Crunching Numbers, Hidden States
This year’s Climate and Society class is out in the field (or lab or office) completing a summer internship or thesis. They’ll be documenting their experiences one blog post at a time. Read on to see what they’re up to.
By Shammunul Islam, Climate and Society 2014
Climate models and analyses of the available data allow us to have a more comprehensive understanding of water and the Earth system. The better we understand this delicate relationship, the better we can make decisions and policies that protect both. The importance of this has been duly acknowledged in the Working Group I contribution to the Intergovernmental Panel on Climate Change’s recent report:
There should be … modeling capabilities, attribution of the changes to causes, predictions from daily to decadal time scales, projections of the longer term future, and an assessment of all of these for use by decision makers.
This requires modeling precipitation, temperature, water vapor and other facets of the climate system and factoring observations into model projections and downscaling them to the regional scale. Global Circulation Models (GCM) simulate the evolution of atmosphere and water. But, as these models split the world into grid boxes hundreds of kilometers wide, they are not well-suited for regional scale modeling of climate processes such as rainfall. Hence we need to modify large-scale GCMs to local level climate variables, a process called downscaling.
I am doing my summer internship at the International Research Institute for Climate and Society investigating how large-scale climate variability like El Niño can drive the rainfall variability on a daily scale over the Bhakra basin, home to one of the largest reservoirs in northwest India. In this work, multiple station precipitation and temperature data has been used and modeled with the help of a unique modeling method. The model assumes that all stations in the Bakhra region have similar states of rainfall which we can’t observe (and hence, hidden states). It allows the probability of rainfall and El Niño — or its opposite phase, La Niña — to influence both the hidden states and the distribution of rainfall in all the stations considered in a Bayesian framework.
As we can see from the panel above, the top figure shows the evolution of daily rainfall classified into six hidden states, ordered from driest (S1) to wettest (S6), drawn with Bayesian sampling. The bottom figure shows the seasonal changes for these six states. From this we can infer that from early June, there normally is a transition to state 5 or 6 corresponding to heavy rainfall, which aligns nicely with the onset of India’s monsoon period, one of the most important weather patterns in the globe. We can also observe that at the beginning and end of a year, this region remains mainly dry or somewhat dry. These figures conform to our expectation of what we could have observed.
This number crunching using hidden states allows us to downscale GCM climate predictions and projections for better water management. It opens up the avenue for replication in other regions and helps in addressing growing water crisis across southern and eastern Asia. The better we crunch the numbers, the better we stand the chance of making hidden mechanisms unhidden.