Warnings about the likely undesirable consequences of climate change have come largely from environmental scientists. However, the language used by scientists to describe their observations and predictions is often hard for non-scientists to understand. A basic knowledge of how science works and how scientists come to their conclusions may help in understanding what they are saying.

The process of science generally starts with observations of the world. Scientists seek to observe the world in a dispassionate way by trying to avoid making assumptions about what they will see or what their measurements will or will not show. Painstaking observations and accurate recording are essential tools in scientific work – for example methodical astronomical observations in the 16th century led to an understanding of the orbits of planets around the sun. 
Science uses data to develop testable questions – known as hypotheses – about what is being observed. To be scientifically useful a hypothesis must be falsifiable which means there must be a logical possibility that it can be shown to be false by observation or experiment.  Beliefs are not falsifiable and cannot form the basis of a scientific hypothesis.  Sometimes hypotheses will come from a deep analysis of data; at other times they are in response to questions people want answered – for example ‘does this new drug work’? 
The scientific method involves checking if hypotheses hold up by making more measurements to see if they are consistent with the hypothesis – for example testing if mice are more likely get better when given a new drug. Scientists design experiments that aim to test hypotheses more widely. In climate science there is rarely the opportunity to carry out carefully controlled experiments in the laboratory, so measurements of the actual climate and past events, such as those that occurred in the repeated sequence of warm periods between the last ice ages, must be studied instead.
Scientific hypotheses can never be ‘proved’ because there always remains the possibility that contradictory data will be found later. However, as hypotheses remain consistent with more and more observations, they are often built up into theories – ideas that are felt to be strong enough to start making predictions. The process does not always proceed in this order – Newton and Einstein came up with many of their ideas based on abstract reasoning, but subsequent testing has shown them to be extremely valuable (and testable) theories in terms of predicting how the world works.
The world is a stochastic place  –many events are determined partly by chance which  means there is often a lot of ‘noise’ or randomness around the data we collect. For example, patients react differently to drugs, some get better and some do not. As a consequence modern science applies statistical techniques to assess data and results. These techniques become more important when it is impossible or impractical to conduct certain experiments, as is true to a large extent in environmental science. 
Statistics is a branch of mathematics. Statistical methods are used to test hypotheses and to determine how certain or uncertain conclusions are. Thus scientists often talk about things like ‘confidence levels’ –that is how confident we can be that a given hypothesis is supported by the data; they do not use statistics to claim that something has been proved (or disproved) absolutely. This might make science sound as if it can never answer any questions definitively. However, as evidence builds up, confidence in theories and predictions increases. 
People can find this apparent lack of certainty disconcerting, but in reality it is no different from the judgements people make in their own lives - for example: in judging the probability that a horse with good form will win a race; or the probability that something bad will happen when they take a risk.
Anyone can undertake scientific investigations, and most important scientific theories and discoveries are freely available in scientific journals for anyone to study. However, to publish a paper in a reputable scientific journal, scientists are expected to follow rigorous methods and be subject to peer review. Results of experiments should be reproducible by others. Many scientific frauds or misinterpretations were finally exposed when others failed to duplicate the claimed results. Journals use other reputable scientists, who may be working on competing hypotheses, to study and comment on the validity of draft papers and articles before they are published.  In this peer review process, papers are often reviewed by two or three scientists working in different institutions.  These checks are not perfect but are thought to weed out most poor-quality work and eliminate errors.
These days, scientists frequently use computer modelling which allows them to make more precise predictions based on measurements and theories. Things that are modelled range from the behaviour of individual atoms up to modelling the Earth’s climate. Scientists make efforts to check the results of models by, for example, trying to duplicate the results with an independently constructed model. Of course computer modelling is also used and trusted in many other applications – for example in modern engineering design and aircraft flight systems.