 Today, I am going to continue to talk about monsoon variability in agriculture as I mentioned last time a very very important topic and we were talking about our study region which is a semi arid region, semi arid part of the Indian peninsula over which ground nut is now grown extensively. And we said that this was not always so, the traditional cropping pattern was very complex. Now how did the transition occur from the traditional to the current cropping pattern which comprises primarily ground nut and a little bit of other crops as well. Now the traditional crops largely provided food grains for local consumption that is what agriculture was for in the olden days. The development programs that the government of India undertook after freedom from colonial rule such as road construction and electrification led to the spread of a market economy in this region. At the same time food grains became available at rates subsidized by the government this is in ration shops. And hence the need for cultivation of food crops for home consumption decreased. Now with the setting up of oil mills the price for ground nut increased and it became profitable for the farmers to change over to a ground nut based cropping system and this is what they did. So ground nut cultivation on a large scale in began in this region in the 70s with the introduction of the TMV2 variety which is grown even today in that region. Now we I mentioned last time that it is extremely important to progress in this area to have an interdisciplinary group of which a farmer is a very very important member. So farmers agricultural scientists and meteorologists were the three people three kinds of people who had to participate in this interdisciplinary interaction. Now first of all let us consider what the perceptions of the farmers are. Now many people have a doubt how much do the farmers know. In fact in order to understand the level of knowledge and approach of the farmers it is important to note that rain fed agriculture has been practiced for a very long time in India. The strong links between climate variability and agricultural production of the Indian region are very well known. In fact over 2000 years ago Kautilya you know who wrote Kautilya's Artha Shastra who is believed to be a contemporary of Aristotle stated that for agriculture and he was referring primarily to the Indo-Gangetic Plains. The optimum distribution of rainfall during the summer monsoon that is June to September is one with one third of the rainfall in the first and fourth month and two third in the middle two months. This was Kautilya's perception of what is the optimum what is happening is going back instead of front sorry. Some of the crops grown today such as sorghum and pigeon pea have been cultivated in India for over 2000 years. Even relatively recently introduced ground crop has been cultivated for more than 100 years. Not surprisingly farmers have considerable knowledge of the nature of the impact of the rainfall variability in the region on traditional crops. This knowledge is reflected in several proverbs such as the following from our study region Uttara Chusi Yattara Gampa wait for the Uttara rain that is rain fall during 13 to 26 September if it fails leave the place. So if the Uttara rain or 13 to 26 September period rain fails then it is much better to simply leave the place and not worry about the ground nut in the field or the following from the Indo-Gangetic Plains were Kautilya resided avat aadar nahi diyo jat no diyo hasta bina kate bina kane dona gaye pahuna or gresta. So while the coming monsoon in during the coming monsoon if there is no rain in one of the nakshatras and while going during hasta there is no rain then both the host and the guest will have to go without food see Ardra is the 22nd June to 5th July so this is when the onset occurs onset of the monsoon occurs over the Indo-Gangetic Plain. So he is saying that if there is no rain during the onset phase that is 22nd June to 5th July and no rain in the phase when it retreats from this region 27 September to 9th October then both the host and the guest will have to go without food in other words the crops will fail. So these people had considerable knowledge of the impact of the variation of rainfall on the crops and in particular which were the sensitive periods in which rain is very important and without rain you would get crop failure. Now notice that the time units used by the farmers Ardra, Punarvasu or Uttara and so on all over India are not weeks or months but 13 to 14 day periods called nakshatras based on the solar calendar. It is very important to remember that what they use are nakshatras based on the solar calendar which I show here and it begins with of course beginning of the year is Ashwini 13th April to 26th April it goes on to Bharane Krithika Rohini and so on then we come to Punarvasu which is 6th July to 19th July which is the time after which people start showing in our region in fact they start showing during Punarvasu itself in ground earth. So Punarvasu, Pushya and so on and so forth and you heard the name Uttara mentioned Uttara is 13 to 26th September and Hastha is 27th September to 9th October. So knowledge of the impact of variability of rain during associated with these nakshatra during these periods the impact on crops is very well known and in fact the proverbs that you looked at just now are a result of this knowledge or a reflection of this knowledge. So now meteorological data are available for several stations in the country for over 100 years but equivalent to a Panchang meteorologists have not derived information on variability of rainfall from these data in these time units that the farmers use. Now I just want to emphasize one more point because the word nakshatra in India is used also for lunar nakshatras and these are the nakshatras that people refer to as the star under which one is born and so on and horoscopes involve these nakshatras. But what the farmers use are solar nakshatras so they are by and large fixed by calendar date with a day or two this way or that way and not much more. So they are really equivalent to using the biweekly or two week periods instead of weekly or monthly rainfall for which we generally make climatological averages. So it is not at all difficult given the data that we have to make average rainfall or decide on what is the likelihood of a certain spell in the nakshatra and so on for each nakshatra but this had not been done earlier. Now I want to talk a little bit about impact of rainfall variability on the present cropping system. So although farmers have considerable knowledge of the impact of the local climate and its variability on traditional crops they have not cultivated the present cropping pattern sufficiently long to provide these insights for the newer crops. So the for the crops that they are cultivating they do not have as many insights as they had for the traditional crops. In fact our collaboration on monsoon variability in agriculture began when one day in early October Shesh Giridhau who was then a student at the centre for ecological sciences in our institute who is also a farmer in the Pavgadananpur region asked me why it rains so often during the harvest season nowadays. This was his question why is it that nowadays it is raining so often during the harvest season of groundnut. Now just to tell you when the harvest season generally occurs generally sowing is done during 22nd June to mid August depending on when the soil becomes moist enough. Now harvest is about 100 to 120 days later this implies that if the sowing is done early in the sowing window that is to say around 22nd June harvest has to be done in early October counting 110 days whereas if it is done in early August then the harvest is in mid November. Now Shesh Giridhau was very surprised when I told him that the nature of the variability of the rainfall over the region is such that the rain is in fact maximum in early October which is when harvest time would come if one had sown very early that is around 22nd June. So the rain in fact is maximum in early October and the chance of wet spells more than 50 percent whereas it decreases somewhat by mid November and in fact in the next few slides I will tell you what we can derive from data for daily rainfall at Anandpur for about 90 years. So first of all we have to note that the rainfall does vary a great deal from year to year so it is important to look at daily rainfall and what you see is just rainfall within the rainy season for 3 years and you can see how different it is. This is the year in which it is reasonably well distributed with one dry spell here then in this year most of the rain came here and very early withdrawal if you like or cessation of rains whereas here there is only one genuine wet spell and small, small wet spells several of them occurred throughout the year this is 1997, 1988 and 1982. So the total rainfall also varies from year to year but within the season also we get considerable variation between dry spells which can be long like this one or this one and wet spells. So there is considerable variability of rainfall from year to year in this region but one can still see talk of the averages and this is actually the weekly rainfall. Now what you see here this is the mean rainfall, mean rainfall, weekly rainfall at Anandpur and what you find is that in fact the maximum is occurring in late September and early October this is the maximum rainfall after which it decreases there is a small peak towards end of May and early June but the major peak is this one and this is the 90 percent limit in other words that 10, 5 percent of the years have rainfall more than this and 5 percent of the years have rainfall less than this. So this is the have rainfall less than this, this is the 10 percent limit. So 80 percent actually of the years have rainfall between these two ranges the red line and the blue line and blue line is the 75 percent and 25 percent is generally just 0. So quarter of the years will always have 0 rainfall irrespective of the nakshatra this is 75 percent and this is 90 percent of the mean. Now what is the weekly rainfall probability of wet spells again the same thing at Anandpur this is the probability of wet spells greater than 1 centimeter this is wet spell greater than 2 centimeter in that week. So weekly rainfall greater than 1 centimeter this is the probability and this is the probability for 2 and you can see again end of September early October you get highest probability of wet spells and correspondingly lowest probability of dry spells these are dry spells with rainfall less than 0.25 or 0 rain this is actually probability of 0 rain and this is probability of 0.25 rain. So you see that this period about which H. Girira was complaining is the period with highest chance of wet spells and lowest chance of dry spells and this we derived because we had data for so many years for Anandpur. So the farmers were not aware of this important facet of the climatology of rainfall over the region when we say climatology we mean average behavior. So this is a very important facet of the average rainfall for the region that it has a peak in late September early October and the farmers were not aware of it this is why Shesh Girira asked me that question. Now this incidence drives home the importance of deriving needed information on rainfall variability with the rich data set already available at India Med Department using time units which are commonly used by a farmers in India namely Nakshatras. See these climatology is derived for monthly and weekly rainfall on a regular basis but this drives home the point that since farmers use Nakshatras as time units it would be a good idea to generate information on rainfall variability with these time units. So we have actually done this. So rainfall for farmers we have used daily rainfall data for 100 years at 170 stations these are all taluk headquarters in the state derived basic information on rainfall variability in the time units used by the farmers that is Nakshatras and this information is available both in English as well as in the local language Kannada. We have also developed an interactive system by which the probability of occurrence of given quantity of rainfall or wet spells or dry spells are given for any period chosen by the user for any station. This is available on the website of our centre at the institute. This is the rainfall variability of at Pawgada now from May to November Pawgada is very close to Anandpur and this is where the farm of Sheshgiri Rao is and this gives the mean rainfall for each Nakshatra for Pawgada and two neighbouring stations three neighbouring stations in fact and what you see is what you saw earlier that there is a high there is a peak in Uttara Nakshatra that is why the proverb said that if it does not rain in Uttara time has come to pick up your basket and leave because the mean rainfall in Uttara is very high and continues to be high in Hastha as well. So this is the nature of the mean rainfall but what is needed is not just the mean but probability of various events and this for Pawgada is the probability of rain of different quantum of rainfall in centimetres here during a specific Nakshatra Poonarvasu which is 6 to 19 July and what you see is that probability of absolutely no rain is very small but probability of very little rain is very high in Poonarvasu and then there is a long tail with smaller and smaller chance much smaller than 10 percent chance of some rain say 3 centimeter, 5 centimeter, 6 centimeter and so on and so forth during this 14 day period. On the other hand you see Uttara which is the one where you have assured rainfall you see that the probability of zero rain is actually less than 10 percent it tends to rain quite a bit up to about 10 centimeters during this Nakshatra after that the probability is somewhat lower for rainfall higher than 10 centimeter. The information like this has been generated for every Nakshatra for every station you can also ask the question what are the probabilities of seasonal rainfall in different ranges. So for Pawgada for example, we have seasonal rainfall between 31 and 40 centimeter 50 percent of the years have that between 41 and 50 26 percent have that 51 and 60 16.7 and 61 and 78.3. Now, we can also ask the question for some of the strategies we may need to know what is the chance of rainfall more than 30 centimeter 100 percent because all the years in the recorded history have had rainfall more than 30 percent 30 centimeter. What is the chance of rainfall more than 40 centimeter we are now talking of total seasonal rainfall well it is 50 percent. So, 50 percent of the years have rainfall less than 40 rainfall more than 50 centimeters likely only in 25 percent of the years and rainfall more than 60 centimeters is very very small it is only 8 percent. So, when one has to plan various strategies this variability has to be taken into account. Now, as I mentioned before there are wet spells and dry spells and these have major impacts on the growth of the plant as well as on pests and diseases because they can trigger pests and diseases in the area. So, what we have done also is to calculate the probabilities of say 3 to 4 days of wet spell once or twice 5 to 6 days wet spell once twice 7 to 8 days wet spell and then 9 days wet spell larger than 9 days wet spell which actually never occurs 7 to 8 days does occur during some of the naksha ras this is ashilesha makha pupba and uttara you can say uttara the chance is very high all together 3 to 4 days spell also is maximum in uttara chance is 19. Then such spells occurring twice is a chance is 2 and once is chance is 5 to 6 days once chance is 5 and 7 to 8 days 2 and 9 day more than 9 days 2 and so on. So, all these things have been calculated so that we know the climatology similarly for dry spells also and for dry spells you can see that again chances of 3 to 4 days of dry spells are also very large in uttara hasta chitra and so on and then there are of course longer dry spells possible which come before the peak and after the peak here. So, this kind of information is available so this kind of information that we have about you know how much it is likely to rain during the total season how many dry spells you are likely to get within a specific naksha tra and so on can be used by intelligent farmers in different types of decision making even on the basis of experience of 2 to 3 decades in cultivation. So, I consider now so this is as far as giving information and making information on rainfall variability available to farmers so that given the knowledge of the crop they are cultivating they can make some informed decisions about various strategies various farming strategies. Now, I can now we will consider how the optimum strategies can be identified by combining this kind of detailed information on rainfall variability with the use of crop models and this as I said is a very powerful tool we have now in our hand which is been validated over several regions for several crops. So, strategies for enhancement of yields in the face of variable climate this is what we are looking for. Now, most of the farm level decisions such as enhancement of the seed rate that is how many seeds you plant how many seeds you sow per acre or per unit area or application of fertilizers or pesticides involve additional cost. Now, it is important to remember that in rain fed regions such as this there are high levels of risk and low levels of production and the resources available for such inputs are very meager. We have already seen how the yield fluctuates from year to year and so the profit margins are low even when the there is no loss and there are years in which there is crop failure as well. So, farmers resources available to the farmers are not large and therefore, farmers tend to avoid investment in such things this is also because the benefit of such investment itself depends on the rainfall. For example, if the rainfall is very poor no matter how much fertilizer you have added you will not get enhancement of yields the yield will be low because the rainfall is very low. So, the farmers are not quite sure of how much enhanced yield they would get by addition of fertilizers in this situation when a rainfall varies from one year to the next and they do not know which a poor monsoon year is going to be. So, the farmers tend to avoid such additional expenditure, but there are some farm level decisions such as choice of sowing window which involve no additional expenditure, but can have a very large impact on the yield this has been documented and in fact, you may ask the question what is the sowing window. So, sowing window is a specific period in which farmers sow grounders seeds when there is adequate moisture in the soil. So, when there is an opportunity to sow that is to say when the soil is moist enough and the date is within the accepted sowing window such and such a date in June to such and such a date in August for example, then that is called a sowing window. Now, in the package of recommendations developed by agricultural scientist sowing in may June is recommended for this ground nut it is suggested that in the absence of sowing opportunities in may and June which means if it did not rain enough in may or June. So, that the soil was never moist enough for sowing to be done then they say it is alright then you can do sowing in July if the if it rains adequate in July. So, that the soil moisture is adequate. However, if no opportunity occurs till the end of July then the farmers are advised not to sow ground nut at all in August in fact, they recommend that some other crops should then be sown. However, on the basis of experience of about 2 decades farmers do not sow until late June. So, although the recommendation is that you can sow in may if the soil is moisture is adequate they never do that. They do not sow until late June and do so in August if no opportunity occurred earlier despite the recommendation to the contrary remember the recommendation is that if you do not get a sowing opportunity till end of July then please do not sow ground nut in August that is the recommendation. But, farmers anyway sow in August despite the recommendation to the contrary because they have experienced that in some years when the sowing was delayed to August they got very good yields not all years, but in some of the years in which the sowing was delayed to August the farmers got very good yields. So, they do not believe that the recommendation is right. So, the window now they have adopted there is from 22nd June to mid August and in this window farmers generally sow at the first opportunity that is to say if the date is within the sowing window and the soil is moist enough they will sow at the earliest opportunity. Now, this is the situation. So, what is the problem posed by the farmers given the background of the recommendations which the farmers did not believe because they were inconsistent with the experience they had what when we are called for a meeting of the farmers and ask them what are the problems that you would like us to address in trying to see how best to enhance yields of ground nut in the face of rainfall variability. So, in that meeting the farmers suggested that one of the most important problems is identification of the optimum sowing window and what did they mean by optimum that which is associated with maximum production in the face of rainfall variability of the region. There is another way to look at the optimum as well when can also look at the optimum sowing date as a sowing window as that which avoids crop failure altogether. So, minimum chance of crop failure or maximum chance of high production would be another set of goals that one could pose. Now, we of course, used modern tool to try and figure out what would be the optimum sowing date and the model we used was the peanut grow model. See we were very fortunate that there was a group at Ykriset led by Dr. Singh who had actually worked on this peanut grow model with the people who developed this model in US, BOOT and others and who later on in collaboration with scientists at the Anandpur Agricultural Research Station compared the simulated yields from peanut grow with the observed yields at the Anandpur Research Station and actually showed that the model did very well. What you see here is simulated and observed yields from 1979 to 1990 and you see that the model is able to actually capture the rainfall the impact of rainfall variability on crop yields from year to year. I must of course, emphasize that this model is indeed for rainfed crops it is not for irrigated crops and this result of Parasim and others which was published in 1994 showed very clearly that in fact, the model is able to capture the variation induced by the rainfall variation on the yields of ground at Anandpur. So, this was a very big asset that we had a model that was already validated for the crop variety and for the reason of interest. So, this was a big asset and so, work done at by Singh et al at Ikrisat showed that this model is able to simulate year to year variation in the yield of the variety TMV2 of ground at cultivated in rainfed conditions at the Anandpur Agricultural Research Station. Now, generally the average yield at the district level is less than the model simulated yield primarily, because of the incidence of pests and diseases. Because in real life there are pests and diseases that are incident and the model does not have any impact of pests and diseases in it. It is only a model in which the plant grows it experiences moisture stress due to you know dry spells and so on, but there are no impacts of pests and diseases incorporated into the model. So, the model yield can be considered as a potential or maximum yield under rainfed conditions. This is the best you could do in some sense in rainfed conditions if you could somehow curb the pest disease incident. Now, the average yield for the region is about 750 kilogram per hectare and crop failure is said to occur when the yield is less than 500 kilogram per hectare. Since at that yield level it is not adequate to meet the cost of cultivation. So, when the money you get by selling the ground nut is less than the cost of cultivation it is a failure it is considered to be a crop failure. So, what we do is we consider a model yield of of course, less than 500 kg per hectare as a crop failure because if the model yield is less than 500 the farmers yield is bound to be less than 500. So, this is a crop failure then we consider between 1000 to 1500 kg a hectare as above average and greater than 1500 kg per hectare as very good yield. So, these are the three criteria we use. Now, what is our problem then the optimum showing that for minimizing risk of crop failure is thus one which corresponds to minimizing the probability of model yield of 500 kg per hectare or less and maximizing the production which is which maximizes the probability of yield above 1500 kg per hectare. Now, sensitivity to different meteorological inputs. Now, we we got this model peanut grow model from Ikrisa from Dr. Paira Singh. Now, it required various meteorological inputs for example, it required daily values of maximum minimum temperature radiation and rainfall. Now, all these data were available at the Anantapur Agricultural Station for 79 to 98 which was a very good thing and that is the period for which you saw that the model was shown to be validated by Paira Singh and others. Now, Anantapur itself has rainfall data from 60 to 279, but does not have data on radiation and maximum minimum temperature. In addition to the data at Anantapur Agricultural Station rainfall data at Anantapur Meteorological Observatory was available to us from 1911 to 1990 at the time we undertook this study. So, we have all the meteorological inputs only over a short period, but if we could use the model for the entire period for which rainfall data is available which is a long period here 80 years, then then actually we would be able to get much more out of the model. So, the first question was are data on temperature radiation and temperature radiation that essential does it make a difference whether you put in the year to year variation of daily values of maximum minimum temperature radiation into the model. So, this is the first thing we did we tested the sensitivity of the model for and we wanted to use the model to study variation of yield with showing date that was very clear. Question is the results we get for variation of the yield with showing date using rainfall and averages daily averages of temperature and radiation are they different from results one would get with rainfall as well as actual data on temperature and radiation. This is the first question we asked and what we found is that really the most critical element for the yield is the rainfall. So, variation of the model yield with showing date based on all the meteorological data is found to be very close to that obtained by replacing the daily temperature and radiation by the daily averages for 79 to 88. In other words we are giving as an input the same pattern of daily variation of temperature and radiation year after year what happens then. What you see is black is the actual and red are the points which are the where we use daily averages of temperature and radiation and you can see that the matching is in fact almost too good to be true which shows that the variation and this is for different years as you can see and in different years the variation of yield with showing date is different. But you can see very clearly that even if we use daily averages of temperature, radiation and so on the basic features are captured extremely well we do not have to worry about getting data which varies from year to year on temperature and radiation. So, this was a bigger set. So, the daily rainfall data from 1911 to 1990 can be used along with the climatological averages of the daily maximum and minimum temperature and radiation for deriving the variation of the yield with showing date. This is the first conclusion and that made it possible for us to use a big long data series rather than being restricted to some 10, 15 years where all the meteorological data that the model demands actually are available. Now, let us see what the results that we got and in these results what we have done is sorted all the results into different types you will see it here. See the pattern this is the first pattern this is the first pattern and these are different years actually shown here and the common thing about these is that by and large the maximum yield occurs for early July or even earlier. This year it is even earlier. So, the peak is occurring here for early July this is the pattern for about 19 years that you see. Then there are 16 years which is shown below in which the peak occurs much later. So, this is when optimum showing date is between 20th and 31st July. So, it is only towards the end of July between 20th and 31st July that you have a peak and these are again several years 16 years these are 19 years. So almost as many years the peak is somewhat later this is in early July or even earlier this is now in third week of July till the end of July. Now these are all years in which the peak is much later and in fact it does not vary all that much you know once the showing date has gone beyond this for many, many years. So, for some years like this you can see that the showing date is early August the optimum showing date is early August maximum yield you get if you showing early August after that it is flat. So, this is a case in which showing date has now become even later and this is 14 years for which it is now early August and in this case the optimum showing date is between 1st and 15th August this is again class 4 which are 15 of them. Now this is where the optimum showing date is either in early June or in late August this is somewhat slightly confusing these 12 years or so. There can be a peak in early June and another very often bigger peak in late August. So, here actually if one had to derive the optimum it would be more late August except for this particular year in which the peak is definitely in June. Then there are some years in which yield is very insensitive to showing date these are either years in which the rainfall is very good and very well distributed these are these years here where no matter where you show you get very good yield. There also years in which the yield is low no matter where you show and these are very very poor rainfall years in which also yield is not very sensitive to showing date. So, note that in almost all the years the yield increases as the showing date is postponed from May to late June. So, this is consistent with the experience of the farmers which led to the present showing window of 20 second June to mid August. The surprising result from these figures is that the yield increases with later showing even beyond July for many years. This is also consistent with experience of the farmers in the regions who refuse to you know stop showing ground at in August when the opportunity to enterize earlier. Now, yield is greater for showing in late June early July only in 19 out of 87 years later showing is associated with larger or at least as much yield in 80 percent of the years. So, in within that showing window itself 22nd June to mid August the first part of the window late June to early July relatively few years have optimum showing date it is the latter part of the showing window where 80 percent of the years have an optimum showing date. So, as the farmer suggested to us the recommendations about the showing window in the official package of recommendations are clearly wrong. Now, the variation of the probability of crop failure that is to say yield less than 500 kilograms per hectare and of probability of above average yield and very good yield is shown in the next slide. So, this is again results from the model mind you and what you see here on top is the probability of crop failure yield less than 500. You can see that this showing date we have started from 30th of April because they had recommended you can start showing from May itself and in fact the probability is very high for early showing in May and so on close to 50 percent chance of crop failure and it decreases sharply and it is actually less than 20 percent by about 25th of June. So, the probability of crop failure has decreased markedly to less than 20 percent by 20th of June and by early July actually it is less than 10 percent and remains flat. So, the farmers were quite right in abandoning the recommendation that the planting should be done in May or early July early June because then the probability of crop failure is huge more than 1 in 3 or more than or around 1 in 2. So, this is a very very large probability here now what is the probability of above average yield on the farm now that actually increases and becomes large from in the beginning of July and remains more or less same decreasing a little bit to end of August. Now, on the other hand probability of very good yield again this is from the same model it is a integral of all the patterns that you have seen. The probability of very good yield is very high if you restrict the showing to late July latter part of the showing window remember their showing window is 22nd June to mid August. So, if we restrict to say last week of July and first week of August then you get very high chance of very good yields. So, this is a very very interesting result. So, it is seen that the probability of failure is high about 40 percent for showing in May or early June decreases rapidly to less than 5 percent by early July. Also for showing in May or early June the probability of good yields is less than 20 percent it increases rapidly with showing date up to 6 August and then decreases a bit. So, for early showing not only is the probability of crop failure very high the probability of very good yields is also very low here and then it increases up to this maximum and then decreases a little bit, but not that much. So, the choice of the prevalent broad showing window 22nd June to 16th August is the appropriate one for minimizing the risk of crop failure. So, what the farmers have done is the right thing because they have now minimized the risk of crop failure. It also implies a 50 percent or higher chance of above average or good yields. So, this broad window which they have come to recognize the which they have adopted on the basis of the experience of about 2 decades is indeed and the appropriate one for minimizing the risk of crop failure and it is also reasonable because it implies more about 50 percent or higher chance of good yields above average or good yields. Now, for showing after mid July the probability of above average and very good yields is even higher than 50 percent. So, within the broad window if we look at the smaller window of showing after mid July mid July to mid August then is the showing window the probability of above average and very good yields is even higher and the chance of crop failure is still very small. So, what this model is investigation has done is to show us that actually within the large showing window is a smaller showing window in which the chance of getting higher yields is even higher than the showing window which farmers have adopted by trial and error. Now, it is very important to try and understand the reasons for this. Now, why is this specifics window optimum? We have to understand the reasons because we would like to eventually extrapolate the results to areas where we do not have to run the models if we can understand why is it that certain set a certain window is optimum for showing for maximizing yield then we would be able to check whether that criteria is valid in other regions and come to the conclusion as to whether the window would be optimum for that region for that crop without having to make all this large number of runs with the model. So, it is very important not only to get the results for Anandpur, but also try and understand why this has happened. So, to understand the reasons for this we have to consider the variation of moisture stress expressed by the plant during different life history stages. Now, why is that we know that in rain fed regimes the most critical element which limits the growth and yields of plants is the moisture stress this is because rainfall is canty and variable. So, we would like to now see what is the relationship between the moisture stress experienced by the plant in different life history stages to the optimum showing date that we have found through this model studies. Now, generally the need for water increases as the plant grows until the leaf development is complete at the end of about 60 or 65 days and remains high thereafter. Hence in a region with canty rainfall such as Anandpur the moisture stress also generally increases to a maximum by about 65 days and see now what we have done here is for a few years we have plotted what is the moisture stress in the model and advantages that the model actually computes the moisture stress on a daily basis while the plant is growing in the computer. So, we have the moisture stress in the model for a few years as a function of days after showing which is the x axis here and what we have done is purposely chosen suboptimal dates. In other words we have chosen dates which are not optimal showing dates for those specific years. We have chosen dates which are outside the optimum showing date and then ask the question how does the moisture stress vary during with days after showing and what you find is what I said before that as the plant grows the moisture stress increases it reaches a maximum when the plant all the leaves are out and then remains more or less steady thereafter this is the 65 days. So, from 60 to 80 days or 65 to 85 days the moisture stress experienced by most of the plants is maximum. These are the plants which are planted on suboptimal showing dates. Now, if we choose the optimum showing dates low and behold a very interesting phenomena occurs. If the optimum showing dates are chosen in this period 60 to 80 days actually the plant experiences no moisture stress at all. This is where ordinarily maximum moisture stress would be experienced by the plants this is when no moisture stress is experienced at all. So, for plants for which the showing date is optimum do not experience moisture stress at all during 65 to 85 days. This suggests that out of different critical stages suggested in literature the most critical is the pod filling stage which occurs say 60 to 80 or 65 to 85 days after showing. See we looked at a considerable amount of literature to ask the question which is the life history stage of the plant which is a critical life history stage in the sense that a dry spell during that stage would have a very large impact on the yield this is the critical stage. And you know the literature there is literature which suggests many almost every stage is critical. So, some papers say the first month after showing is very important then some people and so on and so forth every life history stage is mentioned in some one paper or another as being critical. But what this model has been able to show is that the most critical is the pod filling stage which occurs 65 to 85 days after showing. Now in fact we tested this result by more experiments with the model itself and that is what I will talk about in the next lecture. So, in this lecture then we have seen the impact of using a very powerful tool like a crop model and we have found that in fact the farmers choice of the showing window which was empirically determined on the basis of their experience is a reasonable one because it minimizes the risk of crop failure. But we also found that within that window there is a smaller window which would lead to higher yields than the farmers would get. And so now we are trying to understand why are the yields high in this smaller showing window which we would recommend to the farmers and link it with the moisture stress experienced by the crops and link it also to the critical life stages life history stages of the plant. If we can succeed in doing that if we can actually unravel what leads to the sensitivity of the crop to this dry spells in this particular time or lack of moisture stress in this particular time the pod filling stage then we would be able to extend the results of our study to cases where we do not have as many model runs as we have for this. So, in the next lecture we will continue with analysis of this model there is one more thing we will have to do see I mentioned that we have not taken into account pests and diseases in the model because it is not part of the peanut grow model. But in real life pests and diseases can cause a lot of damage therefore, we will I will talk about a heuristic model we develop for the losses created by incidents of pests and diseases how one can combine it with the peanut grow model and therefore, get closer to reality in terms of the yields one can get on the farmers fields. Thank you.