Summary
1In recent years the use of species distribution models by ecologists and conservation managers has increased considerably, along with an awareness of the need to provide accuracy assessment for predictions of such models. The kappa statistic is the most widely used measure for the performance of models generating presence–absence predictions, but several studies have criticized it for being inherently dependent on prevalence, and argued that this dependency introduces statistical artefacts to estimates of predictive accuracy. This criticism has been supported recently by computer simulations showing that kappa responds to the prevalence of the modelled species in a unimodal fashion.
2In this paper we provide a theoretical explanation for the observed dependence of kappa on prevalence, and introduce into ecology an alternative measure of accuracy, the true skill statistic (TSS), which corrects for this dependence while still keeping all the advantages of kappa. We also compare the responses of kappa and TSS to prevalence using empirical data, by modelling distribution patterns of 128 species of woody plant in Israel.
3The theoretical analysis shows that kappa responds in a unimodal fashion to variation in prevalence and that the level of prevalence that maximizes kappa depends on the ratio between sensitivity (the proportion of correctly predicted presences) and specificity (the proportion of correctly predicted absences). In contrast, TSS is independent of prevalence.
4When the two measures of accuracy were compared using empirical data, kappa showed a unimodal response to prevalence, in agreement with the theoretical analysis. TSS showed a decreasing linear response to prevalence, a result we interpret as reflecting true ecological phenomena rather than a statistical artefact. This interpretation is supported by the fact that a similar pattern was found for the area under the ROC curve, a measure known to be independent of prevalence.
5Synthesis and applications. Our results provide theoretical and empirical evidence that kappa, one of the most widely used measures of model performance in ecology, has serious limitations that make it unsuitable for such applications. The alternative we suggest, TSS, compensates for the shortcomings of kappa while keeping all of its advantages. We therefore recommend the TSS as a simple and intuitive measure for the performance of species distribution models when predictions are expressed as presence–absence maps.

https://besjournals.onlinelibrary.wiley.com/doi/pdf/10.1111/j.1365-2664.2006.01214.x

Assessing the accuracy of species distribution models: prevalence, kappa and the true skill statistic (TSS)

http://abe.ufl.edu/Faculty/jjones/ABE_5646/Xtra%20files/The%20DSSAT%20Cropping%20System%20Model.pdf

The DSSAT cropping system model

There is growing recognition in the human dimensions research community that climate change impact studies must take into account the effects of other ongoing global changes. Yet there has been no systematic methodology to study climate change vulnerability in the context of multiple stressors. Using the example of Indian agriculture, this paper presents a methodology for investigating regional vulnerability to climate change in combination with other global stressors. This method, which relies on both vulnerability mapping and local-level case studies, may be used to assess differential vulnerability for any particular sector within a nation or region, and it can serve as a basis for targeting policy interventions.

Mapping vulnerability to multiple stressors: climate change and globalization in India

Why different interpretations of vulnerability matter in climate change discourses

Many definitions of food security exist, and these have been the subject of much debate. As early as 1992, Maxwell and Smith (1992) reviewed more than 180 items discussing concepts and definitions, and more definitions have been formulated since (DEFRA, 2006). Whereas many earlier definitions centered on food production, more recent definitions highlight access to food, in keeping with the 1996 World Food Summit definition (FAO, 1996) that food security is met when “all people, at all times, have physical and economic access to sufficient, safe, and nutritious food to meet their dietary needs and food preferences for an active and healthy life.” Worldwide attention on food access was given impetus by the food “price spike” in 2007–2008, triggered by a complex set of long- and short-term factors (FAO, 2009b; von Braun and Torero, 2009). FAO concluded, “provisional estimates show that, in 2007, 75 million more people were added to the total number of undernourished relative to 2003–05” (FAO, 2008); this is arguably a low-end estimate (Headey and Fan, 2010). More than enough food is currently produced per capita to feed the global population, yet about 870 million people remained hungry in the period from 2010 to 2012 (FAO et al., 2012). The questions for this chapter are how far climate and its change affect current food production systems and food security and the extent to which they will do so in the future (Figure 7-1).

Food security and food production systems

The CERES-Rice v3. crop simulation model, calibrated and validated for its suitability to simulate rice production in the tropical humid climate Kerala State of India, is used for analysing the effect of climate change on rice productivity in the state. The plausible climate change scenario for the Indian subcontinent as expected by the middle of the next century, taking into account the projected emissions of greenhouse gases and sulphate aerosols, in a coupled atmosphere-ocean model experiment performed at Deutsches Klimarechenzentrum, Germany, is adopted for the study. The adopted scenario represented an increase in monsoon seasonal mean surface temperature of the order of about 1.5°C, and an increase in rainfall of the order of 2 mm per day, over the state of Kerala in the decade 2040–2049 with respect to the 1980s. The IPCC Business-as-usual scenario projection of plant usable concentration of CO2 about 460 PPM by the middle of the next century are also used in the crop model simulation. On an average over the state with the climate change scenario studied, the rice maturity period is projected to shorten by 8% and yield increase by 12%. When temperature elevations only are taken into consideration, the crop simulations show a decrease of 8% in crop maturity period and 6% in yield. This shows that the increase in yield due to fertilisation effect of elevated CO2 and increased rainfall over the state as projected in the climate change scenario nearly makes up for the negative impact on rice yield due to temperature rise. The sensitivity experiments of the rice model to CO2 concentration changes indicated that over the state, an increase in CO2 concentration leads to yield increase due to its fertilisation effect and also enhance the water use efficiency of the paddy. The temperature sensitivity experiments have shown that for a positive change in temperature up to 5°C, there is a continuous decline in the yield. For every one degree increment the decline in yield is about 6%. Also, in another experiment it is observed that the physiological effect of ambient CO2 at 425 ppm concentration compensated for the yield losses due to increase in temperature up to 2°C. Rainfall sensitivity experiments have shown that increase in rice yield due to increase in rainfall above the observed values is near exponential. But decrease in rainfall results in yield loss at a constant rate of about 8% per 2 mm/day, up to about 16 mm/day.

Effects of Climate Change on Rice Production in the Tropical Humid Climate of Kerala, India

Vulnerability of rice and wheat yields in NW India to future changes in climate

The CERES-Rice version 3.0 crop growth simulation model was calibrated and evaluated for the agroclimatic conditions of the state of Kerala in India. Genetic coefficients were developed for the rice crop variety Jaya and used for the model evaluation studies. In four experiments using different transplanting dates during the virippu season (June to September) under rainfed conditions (i.e. no irrigation), the flowering date was predicted within an error of four days and date of crop maturity within an error of two days. The model was found to predict the phenological events of the crop fairly well. The grain yield predicted by the model was within an error of 3 for all the transplanting dates, but the straw yield prediction was within an error of 27. The high accuracy of the grain yield prediction showed the ability of the model to simulate the growth of the crop in the agroclimatic conditions of Kerala. It can be concluded from this study that the model can be used for making various strategic and tactical decisions related to agricultural planning in the state.

https://rmets.onlinelibrary.wiley.com/doi/pdf/10.1017/S1350482798000954

Evaluation of the CERES-Rice version 3.0 model for the climate conditions of the state of Kerala, India

Weekly cumulative rainfall forecasts were made for the meteorologically homogeneous areas of the Indian subcontinent, divided into meteorological subdivisions, by performing 7-day integrations of the operational Indian T80 Global Spectral Model every Wednesday during the six southwest monsoon seasons of 1994–99. Objective evaluations of the bias and accuracy of these forecasts during that 6-yr period are made through various forecast verification methods and are presented here. The skill or relative accuracy of the forecasts and some verification measures are quantified by computing the Heidke skill score (HSS), Hanssen–Kuipers discriminant (HKS), threat score (TS), hit rate (HR), probability of detection (POD), bias score, and false-alarm rate (FAR). The study revealed that the T80 model has a tendency to underpredict rainfall over most of the subdivisions falling on the windward side of the Western Ghats and sub-Himalayan areas. The model exhibited negative bias in rainfall simulations over the...

https://journals.ametsoc.org/doi/pdf/10.1175/1520-0434%282002%29017%3C0832%3ACOWCRF%3E2.0.CO%3B2

Characterization of Weekly Cumulative Rainfall Forecasts over Meteorological Subdivisions of India Using a GCM

The CERES-Rice crop simulation model, calibrated and validated for the    varieties PR106 in NW India. IR36 in central India and Jaya in south India, is used for  nalysing the effect of climate change on rice productivity in the country. Plausible climate change scenario for the Indian subcontinent as expected by the middle of the next century taking into account the projected emissions of greenhouse gases and sulphate aerosols, in a coupled atmosphere-ocean model experiment performed at Deutsches Klimarechenzentrum, Germany, is adopted for the study. The adopted scenario represented an increase in monsoon seasonal mean surface temperature of the order of about 1.5° C over the south India and 1°C over northwest and central India in the decade 2040-49 with respect to the 1980s and an increase in rainfall of the order of 2 mm per day over south India while the simulated decrease of the order about -1 mm and -1.5 mm over northwest and central India respectively. The IPCC Business-as-usual scenario projection of plant usable concentration of CO2 about 460 PPM by the middle fo the next century are also used in the crop model simulation (CERES - Rice V3 Model).
 
Simulation studies carried out with the climate change scenarios over different parts of the country are analysed and interpreted.

S. A. Saseendran

Papers

Effects of Climate Change on Rice Production in the Tropical Humid Climate of Kerala, India

Vulnerability of rice and wheat yields in NW India to future changes in climate

Evaluation of the CERES-Rice version 3.0 model for the climate conditions of the state of Kerala, India

Characterization of Weekly Cumulative Rainfall Forecasts over Meteorological Subdivisions of India Using a GCM

Modelling the impact of climate change on rice production in India