induced?
The following sections detail the proposed research approach in the context of attributing the recent cases of extreme flooding in the Sahel region, notably the right bank of the Niger River near Niamey, Niger.
Although other causal factors exist, it is assumed based on previous re- search findings that climate (precipitation) and land surface properties are two factors that may have the biggest influence on Hortonian runoff in the Sahel. Modeling and attribution efforts will therefore first concentrate on identifying the relative contribution of precipitation and human-driven land surface changes, separating out the precipitation-driven contribution to land surface changes.
Figure 5.1 shows a proposed modeling framework for attribution of hy- drological extremes in its simplest form. Historical changes in precipitation are identified in AMMA-CATCH data and implemented into the rainfall sim- ulator Stochastorm. Scenarios are generated from Stochastorm. In parallel, land cover changes are evaluated using land surface maps. In one scenario, the trends in land cover changes are implemented into hydrological model parameters, but precipitation is maintained stationary (1980s values). In another scenario, the non-stationary precipitation scenarios are used as in- puts to drive the model but land surface properties stay stationary (1980s values). Another scenario could be created where both observed changes are
Figure 5.1: Proposed framework for attribution of hydrological extremes
implemented in the model. The methods described in Section 5.3.2 are used to attribute the observed changes in the GEV distribution of hydrological extremes shown in Chapter 2 to a given scenario.
With climate change analysis, the causal factors are relatively indepen- dent. Natural causes of atmospheric composition changes such as volcanic eruptions and solar irradiance fluctuations are arguably quite independent from human-produced greenhouse gas emissions. This is not the case when determining whether the cause of changes in hydrological extremes is pre- cipitation or land surface properties. Precipitation has a direct impact on vegetation, which has a direct impact on surface runoff. An increase (de- crease) in precipitation could lead to an increase (decrease) in vegetation. Land surface changes can likewise impact local convection and, as a result, local convective rainfall. The interactions are complex, intertwined, and of- ten challenging to study due to lack of data and difficulty of relevant data
collection.
Determining whether changes in extreme streamflow are due to precipi- tation or land use changes is scientifically interesting. However, it bypasses the original question asked in climate sciences of whether changes are due to natural or anthropogenic factors. Land cover changes can be due to both natural and human causes. Human impacts include agricultural practices, grazing patterns, and urbanization. Separating out the relative contributions of natural processes and human impact from land surface changes would be another required step in the attribution process.
If one zooms out enough, precipitation changes (notably intensification) can also potentially be due to anthropogenic climate change and global warm- ing, although that would need to be verified for the specific study region. Many climate variables are known to undergo long-term oscillations, which necessitates separating such trends from those caused by other forcings. In addition, interactions between precipitation at a specified location and larger- scale climate systems are complex, with a high degree of variability. Although global and regional climate models can be useful, it is more difficult to use them to attribute changes at a small spatial scales.
To go a step further, humans themselves respond to changes in the natu- ral environment. Human impact on local land surfaces can in part a response to climate change. Herders in the Sahel changed the area and extent over which they let their animals graze as a response to drought conditions. This had a direct impact on local vegetation cover. Farmers change their agricul- tural practices in response to precipitation increases or decreases, which in turn changes the properties of the soil. Thousands of people migrate due to drought. One could attribute these “anthropic” changes to climate changes. We see here that there are multiple potential issues to be resolved. Before proposing a modeling framework, it is important to clearly identify what the core research question is to be answered, and why. One must also determine the “box” around the hydrological system that determines at which scale the changes are attributed and which factors are considered external forcings or internal system variability.
A proposition is to focus on identifying the impacts of local anthropogenic influence on hydrological processes (for now, disregarding whether the human changes were originally due to climate). The impact of human societies on the local environment is a pertinent issue because it has relevance to how human behavior can influence floods. It can guide whether local populations and governments should change adaptation measures and how, based on which
human activities have impact on flooding.
In climate attribution studies, feedback between variables is integrated into the modeling system. For example, the atmospheric temperature re- sponse to changes in albedo from ice melting is modeled. For the hydro- logical attribution setup, one could envision modeling vegetation response to precipitation, local precipitation response to land surface properties, etc. Modeling feedback effects requires extensive interdisciplinary knowledge of how processes impact each other. Future research efforts would involve col- lecting existing knowledge about factors that affect the hydroclimatic system in West Africa, filling gaps in the understanding of the interactions between processes, representing the interactions mathematically, and incorporating the processes into an increasingly complex modeling system. Some studies have already been conducted to determine the impact one variable has on
another, for example B´egu´e et al. (2011)’s analysis on vegetation responses
to precipitation and land use changes in the Sahelian and Sudanian regions. One could also explore theoretical methods for multiple linear regression that take into account non-independence between explanatory variables.
5.3.5
Land use scenarios
Going beyond attributing the causes of documented floods that have already occurred, one can envision making land use projections to complement the precipitation projections provided by RCMs. Making land use projection scenarios would required extended knowledge of how humans may impact their natural environment.
To better understand the human component of land surface changes, one could start by examining how people currently modify soils. There are projects underway for creating half-moons and banquettes that alter local runoff. Agricultural practices change in response to demographic pressure, changes in cultural practices, and to sudden events. There are also regreening efforts underway, and regional soil rehabilitation projects. What influence do these activities have on hydrological processes? The development of land use scenarios would require interdisciplinary research on what changes in human behavior produce what changes in land surface properties.
As changes in societies are even more challenging to predict than changes in local precipitation, one could develop multiple land use/land surface sce- narios based on population and demographic projections and test them in combination with precipitation projections. The factors can then be inte-
grated into the hydrological modeling system to determine impacts. Attribu- tion methods can be applied on future projections. Appropriate adaptation measures could then be identified based on results.