192 – Twenty Years of RMS for CRFS: Crop Simulation Models

[00:00:06] Lucie: Hi, and welcome to the IDEMS Podcast. My name’s Lucie Hazelgrove Planel, I’m a Social Impact Scientist and anthropologist, and I’m here today with Roger Stern continuing our series of conversations about agricultural research methods.

Hi, Roger.

[00:00:21] Roger: Hello, Lucie. Nice to be back again.

[00:00:23] Lucie: Thank you so much for joining me and I think we are going to continue our last conversation, which mentioned crop simulation models. It’s something that I don’t know much about, so I’m really keen to know more about the whole topic, really, a general introduction about what these models are, how they exist, how did you hear about them perhaps?

[00:00:44] Roger: I’ll give you a quick history because I heard about them initially in the 1980s, so almost half a century ago, where they were just developing, and even before the era of microcomputers. These are computer programs, which, in the computer, model the growth and yield of crops.

Often these models work on what’s called a daily time step, and you’ve got to sort of then picture that what they do is step through the crop growing. You plant it, and then after a time the crop emerges, and then the roots are growing and the shoots are growing, and then they produce leaves, and afterwards they produce the fruit.

[00:01:30] Lucie: So it’s able to model all of those different steps as though it was happening on a daily basis. Wow.

[00:01:35] Roger: That’s right. Some of them are even in more detail than daily, and there’s many of these models, but there’s two general systems that still seem very active and have grown gradually over the last 40 years, and are very impressive crop simulation modelling systems.

One of them is called DSAT, and that is decision support system for agricultural technology, or something like that.

[00:02:03] Lucie: Agro technology transfer.

[00:02:05] Roger: Yes, I don’t quite know what the transfer means, but it’s to help people make decisions. And that’s developed in the United States.

And there is a system called APSIM, Agricultural Production Simulator, which is developed in Australia, and they continue to improve. Currently they’re used quite routinely, both for research and even by individual farmers in Australia and the States and in some other countries.

[00:02:37] Lucie: Are they open to use?

[00:02:39] Roger: I think you can use them freely, I don’t think there’s any license to pay. But you need to have the data to put in. But they supply quite a lot of the data, so you need data on four topics. You need some climatic data, and usually rainfall, temperature, and some form of radiation, like sunshine hours. So you need that for quite a lot of years if you’re going to run the model for quite a lot of years. Then you need information about the crop, you’ve got to know what crop it is. And different varieties of the crop have different parameters, and you need crop information. You need soil information because the crop grows in soil. And you need management information.

So those are the four types of information you need. And examples of this are built into the software. So if you are growing a crop of maize, you’d find different varieties of maize would be there pre-programmed. And if you were growing a new variety, you might have to relate it to the varieties that were there in terms of its properties,

[00:03:49] Lucie: Okay, that’s interesting. So even if someone has developed their own variety of a crop that exists in the model or in a model, then from their own understanding of that new variety, they should be able to sort of determine which model to use and how.

[00:04:01] Roger: Link it in and see which parameters to adapt because it’s like this one but matures more quickly, for instance and so you change the way it matures and therefore it goes through different phases and you would adapt those things. So you might have to adapt things if you think you’ve got a new variety, but quite often you’d find that it has the characteristics built in of crops that are very similar.

Now, what’s different in your case, and often very different, is that you might have a sandy or a clay soil and there’d be some soils built in, but your climate is different. And so most places would put in local information on the climate.

And then, you can run these just for a single season, and then you understand in that season how the crops matured and grew and what yields you got. But you could then model it, and you can design an experiment, just as in the field, you could run through with two different levels of fertiliser to say for that year, which one did I think would’ve done best in the computer.

[00:05:10] Lucie: What I found surprising, previously discussing it with you was that these models often are historic ones, you don’t look in the future as to what you think the crop will do, you look in the past to what it could have done, because you have the historic climate data, you don’t have the data for the climate for what may happen this coming year.

[00:05:27] Roger: This isn’t your magic bullet for climate change, the main thing it adds to field experiments is a field experiment you can only do for two or three years because that’s your grant. Whereas you might have climatic data historically for the last 40 years, and it’s a very interesting question to say what might they have done if I had run them through for lots of years.

That’s a really rather important element because often, although it’s not used in experiments very much, they just report the climate, the variability of rainfall and temperatures, and also radiation, the climatic variability causes tremendous changes in the way the crops grew and the yield they produced in the different years, and that is ignored.

So understanding what would happen in a relatively dry year, well, you can take the dry years and so you can ask those different questions. Or you can ask the questions, what happened generally in the pattern over the last 40 years? Now, when it comes to climate change, these are not models of climate change per se, but we do know, for example, that temperatures are rising.

So there’s nothing to stop you in the model, taking the historical data and for example, very simplistically increasing the temperatures by two degrees and say what would’ve happened if the temperatures had been exactly the same but two degrees higher, which is what we expect might be the case in the next 50 years. So you can add those sorts of things.

What you can’t do is, of course, many of the climate people would be aghast because you’re changing the temperature, but everything in the climate is linked, and so the rainfall could well be changing and the radiation would be changing as well. So you can do simplistic changes in one or more of the elements, but this isn’t so much a prediction.

But, I think, forgetting the climate change even, there’s scope for understanding the climate variability in the past and what would’ve happened in a dry year, in a very wet year, and things like this is extremely valuable adjunct to many of the research experiments, or the research trials.

[00:07:59] Lucie: Some of those different suggestions of both sort of potential futures and possible pasts as well, you can work on those and imagine them.

[00:08:08] Roger: There’s two sorts of different ways you use these models. I just want to mention that if you were, for instance, a large scale farmer in Australia and you were thinking of a different variety of wheat, which is similar to one, but different in particular ways, you could in an evening sit down and say over the last 40 years, if I had grown that new variety, how would it have compared with the variety that I was growing?

And both these varieties are likely to be able to be modelled. And you could have a decision, oh gosh, it would consistently have done worse than the one I’m using now, I’m not going to use it. Or it might have been better in some years and the current variety is better in others.

Now, the point I’m making is that I think there’s some use of things that can be run over quite a lot of years, and that if I was in Australia or in the States, where I have easy access to suitable climatic information, then I could in an evening even do a study.

So in terms of your research people in Africa, I’m not suggesting they change from field experiments to crop simulation experiments, I’m suggesting that sometimes it might be useful to add that small component to the current ways they’re doing.

[00:09:31] Lucie: Exactly. This is what I find really interesting, I mean, we’re seeing yet again that it’s not just one method that people should use, but combining different methods and different approaches, you know, actually sort of just talking with farmers about how they saw things.

[00:09:45] Roger: And, there are teams doing this, that are using these crop simulation models. From memory, there’s a very dynamic team at the university in Ghana, in Accra, doing crop simulation modelling.

However, it’s tricky enough to do in Africa, maybe I’m generalising too far, but generally people either become a crop simulation modeller or they don’t. You can’t use it as a simple addition quickly, that seems to be the problem. You either model or you don’t.

This is the parallel of, for instance, drawing complicated maps, that you look for a map expert rather than thinking, oh, I could do that myself. And similarly with crop simulation models, you either have to get into it in a big way or you don’t do it at all.

[00:10:35] Lucie: Because there’s so many parameters that need to be sort of modified and everything, there’s so many different types of data that need to be included in the model even.

[00:10:43] Roger: There are, but I think that’s a small problem compared in Africa to the problem that you need for each year to have the relevant daily climatic data without missing values in that particular year. Of course, if there’s missing values, you can leave out a few years, but actually getting hold of the daily climatic data that you need, it’s often quite tricky.

So if we ever were to encourage people in your teams to be doing that, we ought to have a parallel exercise that helps them to be able to access suitable climatic data, without a tremendous effort, to be used, because that is a fundamental component of these models.

So if there were a block, it’s ironically not a crop block, it is a climate data block that is making it a more formidable task in Africa compared to the ease with which you can get suitable climatic data if you come from Australia or the United States.

[00:11:52] Lucie: Okay, so for researchers who have got some access to suitable climate data then, I think you’ve said that, you know, it can be useful for specific types of agricultural research, especially long term agricultural research perhaps also, like if you’re doing intercropping.

[00:12:06] Roger: Yes. Well, intercropping, the models build in competition, and particularly I think APSIM, I don’t know the details of these, but they do build in, because there’s a lot of mixed cropping. And even if you don’t have mixed crops on your field, you almost certainly have weeds, which is a form of mixed crops, even those one of those is not as useful as the other.

And so, mixed crops have got to be a component of the model, so they are coped with. There are other aspects, planning, if your interest was in different competing rotation schemes, this is very difficult to do in standard field experiments.

However, the reason for rotation is often that certain aspects of the rotation restore the fertility of the soil, and so they are wonderful candidates to be considering a crop simulation model rather than trying to design a field experiment and finding that it’s really almost impossible to do a constructive field experiments, certainly within the limitation of the few years that many research projects are given.

So there are instances where the field experiments just cannot answer those questions, and also where an aspect of the answer could be useful. I should just mention that, I am now thinking that as part of the research support, helping people to use these models could be very constructive. It would have to include the climatic component.

However, people use these models in two contrasting ways. The modellers enjoy very much the continued improvements that are needed in the model. How well is it modelling the below ground component?

[00:13:56] Lucie: Which must be a lot work to try and get right.

[00:14:00] Roger: Well, those are the parameters within the model. And professional crop modellers continually try to improve the models, and that’s right and proper. I’m not thinking of that. I’m thinking of accepting the models where they’re working okay, and you see, if you’re trying to improve the model, you can spend a lot of time on one year of data.

Whereas I’m thinking that one of the main uses is to understand the performance of your particular crops or options that you are recommending over a long period of years. What would’ve happened in rather wet years, in hot years, in years with longer dry spells and so on?

So that it is the fact that with the model with 40 years, you are modelling over quite a lot of climatic conditions, which you don’t get anywhere near doing with the standard research that people are doing now. So it’s filling a hole in the research for the people that are using the models rather than critiquing and improving them. Both of these are very valid. But I’m wondering whether there is scope for a few more users in Africa than we have currently, and whether that could be interesting for some of your research projects.

[00:15:19] Lucie: Yeah, absolutely. And I think you’ve mentioned before that the models, well, firstly, I know I’ve heard before that no model is correct or something, but some are useful.

[00:15:28] Roger: All models are wrong but some are useful. Yes.

[00:15:31] Lucie: So one of their uses is also, not necessarily in giving answers, but in asking questions.

[00:15:38] Roger: That’s right, yes. Absolutely. And the question, part of the answer to the question might be, it’s not modelling that part very well, but it still would’ve been worth trying. I’m convinced there’s quite a lot of situations for a number of your projects where it ought to be worth trying these simulation models. They’ve become easy enough to use, and they need the support, but that’s why there is a support team for the project. I emphasise that I’m not thinking of it substituting for any of the current research activities, I’m thinking of it making a useful addition and complimenting them.

[00:16:16] Lucie: Can I ask, like, this especially talked about crop simulation models, are there other simulation models that exist for livestock, for pests and diseases?

[00:16:28] Roger: Lovely question. Yes and no. I’ve mentioned just DSAT and APSIM, which are very general modelling systems. They may have changed recently, but they weren’t very good on the pests and disease component, which is an important component of course. And they didn’t cope with livestock.

Now, there’s a lot of other teams that are developing these models, and the pests and diseases have tended to be where you need a special model for each disease or each pest. So they’re more specific models.

[00:17:02] Lucie: Because there’s so much more difference between how each pest acts or lives, you know, how its lifecycle works.

[00:17:09] Roger: That seems to be the situation that these are different enough. So I don’t know how well they are modelled within these general systems. You see these general systems cover many, many crops. But there are certainly models for livestock as well, and there are models for pests and diseases. Wageningen is another organisation that is very strong on the modelling side.

[00:17:31] Lucie: The university.

[00:17:32] Roger: The university in the Netherlands.

So there are many, many other models and my impression is that they’re more specific for the pests and disease component. But I’ve gone beyond the subjects that I’m very knowledgeable about.

[00:17:47] Lucie: But I’m just realising that, of course, really, it should have been George Simmons interviewing you about these simulation models, seeing as he’s working on, well, on CASAS’ pest and disease modelling system, I think.

[00:18:02] Roger: Now, I think it’s really rather important that the support team stays very involved in these. So maybe there’s a parallel discussion where you discuss with George Simmons and we have another podcast on that. Let’s see.

[00:18:18] Lucie: Great. Is there anything else you’d like to, any other advice you’d like to give to researchers listening to this about crop simulation experiments?

[00:18:25] Roger: Very interested in reactions and whether there are researchers that would become keen to take this further. And then of course to understand what your team would be doing to provide the necessary support for it.

[00:18:40] Lucie: Yes, I know, it would be a big new step for us.

[00:18:44] Roger: This could come back to bite you. I’m sorry about that.

[00:18:49] Lucie: Thank you, Roger. Thank you for causing problems to us.

[00:18:52] Roger: Okay. Anytime.