Eyes on Earth Episode 76 – ECOSTRESS and Disease Risk

ANDY MACDONALD:

The ultimate goal in this field is to accurately predict disease risk or epidemics of disease before they happen. With this sort of high-resolution remote sensing information about things like temperature and other variables that are really increasingly available in near real time, could we actually forecast risk and epidemics before they happen?"

JOHN HULT:

Hello everyone, and welcome to another episode of Eyes on Earth. We're a podcast that focuses on our ever-changing planet and on the people here at EROS and across the globe who use remote sensing to monitor and study the health of Earth. I'm your host for this episode, John Hult. Spaceborne sensors orbit hundreds of miles over our heads. Even the most advanced among them struggle to capture high-resolution imagery of individual human beings. Mosquitoes, of course, are far smaller than we are. So sensors on a satellite or a space station can't actually see them. Even so, these sensors can gather a host of information that helps us to understand the movements and behaviors of these pesky little disease vectors, which are responsible for at least a million deaths a year. Mosquitoes are more active under certain environmental conditions, and those conditions can be tracked at wide scales from above. Changes to the land surface can also make it easier for mosquitoes to proliferate. Today's guests recently used remote sensing for a study on West Nile virus risk in the California San Joaquin Valley. West Nile is among the diseases carried by mosquitoes. In the San Joaquin Valley, temperatures fluctuate mile by mile across orchards, agricultural fields and urban areas. They also fluctuate throughout the day, which of course factors into the risk for mosquito bites. Anna Boser is a Ph.D. student at the University of California, Santa Barbara. She collaborated on the study with Andy McDonald, an assistant professor at UC Santa Barbara's Earth Research Institute, as well as with others. The team used data from the ECOSTRESS sensor on the International Space Station to track land surface temperature variations in the valley as a measure of mosquito bite risk. MacDonald has previously worked with satellite data to study the link between malaria and deforestation in the Amazon. This is interesting stuff, so we're really glad to have them on the show. And oh, by the way, ECOSTRESS data is available through NASA's Land Processes Distributed Active Archive Center, or LP DAAC, which is located on site here at EROS. So without further ado, Anna, Andy, welcome to Eyes on Earth. 

ANNA BOSER:

Thanks so much. 

MACDONALD:

Thanks, John.

HULT:

Let's get started by talking about your fields of study. It's my understanding that neither of you came to remote sensing as geographers, at least initially. You each focus on public health and disease. Tell us about your research, and how each of you came to see these spaceborne measurements as tools for this work. 

MACDONALD:

I was trained in ecology and evolutionary biology. I'm what you might call a disease ecologist. I just happen to be interested in organisms that cause infection like parasites, bacteria, viruses, and how they interact with wildlife hosts that they infect, with vectors that might transmit them like mosquitoes, and with humans. My research is often focusing on vector-borne diseases. These are things transmitted by biting arthropods like ticks, mosquitoes, sand flies. Lyme disease, malaria, West Nile, just to name a few. These vectors are all ectotherms, unlike us, so they're really strongly influenced by abiotic conditions like temperature, like humidity, like precipitation, as well as by things like vegetation cover or different types of land use that they might use as habitat to look for hosts to feed on, or for reproduction, for example. The types of questions that I'm often interested in addressing surround how changes in these environmental conditions might influence disease risk or changes in human disease transmission. I could, and sometimes do, go out and deploy something like a mosquito trap in the field to figure out what mosquito species are present, how abundant are they? Maybe I'd put a data logger out there as well to measure things like temperature, humidity around that trap station. But ultimately, that's just one data point, and it's nearly impossible to generalize across really large and heterogeneous landscapes with that sort of data. The power of remote sensing is that you can get at these, close to globally consistent measurements that matter to disease ecology, that matter to transmission, things like temperature, which is the focus of our study, and then use this spatially and temporally explicit information to model something like temperature-dependent disease risk across, you know, large regions and across long periods of time. Or, in the case of a lot of work that I do, to link these measurements with surveillance data, so things like human disease, case reporting or mosquito surveillance to statistically assess the contribution of changes in these remotely sensed environmental conditions to changes in human disease risk. 

HULT:

So basically you are looking at, at bugs, at vectors, who are healthier, happier, and there are more of them if there are certain conditions present. So you study those conditions as a way to better understand disease risk. Anna, how did you come to, I assume, some of the same conclusions?

BOSER:

I was interested in looking at prescribed burns and how they affect air quality. So prescribed burns are just these really low intensity fires that we light on purpose so that we can clear out fuel loads and avoid really high intensity and dangerous fires in the future. Obviously, wildfire affects air quality, but what about these little fires that we do to prevent these high-intensity fires? The issue here is similar to Andy. When he puts out his sensors for mosquitoes, it's really hard to get a lot of them, enough of them to really figure out what's going on. The same idea applies to air quality sensors. There's just not that many around, especially out in the forest where prescribed burns are happening. So instead, I turned to remote sensing to look at smoke plumes from space. Since then, it's just opened up a whole world for me. I do pretty much all of my work with remotely sensed data, and it's really been amazing. You can really answer a huge breadth of questions. 

HULT:

Andy, let's talk a little bit about your previous research now. Talk about the deforestation work. Let's get into the details there. What did you find out when you looked at deforestation and mosquito bite risk?

MACDONALD:

In what is the largest remaining tropical forest, the Amazon in South America, we saw this really large increase in malaria transmission following government-sponsored settlement, of the Brazilian Amazon in particular. So they constructed the Trans-Amazonian Highway in the 1970s and '80s, which really opened up the Amazon to settlement by folks in Brazil. But despite this trend that we see over time, there's conflicting evidence in the scientific literature about whether deforestation does actually influence malaria transmission. You can imagine that this sort of lack of consensus can really hamper development of policy or interventions that you're aiming to use to improve either conservation on the one side or human health on the other. The sort of local scale ecological studies that have been conducted are fairly consistent. So these are studies that are looking at, say, biting rates of mosquitoes or transmission of malaria. Those are actually fairly consistent in predicting an increase in transmission associated with deforestation. Where you start to get more of the disagreement is at these larger spatial scales, where folks are using more remote sensing-based measurements. And so there was actually this sort of back and forth in the literature just a few years back. It got a little bit heated, you know, as academic debates can. Two of the hypotheses that we developed were first that maybe the relationship between deforestation and malaria differs between different regions. And then the second was that maybe there's a feedback from malaria burden back to land use and land clearing activity that could influence rates of deforestation, by impeding development of new settlements, or simply limiting the number of healthy workers that are available to clear forest in the first place. And so we set out to address these hypotheses, using malaria case reporting records from across the Brazilian Amazon for about 13 years. And we coupled this with varying types of remotely sensed data on land use and land cover, including estimates of deforestation, environmental and sociodemographic control variables. So temperature, again, was really important in this context to control for. What we found was when we are looking at the entire Brazilian Amazon as a whole, this really significant positive effect of deforestation on malaria transmission, given higher rates of deforestation in a given municipality-so this is something like a U.S. county, for example-we tend to also see higher rates of malaria transmission in that year. But we also find a significant negative effect of malaria burden, so basically how large those malaria epidemics are in a given year on rates of forest loss. If you have higher rates of malaria in a given municipality, then we tend to see lower rates of deforestation than we would otherwise expect. So this sort of agrees with one of our hypotheses. There might be this sort of feedback between malaria and deforestation in the Amazon. But, you know, if you don't account for this, you might actually underestimate the effect of deforestation on malaria, which would contribute to this uncertainty that we see in the literature. Getting to that second hypothesis about whether the relationship might differ between different regions, we broke our dataset down into on one hand, what's often called the Arc of Deforestation in Brazil, sort of on the edge of the Amazon rainforest, and then the interior of the Amazon rainforest. And we essentially reran our same models. And we actually don't find any evidence for these same effects in that outer Amazon region, where we've sort of transitioned away from lots of intact forest to a lot more sort of large scale agricultural settlements and production. So there's not that much forest remaining in those regions. And so in that context, you know, removing more forest doesn't actually have an effect on malaria. But in the interior, where we do have a lot of intact forest remaining, we see even stronger effects, consistent with our results that we got in the full Amazon. So here, deforestation had this really positive effect on malaria transmission, and then malaria burden in turn had this negative effect on deforestation. 

HULT:

That's super interesting. And you settled it. You totally settled it, right? 

MACDONALD:

Well, you never settle anything in science ... but, um, we like how the results came out. 

HULT:

Anna, let's turn to you talk about some previous research.  You wanted to talk about your work with irrigation efficiency in California and remote sensing. Let's get into that. What did you learn there?

BOSER:

What's cool about this is that it's using the same sensor that we used for the mosquito work. In the mosquito study, we were looking at temperature. But we can actually use information on temperature to kind of infer how much the Earth is, in a sense, sweating. So how much evaporation is happening, or transpiration, which you can think of as evaporation from plants. So how much water are we losing to the atmosphere? We can get some pretty solid estimates of that at a pretty high resolution using remote sensing. I look at this movement of water over agricultural landscapes, and I can figure out how much more water is moving into the atmosphere than would be were these lands natural. From that, I can figure out how much water we're consuming in agriculture in California. As a Californian, this is something that hits home for me, because I grew up hearing, you know, take shorter showers, because, you know, we're in a drought, and saving water is a really big deal here. Especially for agriculture, because 80% of the water that we divert for human use in California goes to agriculture. We can see how much water agriculture is consuming from space, which is something that we can't really do on the ground. Like I was saying, it's really hard to do. We can use these eddy covariance flux towers, there's these amazing towers that can kind of get a sense of how much is moving into the atmosphere, how much water is. But the problem with those is they don't actually work particularly well in very heterogenous landscapes. You can think of agriculture, you know, you'll have one field of corn right next to a field of almonds. And it can be really hard to figure out the difference between those two. So once I've got this idea of how much water we're consuming, I can see are some crops consuming more water than others. It's something that's been of a lot of interest lately. And then the other thing that I can see, which I'm especially interested in, is I can compare it to how much water we're irrigating with, and see how much is actually being consumed from that. If we're irrigating a lot of water, but only some of it is actually ending up going into the atmosphere, growing our crops, what is happening to the rest of it? This number is actually very small, the efficiency of our irrigation. I don't have a final number on it, but it's less than 30%. So where is the rest of this water going? Most of it is probably lost as runoff. It just is going to go right back into rivers, or maybe it can be lost in conveyance, you know, we take water from wherever we're taking it, either rivers or a reservoir or the ground, and just as we're trying to move it, it gets lost. So getting to the bottom of what is happening to our water, I think, is a really big priority.

HULT:

We have talked about ECOSTRESS and evapotranspiration on this show before. There's some pretty exciting stuff going on. And part of the reason that ECOSTRESS is so useful, as I understand it, is because you get multiple measurements throughout the day. With something like a Landsat or a MODIS sensor, they get information at the same time of day every time, so ECOSTRESS is valuable, as I understand it, because you get sometimes a couple of different measurements in the day and the nighttime. Tell me about your recent study. Why did you decide to use ECOSTRESS data? How did your ability to look at those diurnal patterns impact your decision to use ECOSTRESS? Was that a part of it at all? What did the sensor bring to the table? 

BOSER:

Yeah. So like you were saying, ECOSTRESS is really great because it's got this ability to take measurements at different times of day. And also, the measurements that it takes is a really high resolution. So it's 70 meters. It's kind of difficult to think of what that might mean, but it's smaller than an agricultural field, so we can resolve a lot of spatial variation, but also temporal variation. And the reason that that was so important to us is if you were a mosquito, you probably don't want to live in an area that has super, super cold nights and super, super hot days, even if on average that's a pretty comfortable whatever, you know, 70 degrees. You'd rather live in an area that's 70 degrees all around. The benefit of the ECOSTRESS data is that it cannot only get those averages, but it can actually get at the variations.

HULT:

Ultimately, you guys ended up producing 65 maps of land surface temperature using ECOSTRESS across the valley. What did you learn? What did you put together? 

BOSER:

The 65 maps were from across the diurnal cycle, but then we just average over all of these different measurements to get one map of how mosquito borne disease risk, or specifically West Nile virus risk, changes across the landscape. What did we learn from these maps? Well, the first thing that we learned is that this variation that we were just talking about is super important, not only in time, but also in space. So had we not had those 65 different maps to aggregate into this single one, we would have gotten a completely different result ... if we had just used an averaged air temperature. And the same is true if we had just averaged across space. We would have gotten a completely different answer on what the average biting rate or transmission probability of West Nile virus is. The first thing that we found is that ECOSTRESS really is super, super valuable for resolving that variation and giving us accurate maps of West Nile virus risk. The second thing I think we learned is that we can really get at some really interesting variation across space for disease risk. So we can see that in agricultural areas the risk is pretty different from urban areas, which is different from natural areas. And when I say risk and being a little bit loose about it, but what's really exciting I think about this study is we were looking at two different traits. We were looking at both mosquito biting rate, and also the transmission probability. And what I mean by transmission probability is, given the mosquito has bitten you, what's the probability you're actually going to get West Nile virus from that bite? And we can see that the variations in both of these types of risk are different across space. So we can get at some really ecologically interesting patterns there. 

MACDONALD:

It was actually quite surprising to me to see how strongly temperature was associated with those underlying land uses and how much we were able to resolve that with this approach. And, you know, mosquitoes are responding much more directly to this sort of microclimatic variation, not to large scale of climate that, say, a global climate model is working with. So I think there's a lot of potential here for linking large scale climate to microclimate through these sorts of relationships between land cover and say something like ECOSTRESS temperature measurements that might help us improve future forecasting for vector borne disease responses to climate change. 

HULT:

Well, see, that's where I wanted to go next. That's perfect, because I wanted to know, I wanted to know how this information might be useful in our response to mosquito borne disease. I mean, are we talking in the future about some kind of a portal where you can go as a regular person in California and say, "well, what's my risk?" Are we getting to that level? What might be done with this information.

MACDONALD:

The ultimate goal would be feeding this into, say, an epidemic forecasting system that can be used to, say, target interventions like mosquito control activities before the sort of amplification of the virus is happening in mosquito populations, or even targeting public health messaging during particularly risky times for transmission. 

BOSER:

I think what is especially useful about this study, well, number one, we can kind of forecast how things like land use change, and Andy was talking a little bit about climate change, as well, is going to be affecting mosquito borne disease risk. So we specifically can see differences over different land covers on how they affect risk. And in California, water is very scarce, and we are likely to be changing how we use land in California pretty significantly to try to accommodate for that. And so there's a lot of open questions there, like what is going to happen if we have less agriculture? Are we going to have more or less disease risk, for example. Or, you know, as the climate changes, is that going to make us more or less susceptible to diseases?

HULT:

We've been talking to Anna Boser and Andy MacDonald about tracking mosquito bite risk with remote sensing from space. It's been a fascinating conversation. Anna, Andy, thank you so very much for joining us. 

MACDONALD:

Thanks for having us, John. 

BOSER:

Thanks so much, John. It's been a pleasure. 

HULT:

And thank you to the listeners as well. Be sure to join us next time to learn more about satellites, remote sensing, land change and more. You can find all our shows on our website, usgs.gov/eros. That's U-S-G-S-dot-G-O-V, forward Slash E-R-O-S. You can also subscribe to us on Apple Podcasts or Google Podcasts, and the new shows will just show up for you. This podcast, this podcast, this podcast is a product of the U.S. Geological Survey, Department of Interior.