Abstract. In this study, land surface related parameter distributions of a conceptual semi-distributed hydrological model are estimated by employing time series of satellite-based evaporation estimates during the dry season as explanatory information. A key application for this approach is to identify part of the parameter distribution space in ungauged river basins without the need for ground data. The information, contained in the evaporation estimates implicitly imposes compliance of the model with the largest water balance term, evaporation, and a spatially and temporally realistic depletion of soil moisture within the dry season. Furthermore, the model results can provide a better understanding of the information density of remotely sensed evaporation. The approach has been applied to the ungauged Luangwa river basin (150 000 (km)2) in Zambia. Model units were delineated on the basis of similar land cover. For each model unit, model parameters for which evaporation is sensitive, have been conditioned on the evaporation estimates by means of Monte-Carlo sampling. The results show that behavioural parameter sets for model units with similar land cover, are indeed clustered. The clustering reveals hydrologically meaningful signatures in the parameter response surface: wetland-dominated areas (also called dambos) show optimal parameter ranges that reflect a relatively small unsaturated zone (due to the shallow rooting depth of the vegetation) and moisture stressed vegetation. The forested areas and evergreen highlands show parameter ranges that indicate a much deeper root zone and drought resistance. Unrealistic parameter ranges, found for instance in the high optimal field capacity values in the highlands may indicate model structural deficiencies. We believe that in these areas, groundwater uptake into the root zone and lateral movement of groundwater should be included in the model structure. Furthermore, a less distinct parameter clustering was found for forested model units. We hypothesize that this is due to the presence of 2 dominant forest types that differ substantially in their moisture regime. Therefore, this could indicate that the spatial discretization used in this study is oversimplified. This constraining step with remotely sensed data is useful for Bayesian updating in ungauged catchments. To this end trapezoidal shaped fuzzy membership functions were constructed that can be used to constrain parameter realizations in a second calibration step if more data becomes available. Especially in semi-arid areas such as the Luangwa basin, traditional rainfall-runoff calibration should be preceded by this step because evaporation represents a much larger term in the water balance than discharge and because it imposes spatial variability in the water balance. It justifies that land surface related parameters are distributed. Furthermore, the analysis reveals where hydrological processes may be ill-defined in the model structure and how accurate our spatial discretization is.
Anopheles fauna of coastal Cayenne, French Guiana: modelling and mapping of species presence using remotely sensed land cover data
