts of organic inputs (i.e. crop residues, manures or others) into the soil are necessary to achieve this. Such inputs are not currently applied to such soils as part of land management in that part of the watershed, and it is difficult to visualize farmers not deviating the crop residues from their fields in order to feed their livestock.
SOM simulations with RothC-26.3
The scenarios created from the simulations of SOM turnover by the RothC-26.3 model showed the existence of a spatial and temporal pattern of C in the soils of the watershed. Carbon sequestration occurred in the alluvial soils and Vertisols of the flat lowlands of the ex-lacustrine zone under irrigated horticultural crops, specifically onions. All fractions of SOM showed increased values after 12 years. The resistant fraction (RPM) accumulates 8 tonnes/ha whereas the humic fraction (HUM) accrues almost 6 tonnes/ha. There is an increase in BIO of 1.2 tonnes/ha whereas the active fraction remains almost unchanged, and about 55 tonnes of C as CO2 are lost to the atmosphere in the 12-year period. The same LUT sequesters C in soils (Haplic Phaeozems, Eutric Cambisols and Regosols) of the middle portion of the watershed on gentle slopes of hillsides and ravines. The compartments of SOM that increased in the 12-year period are RPM with 8.02 tonnes/ha, HUM with 6.35 tonnes/ha, BIO with 1.165 tonnes/ha, producing 54 tonnes/ha of CO2 for the 12-year period. Although they may seem small, these figures on carbon sequestration should be viewed against the background of dry climate, degraded soils and land and crop management strategies that are not conducive to carbon storage in soils. Moreover, they result from the application of assumed minimum management improvements, such as small additions of FYM. Another LUT in the middle zone of the watershed, irrigated beans, shows accumulation of C in SOM in the humic fraction (HUM = 12.22 tonnes/ha) in the 12-year period, where slight improvements in crop and land management are assumed. However, the CO2 losses amount to 105.5 tonnes/ha for the same period.
Even where staple crops such as maize, beans and squash are grown under rainfed conditions in the middle lands of the watershed, they show small accruals of certain fractions of SOM. For example, for squash, the accrual of BIO is 0.043 tonnes/ha, whereas that of RPM is 1.32 tonnes/ha. For rainfed beans, there are 1.06 tonnes/ha of the humus fraction accumulated in the period. These accumulations occur when improvements in management (amounting to small additions of crop residues) are assumed. In other instances, there is a redistribution of C within the different fractions of SOM. That is to say, losses in one SOM fraction end up as gains in another. For example, for irrigated beans in the middle zone, there is a slight humification (HUM) of BIO, or of BIO and RPM. However, for all other crops in the proposed LUT in the flat lowlands and the middle zone of the watershed, all initial contributions of crop residues end up as emissions to the atmosphere owing to land management.
In the upper watershed, consisting of degraded Eutric and Dystric Cambisols, Lithosols and Haplic Phaeozems on sloping lands of the transition between the middle and the sierra zone, a squash crop reports depletion of BIO, which increases CO2 emissions. For all other crops in this agricultural fringe, there are only carbon losses to the atmosphere as CO2. This can also be interpreted as part of the degradation of the initial organic matter in past forest soils open to cultivation for about half a century (first half of the twentieth century) in this watershed. The only LUT that shows evidence of carbon sequestration in this upper zone of the watershed is a forest balsam fir (Abbies religiosa). This area is an area of rich Humic Cambisols and Mollic Andosols with reasonable amounts of litter on the soil surface on steep slopes of the sierra zone. In these soils, RPM increases by 12.81tonnes/ha and BIO by 0.06tonnes/ha. However, the humic fraction decreases by 1.29tonnes/ha in the 12-year period. This indicates that the humic fraction is being attracted by the BIO, converting it to a resistant (sequestered) fraction. About 55 tonnes/ha are lost to the atmosphere in this process during that period.
Finally, as with CENTURY, it was possible to generate scenarios that sequester C in the watershed from RothC model simulations. On the gentle slopes of the middle of the watershed, the scenario shows that rainfed maize can achieve 48-56 tonnes/ha of C as SOM, but only if 20-30 tonnes/ha/year of organic inputs are applied. Rainfed beans achieve 15.3-20 tonnes/ha of C as SOM with similar organic inputs for the same soils. On the other hand, in the lower plains of the watershed, maize with irrigation from a mix of sewer and storm waters can achieve 130-153 tonnes/ha of C as SOM with the same crop residue inputs. For the same area and soils, alfalfa irrigated with sewer and storm waters achieves an accrual of 40-60 tonnes/ha of C as SOM.
Comparison of results from SOM simulations with CENTURY and RothC-26.3
Judging from the results of the Texcoco case study, it can be said that the similar or comparable trends in the spatial and temporal distribution of carbon dynamics in soils can be derived from the simulations by both models. CENTURY provides a greater number of compartments and recognizes a larger number of fractions and pools of SOM than does RothC-26.3. Therefore, it provides a more detailed breakdown of pools and of the soil processes involved in the fate of organic inputs into the soil. This far greater detail comes at a price. The number of input variables necessary to fully parameterize the model could potentially be in the hundreds (about 650 in total). The number of output variables is similarly large. A situation may arise where the model user does not request variables on output that would help to explain the partition of forms of SOM and to account for the fate of the total of organic inputs, as the number of possible output variables are so many. Another and perhaps more serious problem is that some of the input variables are not commonly recorded, measured or observed. For example, fractions of lignin in types of tree foliage is a variable that ecologists with a definite research purpose may collect data on, but it cannot be expected in conventional studies of vegetation or soils conducted by agencies.
Concerning this problem, the customization of the CENTURY model undertaken through this project (“Soil-C”) could become a rather useful tool to guide non-specialist users to carry out simulations with the model, starting from relatively ordinary data sets for soil, climate and vegetation/crop management.
For answers to simple questions, such as how much CO2 is lost to the atmosphere by a given LUT and land management practices after a finite period of time, or the amount of resistant organic matter left in the soil after the same period, it may be better to use the RothC-26.3 model. However, there are no known interfaces to link this model to standard GIS software in order to generate spatial scenarios. Moreover, the RothC model does not simulate the C-N interactions (which are so important in the attack of SOM by microbial activity and in its fate) as well as and in the detail achieved with CENTURY. Therefore, where accuracy of results and a greater level of detail in the fate of SOM are required, it may be better to use CENTURY or a similar model, with the added advantage of a computer interface that would facilitate its use.
An important consideration is that both models, and indeed any simulation model, should be well calibrated to the conditions in which they are applied, with experimental data from the area. This could only be done in this project in one case study (Bacalar, Mexico), for which there were data for calibration and validation of the model. Further studies ought to be conducted in the region (Latin America and the Caribbean). These should collect data from any long-term experiments in the area, and set up such experiments for a variety of ecological zones and conditions in order to determine the range of conditions for the applicability of such models in Latin America and the Caribbean.
Biodiversity assessment
Related to the spatial distribution of biodiversity (in this case plant diversity) in the Texcoco watershed, the indices calculated and mapped out showed that the middle portion of the watershed is characterized by the highest values of species richness and fewer “abundant” species and more “rare” species. This indicates that these classes are not dominated by a few species, which is confirmed by the high values of the Reciprocal Simpson Diversity Index and the Shannon Diversity Index in the area. This could be explained in terms of the introduction of horticultural species in backyard intensive plots and the greater richness of weed species. Hence, the lower the altitude of the rainfed agricultural plots in the watershed, the more diverse the plant population becomes, with the exception of the intensive mechanized monocropping of irrigated alfalfa and maize for forage, which tends to be grown on the lower and alluvial plains at the bottom of the watershed.
In summary, the results suggest the possible existence of a pattern in terms of the distribution of plant diversity and vegetation classes (landscape element types) throughout the watershed representing different ecological conditions and different stages of forest succession (in the upper part of the watershed). The older the stage of forest succession, the more species there are and the less is the dominance of any given species. Thus, losing these agro-ecosystems in the middle portion of the watershed (e.g. through urban encroachment) and those forests of the upper portion of the watershed (through agricultural encroachment causing deforestation) would be more costly in terms of plant biodiversity than anywhere else in the watershed.
Land degradation assessment
The three types of land degradation assessed in the study, chemical, physical and biological, also show a spatial pattern across the watershed. The lands most chemically degraded, owing essentially to salinity, alkalinity and possibly to the presence of heavy metals in excess (sewer water irrigation), are those located at the bottom of the watershed, near the edge with the salted lake bed. Elsewhere, the lands of the watershed show only slight signs of chemical degradation at most. However, as no analytical work was part of the assessment, no conclusive evidence of other possible sources of chemical degradation could be detected (e.g. pesticide in soils). Thus, the lower the altitude the greater the chemical degradation in this watershed.
As far as physical land degradation is concerned, soil compaction was considered moderate to high in the low plains, owing to the intense use of farm machinery and tillage, whereas in the rest of the watershed it was found to be only slight. The middle and upper middle portions of the watershed, which are the areas of rainfed agriculture, are the areas more affected by both aeolic (10-50 tonnes/ha/year and even 50-200 tonnes/ha/year of soil losses in some areas) and water erosion (10-50 tonnes/ha/year in most of the middle part and all of the upper part of the watershed, except for a few areas in the middle part, which lose 50-200 tonnes/ha/year of soil), as estimated from both experimental runoff plot data and from applying the USLE. These patterns are realistic and coincide with the empirical field observations on the state of soil erosion, which persists at such rates in spite of terracing and other conservation measures.
Finally, with regard to biological land degradation, it became clear that, given the ongoing processes and the current land management practices, and except for one or two landscape units, the lands in the entire watershed are affected by high to very high biological degradation. This is mainly because of the gross decline in the levels of organic matter in the soil and the environmental, cultural and economic conditions in the watershed. These conditions will ensure the steady decline of SOM unless strong remedial actions (e.g. dramatic increase in crop residue inputs, mulching, no tillage, soil conservation and water harvesting) and different land management practices are adopted. The land-use patterns and PLUTs suggested in the scenarios should involve the application of such improved land management practices.
The tropical forest lowlands of Bacalar, Quintana Roo,Mexico
This study area can be thought of as the antithesis of the study site at Texcoco in terms of vegetation type, vegetation density and agricultural activity. This study site also represents a very contrasting situation to Texcoco in terms of population density and, therefore, in terms of anthropic impacts on ecosystems. In Bacalar, there are no permanent residents in the study area, and the surrounding areas are sparsely populated. By contrast, Texcoco has a very high population density. The main socioeconomic activity consists of subsistence farming and of forest- and agriculturally-based family production systems. The staple foods produced in these systems are a result of slash-and-burn agriculture (SABA) or shifting cultivation. This type of use of the forest introduces the most dominant dimension in the carbon dynamics in this part of the world: the management of forest biomass. The use of forest biomass, at different stages of succession, to boost soil fertility through the ashes of the burned forests has enabled crop growth and yields to sustain Mayan populations for millennia. However, much C has been lost to the atmosphere as CO2 in the process.
The soils of the Yucatan Peninsula are naturally very shallow, and are formed by the in-situ weathering of the emerged thin layer of marine sediments on the limestone shield. In the lowlands and plains, the soils are Luvisols and Cambisols, and they can be described as relatively deep red soils. On the small hills and slopes of the undulated landscape, weathering of the shield and its residual materials has created Rendzinas and Lithosols, soils with fine textures and varying degrees of coarse fragments, stoniness and rock fragments. These soils can only sustain crop growth and crop yields through a rapid recycling of organic matter. The nutrients from the ashes of a 40-50-year-old forest incorporated into the soil can sustain crop yields for three to five years. Then, as yields decline, the agricultural plot has to be abandoned and moved to a new patch of mature forest, which is cleared for growing staple crops (maize, beans and squashes).
There is a strong correlation between soil type and position on the landscape: Rendzina (called Bosh-lu’um in the Mayan soil classification) and Lithosols (Tzek’el) at the top of small hills; Regosols (Ho’l lu’um and Chich lu’um) on the slopes of undulations; and Luvisols (K’ankab) and Cambisols (Chack lu’um) on the small plains and flat areas of red and reddish-brown soils. Gleysols or Vertisols (Ak’alche) could replace the latter depending on the degree of hydromorphism of the soil in flat lands, savannahs and concavities, where periodic flooding may occur. The Maya soil classification, derived from millennia of experience and empirical knowledge, recognizes such soil-landscape associations. These associations are extremely important as they can determine vegetation succession and, therefore, the carbon stock as biomass in such forests. Hence, forest biomass and its carbon stocks are related closely to vegetation classes, as recognized by the indigenous knowledge: Monte Alto, Kelenche, Juche, Akalche, Sabanna and Canada.
Biomass estimates
The estimation of biomass was complicated by the sheer amount of biomass in terms of the very large number of trees to measure, the amount of biomass and the undergrowth, shrubs and saplings in a sometimes impenetrable tropical forest. Substantial aboveground carbon stocks in the biomass of tropical forests characterize the Bacalar case study area.
The measurements at the quadrat sites took several orders of magnitude in time more than the time taken to measure quadrats in a more open temperate forest, such as in Texcoco. This is a factor to consider in terms of its impact on time, effort and costs to obtain the ground measurements for the estimates of biomass.
The regression equation method, based on biomass as a function of measurements of volume, proved to be the most adequate method for estimating the biomass of the type of tropical forests in Bacalar in the Yucatan Peninsula. The Bacalar study area receives about 1 000 mm of precipitation per year and is characterized as a lowland area. Thus, the results using biomass regression equation FAO-1 for “dry” tropical areas receiving more than 900 mm of annual precipitation (FAO, 1997) were found to be the most suitable method for predicting biomass as a function of measurements of volume.
The vegetation types recognized by the Maya, and mapped out through the supervised classification of a Landsat TM multispectral image of the area, proved extremely useful as both a frame for sampling design and as a mechanism for upscaling estimates calculated at quadrat sites to the entire area. A close relationship between the radiance values of pixels in the multispectral image and the biomass estimates from quadrat sites on the ground was found and expressed in terms of a regression equation. This equation allowed for the conversion of the digital values of a satellite image into values of biomass at each of the pixels. Where based on a good regression model fit, this type of “biomass transfer function” can be extremely useful in generating estimates of carbon stocks in biomass of dense tropical forests such as those of the Yucatan Peninsula. However, it is doubtful that a similar approach would work as well in less densely forested areas. The map of forest biomass obtained through the satellite image shows a pattern of distribution of values of biomass, which followed closely the distribution of classes of vegetation as from the supervised classification of the image in terms of the Mayan classes.
However, the degree of accuracy of biomass and carbon stock estimates in the tropical forests of Bacalar is unknown. Their accuracy can only be determined through independent error assessment in a follow-up study to this project. Indirect indicators of the accuracy of estimates can be the value of the coefficient of determination (R2) and the standard error of the regression models used.
The biomass in forest litter and debris in Bacalar forests is considerable and contributed significantly to the inventory of carbon stock. Where the layer of litter is removed, as in deforestation for SABA, considerable losses in biomass occur through burning and through degradation of organic matter. Forest litter was found to play a crucial role in understanding the dynamics of SOM turnover in shifting cultivation.
Scenarios generated by SOM simulation models
In Bacalar, it became imperative to use a slightly different approach to carbon accounting in SOM from that used in the other two case studies. Instead of considering individual LUTs one at a time, and evaluating them in their potential for carbon sequestration, a “farming systems” approach was deemed appropriate to describe the flows of energy and materials involved in all productive activities in the family unit production system, as such flows are much more dynamic in this area. The approach would help in accounting for flows of organic materials and “leakages” to other subsystems (e.g. from forest to agriculture, and from this to backyard poultry and livestock, and back to the agricultural plot as FYM).
Another consideration concerned the unique characteristics of SABA. In Bacalar, the most effective measure to reduce CO2 emissions from forest burns was to consider the stabilization of SABA from shifting cultivation to continuous cropping in the same land plot. This would prevent deforestation through forest slash and yearly burnings of numerous patches of forest. In the family unit farming system, the representative subsystems accounted for, through their contributions of organic inputs through modelling, were: the felled forest of the agricultural plot, crop residues from the previous cropping cycle, the family backyard orchard, and FYM from livestock (including backyard poultry).
With food security as an objective, the modelling scenarios aimed at determining the management of organic inputs to the land growing a mix of staple crops (maize, beans and squash) that would guarantee sustaining or increasing yields, while sequestering enough organic C in the soil, over time, so as to make the system sustainable.
After model parameterization, model calibration was achieved with the only five years of SOM measurement data available. For example, the fit of the RothC-26.3 model predictions of SOM to the actual SOM values measured is quite good, and lent confidence to the use of these models in spite of the limited number of data points for the calibrations.
The scenarios computed for staple crops in continuous cropping with a variety of organic matter inputs from different sources (i.e. subsystems of the family unit farming system) show that in all of the scenarios there are C losses and no carbon sequestration in the soil. The amount of C lost to the atmosphere is related to management. Carbon sequestration only begins to occur after continuous cropping of staple crops, when the amounts of organic matter as inputs to the soil are high enough to provide substrate to begin the further accumulation of SOC in its different compartments or pools. It was demonstrated that the management of organic residues is crucial as a carbon sequestration strategy. It was shown that carbon sequestration (as measured by total C in the soil) only occurs under careful land management. Thus, after investigating various organic matter inputs into the soil as recommended by various workers in the specialized literature (e.g. Lal et al., 1998; Sanchez et al., 1989; Szott, Fernandes and Sanchez, 1991), it was found that scenario SK15, which uses sources characteristic of the SABA occurring on the Yucatan Peninsula, was the best scenario in terms of carbon sequestration. The plots are left fallow for one year, a SABA event, followed by continuous cropping with annual FYM inputs. The total carbon inputs for this scenario are 3.3 tonnes/ha from the SABA event for the first year only, followed by 5.39 tonnes/ha from the cropping residues, 20 tonnes/ha from orchard residues and 25.58 tonnes/ha in FYM annually.
The production of crop yields as a function of SOM developed by regression analysis in Bacalar allowed for the estimation of yields as a function of the SOM simulated by the models. These computations permitted the generation of predictions of crop yields values into the future scenarios of carbon sequestration, thus providing a measure of food security. It was found that with the SK15 scenario, the sequestration of C in SOM was such that it would allow for the production of enough maize grain to sustain a family of six individuals.
According to RothC-26.3 simulations, the C sequestered in the Chromic Luvisol soils (Kan’kab) under Juche vegetation and past cropped areas (“canyada”) changes from 730 to 740 tonnes/ha of C in SOM. This represents a gain of about 10 tonnes of C in SOM per hectare in a 12-year period under continuous cropping and assumed organic matter inputs as in the scenario SK15. It is not known what the practical requirements are in order to achieve the levels of organic inputs demanded by the SK15 scenario. Nor are the practical constraints imposed on farmers in the area known. However, it is clear that the carbon sequestered would lead to sustainability of crop yields and to food security for thefamily.
A comparable simulation of this scenario SK15 with the CENTURY model produced similar results. However, the values of C in SOM were more modest. The changes occur between 116.29 and 121.35 tonnes/ha of C in SOM; a gain (sequestration) of 5.06 tonnes/ha of C in SOM for the red Kan’kab soils.
The “business-as-usual” scenario for the 12-year period involved changes in total C in soils such that all systems, except for SK15, behaved as net emitters. For tall and old (30-50 years) forest (Monte Alto and Kelenche) on thin Lithosols and Regosols, 500-550 tonnes/ha of C were transformed to 350-400 tonnes/ha of C after SABA during the 12-year period. For a young successional forest (i.e. Juche), the values changed from 150-200 tonnes/ha of C to 50-100 tonnes/ha of C on the relatively deep Chromic Luvisols (Kan’kab). Where the same type of young successional vegetation (Juche) is found on thinner soils such as Regosols or Lithosols, then the losses are larger than even for the old forest: from 500-550 tonnes/ha of C, to 300-350 tonnes/ha of C in SOM.
Finally, a scenario which includes the addition of only 2 tonnes/ha of C as FYM to the soil under maize-beans and squash cropping and modelled by RothC-26.3 did not have any effect on enhancing carbon sequestration in Chromic Luvisols (Kan’kab). This is because it depleted C as SOM, which changed from 150-200 tonnes/ha of C to 100-150 tonnes/ha of C. This indicates that it is not worth adding organic materials to the soil unless the quantities are sufficient as to create a substrate of SOM to begin sequestration in soils, as in scenario SK15.
Biodiversity assessment
In order to discuss the results of evaluating biodiversity in Bacalar, it is worth reviewing the key concepts that have been applied in calculations of indices.
Intuitively, biodiversity, or species diversity, is understood as the number of species in a given area, habitat or community. However, the formal treatment of the concept and its measurement is complex.
A biodiversity index “seeks to characterize the diversity of a sample or community by a single number” (Magurran, 1988). The concept of “species diversity” involves two components: the number of species, or richness; and the distribution of individuals among species, or evenness. Most indices try to encompass both of these dimensions. Many of the differences between indices lie in the relative weighting that they give to evenness and richness.
The simplest measurement of species diversity is a species count. Simple species counts remain the most popular approach for evaluating species diversity and comparing habitats and species assemblages. Species counts have proved to be a very useful index in areas that are not densely vegetated, such as in the Texcoco or Cuba case studies. However, in densely vegetated areas, such as in Bacalar, they are often considered an early step in many ecological and community studies. The number of species per se provides little insight into the underlying ecological mechanisms that define biodiversity, nor does it encompass evenness. Species counts are insensitive to the ecological placement of species, including rare species, that may be present in tropical forests. For example, species counts in a forest would equally consider species with totally different ecological roles, such as trees and herbs. Species richness provides an extremely useful measure of diversity when a complete catalogue of species in the community is obtained (Magurran, 1988), and this was not the case for Bacalar. Thus, computation of the other indices of biodiversity to complement species richness became an important step in the assessment in this case study.
Concerning the number of species, there are clear spatio-temporal trends in the Bacalar study area. A clear association exists between age of forest succession and species richness. Thus, a spatial pattern emerges with forest succession age. For example, the Monte Alto vegetation class, which corresponds to the oldest forest succession, holds the greatest species richness (more than 50 species), followed by Kelenche (30-50 species), the second oldest forest succession class, Juche (10-30 species), which is the third oldest vegetation succession, etc. The vegetation areas with the lowest species richness are areas with intermittent flooding, such as Akalche and Sabana, where species adaptation to this type of flooding stress has played a selective role.
To a certain extent, the spatial distribution of these vegetation classes represents the distribution of species richness in the area studied. Hence, affecting any of these classes by deforestation would bring implications of possible losses of such species richness on a given patch of land, and considerations about the understanding of the role of SABA on species succession and ultimately on species diversity.
The fact that there is a strong association between species richness and vegetation classes, and that such classes could be recognized and discriminated through multispectral satellite image analysis, made it possible to map species richness through a conventional supervised classification of an FCC of the satellite images.
Perhaps the most widely used index of species diversity is the Shannon Diversity Index. The Shannon Diversity Index is very similar to the Simpson Diversity Index except for the underlying distribution. The Simpson Diversity Index assumes that the probability of observing an individual is proportional to their frequency in the habitat, while the Shannon Diversity Index assumes that the habitat contains an infinite number of individuals. The equation for the Shannon Diversity Index is:
H = S (pi - ln (pi))
where pi is the proportion of individuals ni of species i in the total sample N. That is, pi = ni/N. This index considers both the number of species and the distribution of individuals among species. For a given number of species s, the largest value of H results when every individual belongs to a different species, that is pi = 1/n, which allows for a relative measure of diversity, “evenness”: E = H/ln(S). However, comparisons among communities or habitats based on E are possible only where the sample size is the same. The value E or “evenness” is a measure of how similar the abundances of different species are. Where there are similar proportions of all species, then evenness is one, but where the abundances are very dissimilar (some rare and some common species), then the value decreases. Evenness (E) has a high value where there are equal numbers of individuals in each species.
In the Bacalar case study, the calculated “evenness” for each vegetation class mapped showed a split in terms of evenness values down half of the vegetation classes and vegetation succession. Values closest to one were observed for the oldest stages of succession, i.e. Monte Alto, Kelenche and Juche. These values of evenness for such classes are in descending order, but very close to one another. This would indicate that after a period of recovery from SABA, equivalent to the age of a Juche (10-20 years), most of the species that would establish themselves in succession have already established themselves in comparable numbers of individuals. Hence, the transition from Juche to Kelenche does not bring more “rare” or new species, nor does the transition between Kelenche and Monte Alto, thereby indicating that the vegetation reaches some degree of equilibrium.
In contrast, Saakab and Akalche classes showed the lowest evenness values, indicating that these classes not only have the lowest richness, but also the most “rare” species, typical of wetlands or lands experiencing intermittent flooding. In this respect, the Sabana class can be considered an “anomaly”, given the fact that it showed the second largest evenness values after Monte Alto.
The Simpson Diversity Index measures the sum of the probabilities that two randomly chosen individuals belong to the same species, summed over all species in the sample. The Simpson Diversity Index assumes that the proportion of individuals in an area adequately weights their importance to diversity. The equation for this index is:
D = 1/(S (pi2))
where D is the diversity and pi is the proportion of the i-th species in the total sample. This can also be written as:
D = 1- [S ni (ni -1) / N (N - 1)]
where ni is the number of individuals of the i-th species, and N is the total number of individuals. The value of D varies widely as the total number of species increases, depending on the type of species-abundance relationship used to calculate the index. This index goes from zero to the total number of species. An index of one indicates total “dominance” of one species, that is to say, that all of the individuals in the area belong to a single species. Where D = S, then every individual belongs to a different species. The Simpson Diversity Index is a commonly used dominance measure because it is weighed towards the abundances of the most common species rather than providing a measure of species richness.
The values of the Simpson Diversity Index computed for Bacalar show that in the oldest successional stages of the tropical forest (i.e. Monte Alto, Kelenche and Juche vegetation classes) a few species tend to dominate, even though the number of individuals in each of such species is quite even. The values of “dominance” approach 0.90. In contrast, the index values for the vegetation classes affected by flooding remained with less dominance of the species found in such habitats.
Other indices to measure species diversity have been proposed, but they have received little attention or are mathematically related to the more popular indices (H and D). However, one common characteristic of biodiversity indices is their requirement for statistically sound sampling. Sampling for species richness requires appropriate area and seasonal coverage in order to ensure that the sample includes a significant subset of all species. Indices that include both species and individuals/species data for their calculation require more intensive sampling. In the light of these observations, the procedures advanced in this methodology provide a picture of the status of plant diversity in the area of concern. This may require further investigation in order to understand the underlying ecological processes leading to the distribution of species in the area and the diversity they represent.
Sixty-seven quadrat sites were sampled in Bacalar for the purpose of estimating the biodiversity indices. This number was considered adequate for the purposes of providing an initial assessment of the status of biodiversity, in this case plant diversity. This was confirmed by the shape of the curves of each of the three biodiversity indices calculated, when plotted against the number of samples (i.e. quadrat sites). At about 65 samples, the curves of biodiversity indices became asymptotic to the horizontal axis of the number of samples, which was considered an adequate compromise between information and costs.
Land degradation assessment
Land degradation in the Bacalar case study was reduced to biological land degradation, as detected by early observations from field quadrats. The absence of anthropogenic impacts on these forests and lands, other than SABA, owing to the rather low population densities, makes the presence of chemical degradation almost undetectable. There are reported uses of herbicides and pesticides together with chemical fertilizers. However, there are no records on the application rates of such chemicals and on the effects they may have on the status of the land.
Physical land degradation, particularly soil erosion by water and wind and compaction due to machinery, was considered negligible on these lands. The high forest protective cover and the microtopography and shallowness of the soils limits the use of machinery for tillage to the point of having negligible effects on the land. Therefore, for the purposes of this case study, both chemical and physical land degradation were considered as being negligible. In contrast, organic matter plays a key role in determining the productivity of these lands, to the point of obscuring any other possible types of degradation. SOM and the rather tight cycling of nutrients from vegetation in these shallow and stony soils play a key role in determining the primary productivity of these lands, the types of vegetation succession, and crop growth and yields. Thus, only biological degradation was considered significant and evaluated in this case study through the decline of SOM in the soils of the area studied.
As far as the decline in organic matter is concerned, the state of biological degradation was assessed through the scenarios of future SOM turnover. Through these, it became clear that careful selection of crop mixes and additions of organic matter from different subsystems would create the conditions for carbon sequestration and accumulation of organic matter over time.
The Rio Cauto Watershed,Cuba
The results obtained from the Rio Cauto Watershed in eastern Cuba confirmed some of the findings from the two Mexico case studies. However, some other results were unique to the Cuba case study. There are major differences in the decision-making processes concerning land use and LUCs in Cuba compared with the other two case studies. Central planning has produced a land-use pattern that is less complex and more straightforward than in the Texcoco case study. There is continuity and larger areas in the mapping units of the land cover map generated in the Rio Cauto Watershed owing to the allocation of land use to parcels of land according to central planning. The decision-making process, also in terms of land management, is determined centrally and, therefore, more uniform for relatively large areas. These factors play a role in simplifying the generation of scenarios for PLUTs and LUCs in the area studied.
Access to data and information was initially difficult. However, the process became highly participatory once the cooperation of the local and national agencies was obtained. Obstacles relating to logistics and scarce local resources also played a role in constraining the number of samples that could be obtained in the field. However, the field component of the project was considered a success in terms of the amounts of data gathered in the field by a quickly assembled multidisciplinary team in a rather shorttime.
The multispectral satellite image obtained of the area had considerable cloud cover and it was not possible to obtain any other image at the time. This hindered the definition of land cover classes by introducing confusion in the separation of the classes in the feature space, while performing a supervised classification of the image. Moreover, a compounding factor was the use of a relatively simplistic algorithm for classification (i.e. “box” classification), which was the only one available in Cuba at the time. As a result of these compounding factors, the initial land cover map developed from the supervised classification had about three misclassifications out of 17 classed defined (e.g. what was defined as “water body” is essentially part of the cloud cover; the class “cloud” was misclassified as “quarry”; etc.). The classes as determined by satellite image interpretation required extra fieldwork in order to be completed satisfactorily. This type of circumstances can be typical of the working environment in developing countries. Hence, procedures for dealing with missing data or software should be part of the contingency plans of any assessment and monitoring team.
After field validation, it was possible to establish as land cover classes: sugar cane; two classes of pastures (short and tall grasses in seven different mapping units); forests (two kinds: on sloping lands and on the banks of the river, in six mapping units); orchards (family orchards and small plantations, in three different mapping units); crops (three mapping units); maize (two types); urban land use and roads. The units mapped out correspond to what could be considered LUTs, for their differences lay in the management and level of inputs or infrastructural support, given the same type of activity (i.e. forest, pasture, orchard or crop). The area covered with sugar cane came out as being considerably smaller than the actual area covered by sugar cane. Hence, this was one of the major adjustments to the map after validation.
Biomass estimation
Biomass was calculated across the study area using measurements taken in 10 m × 10 m study quadrats. Similar to the Texcoco study area, the biomass calculated per quadrat is taken to be representative of the LUT polygon that it occurs within, averaging results between quadrats when more than one occurs within each polygon. Several methods of biomass calculation were applied to these data. The aboveground biomass of forested areas in the Rio Cauto study area were estimated using regression equation FAO-1 for “dry” tropical areas receiving more than 900 mm of annual precipitation. This equation gave the “best” fit to data in the least-squares sense.
The results indicate that forests contributed the most to the stock of aboveground biomass. This is in spite of the fact that much of the study area is agricultural land and forested land are only remnants of old forests on slopes of hills and on the sloping banks of the Cauto River. The level of contribution to the stocks of biomass and C is proportional to the level of health and degradation of these forests, some of which are enduring great ecological and human stresses (e.g. drought and selective cutting). Relatively healthy forests contributed about fourfold the total biomass of crops for the entire area. Stressed forests contributed twofold the biomass of crops in the area. The forested area constitutes about 15 percent of the total area studied.
Sugar cane was shown to contribute substantially to the stock of aboveground biomass. The biomass estimates for crops, particularly those for sugar cane (which covers about 65-70 percent of the total area and about 90 percent of the area cropped), were obtained from records of the National Institute for Sugar Cane Research (INICA) and verified by the institute’s local experts. In this situation, these records are considered more accurate than any estimates obtained from crop growth modelling. Variations in sugar-cane biomass estimates by mapping units are due to the variety and age of the semi-perennial crop. On the other hand, biomass in pastureland was estimated from the 1 × 1 m quadrats and then upscaled to a hectare.
The spatial distribution of total biomass (aboveground and belowground) estimates shows the remnant forests on the slopes of the riverbanks and on gentle slopes of hills, stocking 20-80 times more biomass (1 657 and 8 176 tonnes/ha respectively) than sugar cane and pasture (51.5 and 7.5 tonnes/ha respectively). The resulting carbon estimates in biomass are simply a proportion (about 0.55) of the biomass stock. This fact draws attention to the potential benefits that could accrue from implementing afforestation or reforestation in the area, or from implementing some form of agroforestry systems, which may include fruit trees and woody species of economic importance.
There were some mapping units in the land cover classification map for which no biomass estimates were possible owing to the lack of sampling quadrat sites. The quadrat sites that could be measured, within the constraints imposed by circumstances, were a sufficient but not very large number. Hence, for the purposes of this study, it was decided to leave such mapping units without estimates, rather than generate estimates from a relatively weak interpolating procedure.
Results of SOM modelling
The soil data from the soil mapping effort by the INICA in Cuba was useful for parameterizing both models of SOC dynamics: RothC-26.3 and CENTURY. However, the soil inventory is more than 30years old. Therefore, some of the analytical data, particularly chemical soil properties, may not reflect present conditions. The soil parameters were taken from this database to parameterize the RothC and CENTURY models. Running the RothC model “backwards” to equilibrium, until the current level of organic C in the soil (as per analysis) was achieved, proved to be a fruitful strategy for finding out the partitions of SOM in its different forms or pools, which are part of the current SOM levels in the Rio Cauto soils. These were taken as the starting point of the simulation “forward” into the future for both models, in lieu of the lack of detailed analytical data on the different fractions of SOM in the soils of thearea.
This initial simulation showed that for 2000, the forest soils (for those soils for which there were data) had the highest levels of SOC (up to 112.6 tonnes/ha). However, the stock in these soils is not significantly larger than that of the soils under sugar cane (102.3-108.93 tonnes/ha). This indicates that soils under sugar-cane cultivation for a period of up to seven years may accumulate amounts of organic residues comparable with soils under forests.
Scenarios that included SOM turnover simulations for 12 years (to 2012), with a combination of sugar-cane varieties grown on different soils and with two levels of additions of organic residues (0 and 2 tonnes/ha), were calculated. Such scenarios were considered conservative as far as the additions of organic residues were concerned. The scenarios were based on crops already part of the present cropping pattern in the area. No new crops in PLUTs were considered in these scenarios, as the suitability of the land for potential crops was not evaluated in Cuba. This was a decision resulting from considerations about the low likelihood of implementing any LUCs that may clash with the centrally-designed land-use plan.
Results of modelling with RothC-26.3
The modelling scenarios with RothC-26.3 included combinations of sugar cane, pastures and forests on dark gley plastic soils and on light grey carbonated soils, both with and without additions of 2 tonnes of C per hectare.
Of all the scenarios simulated for the dark gley soils, which are used mainly for sugar cane, only grasses achieved sequestration of C in SOM, both with 2 tonnes of C per hectare of organic residues as FYM and without any additions at all. When no FYM is added, grasses accrue 2.22 tonnes of C per hectare (from 55.8 to 58 tonnes of C per hectare in 2012) as SOM. When FYM is added, up to 9.75 tonnes of C per hectare as SOM are sequestered in the 12-year period (scenarios Nmp1 and Fyp1).
In contrast, for these same soils, sugar-cane looses total C from SOM to the atmosphere, without any additions of FYM, from -8.26 tonnes of C per hectare (a change from 75.7 to 68.1 tonnes of C per hectare; scenario Nmc1) to -13.43 tonnes of C per hectare (a change from 65.85 to 52.43 tonnes of C per hectare; scenario Nmc3). However, when 2 tonnes of residues (FYM) per hectare are added, the losses are reduced to only one-quarter or one-sixth of the losses without FYM: from -1.13 tonnes of C per hectare (a change from 64.5 to 63.45 tonnes of C per hectare; scenario Fyc2) to -2.89 tonnes of C per hectare (a change from 65.52 to 62.63 tonnes of C per hectare; scenario Fyc4). These figures indicate that the additions of organic residues (FYM) at the rate of 2 tonnes of C per hectare to sugar cane reduce the losses of C to the atmosphere substantially, and presumably they would sequester C if the additions of FYM were in a greater quantity. Thus, the additions of organic residues as FYM of 2 tonnes of C per hectare are too little to achieve carbon sequestration within the 12-year period. Greater amounts would enhance carbon sequestration and collateral ecological benefits, such as increasing soil biodiversity and preventing further land degradation. Therefore, the solution is based on the implementation of appropriate land management strategies based on SOM.
The simulations computed with this model demonstrate that for the entire landscape, sugar cane and, in some soils, grasses as LUTs, with no additions of organic inputs, are net emitters of C to the atmosphere. Conversely, a mix of sugar cane and grasses with organic inputs to the soil of more than 2 tonnes/ha/year for the 12-year period would sequester C in SOM.
Results of modelling with CENTURY
On the whole, the forested areas that are left intact accrue SOC. Regardless of the land use, additions of organic matter to the soil are crucial to SOC sequestration. A scenario of land-use conversion of sugar cane to pasture illustrates the same point. The resistant C in the “slow” pool or fraction is shown to be less affected by management of organic residues.
The scenarios developed through modelling with CENTURY explored the use of organic inputs and no inputs at all, and an LUC from present land use to grassland. Total SOC and the resistant or “slow” fraction of SOM (SOM3C) were modelled.
Total SOC is in moderate to high contents in grasslands (108-112 tonnes of C per hectare) and in not much smaller quantities in sugar cane soils (102-108 tonnes of C per hectare). In fact, these figures indicate that the total carbon contents of soils under both types of land use are comparable. Then, when 2 tonnes of C per hectare of organic inputs are added to the soil yearly, the soils under sugar cane (dark gleysols) can sequester up to 8-10 tonnes of C per hectare for the 12-year period (a change from 102-108 tonnes of C per hectare to 112.6-116 tonnes of C per hectare). This contrasts with the results modelled with RothC-26.3, where a negative change was observed with the same amounts of organic inputs and in the same soils. This may be because of the sensitivity of the CENTURY model and the result of accounting for the C and N interactions, which were not modelled with RothC-26.3. Grasses also sequester C (116-121 tonnes of C per hectare) from an initial range of 108-112 tonnes of C per hectare, representing an accrual of 8 tonnes of C per hectare on average, over a 12-year period, when 2 tonnes of C per hectare are added as inputs to the soil. Forests maintained 97.7-102.3 tonnes of C per hectare where left undisturbed.
The scenarios of LUC to grasslands (except forests) indicated that, without organic inputs to the soil, areas under sugar cane would lose an average of 20 tonnes of C per hectare for the 12-year period, and areas with other crops would lose as much as 40 tonnes of C per hectare as SOM in the same period. By contrast, undisturbed forests would sequester an average of 20 tonnes of C per hectare for the 12-year period. These results indicate that it may not be advantageous to convert land under sugar cane and other crops to grasslands. Furthermore, even where organic inputs (2 tonnes of C per hectare) are added to the conversion, former sugar-cane lands lose about 10 tonnes of C per hectare, which is half of the loses without organic inputs. Thus, the organic inputs only ameliorate the losses of the LUC.
The resistant fraction of SOM (SOM3C) experiences increases (sequestration) after the LUC and 12 years. Without organic inputs, 1 tonne of C per hectare of SOM3C is sequestered. About 2 tonnes of C per hectare of SOM3C are sequestered when organic inputs (FYM) are added to the soil.
The figures indicate that there is no point changing land use from sugar cane and other crops to pastureland. It may be more worthwhile to increase substantially the level of organic inputs to the soils under sugar cane and pastures as they actually are, and to increase or at least maintain the status of forest litter and forest cover in the area studied.
On the whole, the results obtained with CENTURY are comparable with those obtained with RothC-26.3. However, the CENTURY simulations allow for greater detail in the partition of the SOM fractions and may be considered more accurate. These considerations need to be balanced against the trade-offs of easy access and processing of the models and their data requirements.
Biodiversity assessment
As mentioned above, species richness provides an extremely useful measure of diversity where complemented by measures of evenness of abundance and dominance.
There are clear spatio-temporal trends in Rio Cauto. The greater number of species (S) can be found in the forests. Forests on hill-slopes hold between 13 and 34 species of plants, whereas forests on riverbanks hold an average of 12 plant species. Then, there is a hiatus in terms of plant diversity. Grasslands hold between 6 and 12 species and sugar-cane fields an average of 5 species. However, these apparently diverse forests are dominated by one or a few “rare” species, creating uneven plant populations. This can be appreciated through the values of the Shannon Diversity Index and the Simpson Diversity Index.
The Shannon Diversity Index (H) calculated for the recognized vegetation classes in Rio Cauto, Cuba, allowed for the calculation of a measure of the “evenness” (E = H/ln[S]) of species distribution. The values of “evenness” calculated for all vegetation classes are generally low. Values of 0.32 for forests on hill-slopes, 0.26 for forests on riverbanks, 0.24 in sugar cane fields and 0.14 for grasslands indicate that there are disproportionate distributions of individuals to the species found. This means that while some species counted may have several individuals, there may be only one individual or at most a few individuals for other species. A value of E = 1 would indicate that there are similar proportions of individuals (abundances) in all species. In the case of Rio Cauto, the values are small, indicating uneven abundance or dominance of a few species. This is not counterintuitive when one considers sugar-cane fields, but it is when considering forests. Therefore, forests are not very diverse.
The Simpson Diversity Index measures the “dominance” of species over others. In the case of the Cauto Watershed, the forests on hill-slopes show values of 0.85-0.90, which means that there is great dominance of one or two tree species in the forest. In the forests on riverbanks, the values range between 0.31 and 0.85, which means that these populations are less dominated by one or two species and can therefore be considered more diverse. The grasslands are relatively diverse and non-dominated by one or two species, with D values of 0.19-0.31.
On the whole, it can be ascertained that the landscapes in Rio Cauto represent low diversity of species and great dominance of a few species in terms of number of individuals and their spatial coverage. It is reasonable to assume that the diversity of species would be increased if LUCs were introduced, particularly along the lines of agroforestry systems and appropriate management of the land and SOM.
Land degradation assessment
The lack of data to compile the indicators of land degradation for the Rio Cauto Watershed impeded the calculation of indices of chemical, physical and biological land degradation. However, informal observations reported through the field visits are summarized below.
In terms of physical land degradation, moderate to severe soil erosion by water and wind was observed. The presence of gullies and rills was quite evident, particularly in denuded soils with degraded grasses, and in the hill-slopes and, particularly on the banks of the Cauto River. Sheet erosion by water was also quite evident in areas not covered by sugar cane. On the other hand, land compaction is expected and was also quite evident, particularly in sugar-cane fields in rotation, where the denuded soil could be observed. This is an expected phenomenon owing to the extreme reliance on heavy machinery and the size of the area under sugar-cane cropping.
In terms of chemical degradation, no evidence was observed of chemical contaminants in the area (e.g. pesticides and herbicides). However, it is expected that at least moderate levels of this type of contamination exist given the current paradigm of land and crop management. In contrast, salinity is a relatively important problem, particularly in areas planted with sugar cane and other areas near the banks of the Cauto River. Land reclamation efforts are being applied to the desalinization of some of these soils (dark plastic gleysols). Nevertheless, the salinity and sodication problems in specific low-lying areas range from severe to very severe. No spatial data could be obtained to determine the spatial extent of the problem.
In terms of biological degradation, the SOM imbalance throughout the entire watershed is clear. It is to be expected given the evapotranspiration deficits and the conditions for easy attack and oxidation of the organic matter with consequent losses to the atmosphere. As has been demonstrated through modelling, organic matter is key to managing these soils.
