BNPP/ASB Functional Value of Biodiversity Project – Phase II 

i Assemble more detailed information on biodiversity-rich tropical habitats (IFPRI lead initiative)

iv Undertake synoptic modeling of hydrological impacts of land use change

1Bi.  Characterize areas vulnerable to changes in hydrological function and identify hydrological "hotspots"

The WBM hydrological assessment will evaluate differences in the annual distributions of total water yields, dry weather flows and peak flows associated with each scenario. For the local hazard assessment these differences only need be calculated within the tropical forest biomes. For the far field assessments the differences will be calculated at key river control sections throughout each river basin (but not including the tropical forest biome since that's covered by the local hazard assessment?). By default (and subject to confirmation by UNH) such control sections will be defined as occurring at major confluences, and also at the location of larger human settlement along the river (or known to be located in a river floodplain). The potential hazards may be summarized as: 1. Changes in average annual water yield: reservoir inflows for water supply, hydro-electric power 

2. Changes in dry weather flows: reduced effluent dilution capacity, reduced navigational and ecosystem services, insufficient flushing of saline water in rivers or aquifers 

3. Changes in peak flows: flooding hazard, increased soil erosion, river bed/bank scouring, landslides.

Since it is challenging to summarize the differences in the frequency distribution of each of these hydrological metrics a more practical and policy relevant format may be simply to tabulate percentage differences for a number of specific likelihoods of occurrence. For example, percentage change in the average, 1 in 5, 1 in 10, 1 in 20, 1 in 50 and 1 in 100 year extreme dry and peak flows could be reported for each river control section. The location of larger human settlements and population densities and totals will be taken from Activity 1.A.ii above. Population estimates will be overlaid with estimates of change by hydrologic function and return period and a typology of potential type and level of hazard developed.

There is scope for this stage of the methodology to be extended and refined as the WBM simulations are being run and the preliminary outputs obtained.

General notes, 

Comments 

Amendment to Activity 1 Implementation Protocols based on discussions during BNPP team meeting in Prague, Czech Republic, 11-12 October 2003:

Decision, based on characteristics of the UNH Water Balance Model (WBM):  minimum basin size for application of WBM is 30,000 km2 (12 pixels).  108 basins in the study domain are that size or larger.  Note that, by this construction, ‘coastal’ basins (i.e., small basins near the coast) fall out of the WBM analysis.

Suggested overlays regarding the basins in the WBM domain: human population in each.   

Suggested hydrological ‘leverage’ indicator for total yield:

Percent of annual average of precipitation for the entire basin / pixel 

Suggested alternative graphical representations of results for total yield at the basin level:

x-axis: percent basin area converted from forest – or – percent of basin area forested

y-axis: absolute and relative changes in total yield; absolute change in total yield divided by total water demand for the basin. 

Suggested representations for water demand at the basin level are:

Change in total yield / human population of the basin

Change in total yield / (total yield – total demand), with demand referring to use or offtake by humans for domestic needs, agriculture, industry, etc  

Suggested alternative spatial representations for changes in total yield at the basin level, delineated by biome: absolute and relative change in yield for each pixel

Suggested hydrological ‘leverage’ indicator for flooding risk: an upland hazard index was proposed. 

Development of a potentially useful index for identifying vulnerability to upstream hazards.

E.M.Douglas, 10/18/03

Description of the Upstream Hazard Index:  This index was developed based on my assumption that narrow and steep basins would be associated with a potentially higher hazard vulnerability of downstream populations living on floodplains, relative to wider, shallower basins. Presumably as the basin width narrows and the slope increases, the possibility of upstream hazards increases (i.e., due to minimal attenuation of peak flow).   I wanted an index that would identify basins or sub-basins that were both narrow and steep.  For any grid cell, the UHI was computed as

UHI = cumulative (upstream) area / cumulative (upstream) distance * mean slope for that grid cell.

In the figure above I have plotted the UHI (see legend) and also the occurrence of population living on floodplains (pink/purple color) for southern and Southeast Asia.  The Salween and the upstream portions of the Mekong and Yangtze have a high (>5) UHI.  In the case of the Mekong and the Yangtze, the occurrence of populations on floodplains can be seen directly downstream of where the UHI changes from >5 to <1, indicating that these populations may be vulnerable to upstream hazards (i.e. extreme floods or landslides). However, the Chao Phraya has a very low UHI, which may be more indicative of the scale of this analysis (30-min) than the usefulness of the index.  The Chao Phraya contains only 10 grid cells at this resolution.  

Cumulative distance / cumulative area is essentially an “effective upstream basin width”.  Because I wanted the index to increase with narrower basins, I used the reciprocal of the effective width.   I chose to use the mean slope rather than a moving average slope because I wanted a dramatic change in slope to cause an identifiable change in the index.  Because of this, the index is dominated by, but not completely prescribed by, the grid cell slope.  The mean grid cell slope (in m/km) was computed as the mean slope derived from the GTOPO30 1-km DEM within each 30-min grid cell (see Meybeck et al., 2001). Cumulative distance (km) and area (km²) were computed from STN-30 (Fekete et al., 2001).    The index units are m/km² which are not very physically meaningful.  I tried to create a unitless version of this index or one at least with physically meaningful units (i.e., m/km), but other versions had extremely small or large values ranging over many orders of magnitude and none seemed to be as good at identifying the narrow steep basins or sub-basins as this one.

I have not yet tested the index with respect scalability or a threshold value for indicating hazard/no hazard, both which will need to be done to really make this index useful.  However, we are considering using the UHI as a means of identifying forested areas that should be protected in Activity 1 landcover scenario 4.  We would like the thoughts of the Activity 2 team with respect to the correctness and usefulness of this index.

References

·        Fekete, B. M., C. J. Vorosmarty, and R. B. Lammers. 2001. Scaling gridded river networks for macroscale hydrology: development, analysis and control of error, Water Resources Research, 37 (7): 1955-1967.

·        Meybeck, M., P. Green and C. Vorosmarty. 2001. A new typology for mountains and other relief classes: an application to global continental water resources and population, Mountain Research and Development, 21 (1): 34-45.

Comments and questions on UHI from Ken Chomitz [10/19/2003]:

1. Remind me on the definition of upstream length -- it is the longest upstream path?  Also, I think that the correct formula is length/area as in the figure, not area/length in the text.

2.  Why slope at just the point of calculation?  At 50 km resolution that's a pretty crude slope measure anyway. And maybe the hazards are greatest where the local slope is lowest? It's probably worth being a little more precise about the intuition here.  Are we talking specifically about flooding?  You seem to want to identify flooding risks as "the place where the index changes from >5 to <1". But that complicated phrase suggests that the index isn't doing its job -- somehow you want it to peak where it hits the floodplains.

3.  Somehow rainfall should enter into the index, it seems to me.  Why not multiply by total upstream rainfall.  Holding basin width constant, don't you expect impacts to scale up with rainfall?

4.  Why not somehow incorporate a variant of the Chomitz-Nelson forest-nonforest interface?  Where the forest is all gone, there's no leverage.  Where the forest is intact, there's probably no threat.  So a quick proxy for potential land use change is:

  [sum over upstream cells of (proportion of forest cover)(proportion of nonforest cover)]/(number of upstream cells) You could multiply by 0.25 to get something that scales from 0 to 1. Probably works best, the smaller the cells.

5.  How to test or calibrate?

Design and update: Sandra Velarde

                                                                                                                                               back to top

Global Partnership Amazon Congo Basin Southeast Asia
Climate Change Biodiversity Sustainability Poverty, Policy and Deforestation
Partners Publications Policy Briefs Gallery Funding Priorities Get Involved
e-mail: asb@cgiar.org   Home Page

Last updated: 28 November, 2003     ©2003 ASB. All rights reserved.