Brief History on the HAND model Development

A contextualizing narrative

by Antonio Donato Nobre

From 1999 to 2002, our LBA Project’s research team installed the first instrumented hydrological catchment in the Amazon, at the Igarapé Asu, in area of undisturbed forest near Manaus (described in Waterloo et al 2006 and Cuartas et al 2007). Some years after the installation, we faced the challenge of significantly representing the soil properties and inferring soil water dynamics to a large, hard to reach area, using only the point data generated by the monitoring instruments. Due to the impenetrability of the forest canopy, the use of passive two-dimensional surface imaging techniques to estimate the possible distribution of soil properties was limited. However, at that time, orbital radar sensing missions generated surface elevation models (3D digital topography, JERS-1, SRTM). We began to search for an innovative soil mapping approach that could combine hydrological inference with the new topographic data.

In describing the Amazon landscape, many researchers routinely use qualitative classifications of terrain types, according to the water saturation regime and local relief position (eg, Sioli 1984, Chauvel et al., 1987). For the Amazonian Hileia (wide expanse of the basin lying below 600 m above sea level), it is quite common to use geomorphological / physiognomic (or physiographic) terms, as – lowland (baixio), -slope (encosta) and –plateau (platô). Such a classification can be taken for its practical value, based on local perceived terrain properties, as a semantic description of landscape compartments. Yet, this colloquial form of referring to classes of terrain carry little or no quantitative dimension. Its predictive capacity was thwarted by the difficulty of making any significant and reproducible physical association with the topography (defined in relation to sea level). Consider a South American example: a swamp can occur at any altitude, from near the mouth of the Amazon River – almost at sea level – or all the way up to the shores of Lake Titicaca in the Andes at 4000 m A.S.L. It is clear, then, that height above sea level mingles and confuses local hillslope gradients with continental scale landscape gradients.

We needed to isolate local environments and place them in a comparable topographic framework. This was my insight that motivated us to seek the development of a new normalized terrain model. The idea seemed very simple, yet, to our knowledge, no one had seen it before. A few years later, under the GEOMA modeling project, I led an effort to build a computational tool that could normalize DEMs just as I had imagined, so that local gradients could become physically comparable over larger areas. The conceptual development of the landscape classification in hydrologically relevant terrain classes (described in Nobre et al., 2011) preceded the HAND algorithm and, in fact, guided its development. We had in the group some disbelief that such topographical normalization had not yet been made, so we revised the (vast) literature on topography-hydrology, only to confirm the inexistence of something similar to the HAND model.

Confident in the originality of the idea, around 2005 we made the first promising tests for the Asu catchment. With the prototype of the DEM tool, we focused on the field verified classes. To test the quantitative capacity of the new normalized terrain classification, we performed an analysis of the soil water distribution data. Basing the class allocation onto our depth to water table data, we nominated the matching field classes as waterlogged (saturated to the surface), ecotone (shallow water table), slope and plateau (deeper water table). The HAND landscape classification was initially intended to solve an interpolation for a small river catchment (13 km²). As we applied it to surrounding ungauged catchments, near and far from Asu, we became convinced that we had discovered a robust physical property of the landscape: the elusive topographical coherence of soil-water, correlated to a new dimension of terrain, the HAND normalized topology.

In Nobre et al (2011) we introduced the HAND terrain model, which captures with physical substance the topographic relation between soil and water, demonstrating it with ample validation. The land classification based on the HAND model was applied and verified in the Asu catchment, with robust results also for two surrounding areas, the Cuieiras river (500 km²) and the lower part of the Rio Negro basin (18,000 km²). After this initial development in central Amazonia, we applied and verified the HAND classification for even larger heterogeneous areas, dispersed throughout Brazil (more than 300 thousand km², Nobre et al. 2011b). The HAND classification was also validated independently for different geomorphologies, with equally robust results. With Nobre et al. (2011) we presented the physical foundations and a new quantitative classification of the landscape. Such classification has become a very informative and useful reference, whose classes have recognized global relevance. Throughout the 10 years of initial development (2001 to 2011) we managed to demonstrate that the old semantic classification, based on intuitive perceptions of the groundwater regime and relative position in the relief, could finally acquire quantitative and predictive capacity through the HAND model. The normalization of relief via the HAND topology, and the classification of the terrain derived from it, allowed an entirely new and revolutionary hydrological view of the landscape.

References

Chauvel, A., Lucas, Y., and Boulet, R., 1987 On the genesis of the soil mantle of the region of Manaus, Central Amazonia, Brazil. Experientia 43, 234-240.

Cuartas, L.A., Tomasella, J., Nobre, A.D., Hodnett, M.G., Waterloo, M.J., Múnera, J.C., 2007. Interception water-partitioning dynamics for a pristine rainforest in Central Amazonia: marked differences between normal and dry years. Agricultural and Forest Meteorology 145 (1–2), 69–83.

Nobre, A.D., Cuartas, L.A., Hodnett, et al 2011. Height Above the Nearest Drainage – a hydrologically relevant new terrain model. Journal of Hydrology 404, 13–29.

Nobre, A. D.; Silveira, A … Cuartas, L.A., 2011b. Aspectos físicos e geográficos das áreas ripárias no Brasil: análise preliminar da legislação. Ciência para o Código Florestal. São José dos Campos: Centro de Ciência para o Sistema Terrestre – INPE, 110 p. Relatório Cientifico.

Sioli, H., 1984. The Amazon: limnology and landscape ecology of a mighty tropical river and its basin. W. Junk, 763 p.

Waterloo, M.J., Oliveira, S.M., Drucker, D.P., Nobre, A.D., Cuartas, L.A., Hodnett, M.G., Wilma, I.L., Jans, W.P., Tomasella, J., Araujo, A.C., Pimentel, T.P., Munera Estrada, J.C., 2006. Export of organic carbon in run-off from an Amazonian rainforest black water catchment. Hydrological Processes 20, 2581–2597.