TRIGONOMETRIC PARALLAXES FOR 1507 NEARBY MID-TO-LATE M DWARFS

We selected images for inclusion in the astrometry fitting if they met the following criteria: The FWHM for the image is less than 5 pixels (3.5 arcsec), the average ellipticity of the target stars is less than 0.5, the target star is not more than 15 pixels away from its assigned location on the CCD, and the airmass at which the image was taken is not greater than 2.0. These selection criteria typically eliminate 50% of the MEarth images for each target, but can sometimes eliminate up to 80% of images. Most eliminated frames are eliminated due to a large FWHM, either due to naturally poor seeing or wind-shake of the telescope. For each target star, we selected a master image through an automated routine that selects an image that is of good image quality (low FWHM, ellipticity, sky noise), and good photometric quality (brightness of the stars is not significantly different from a typical exposure). The sky-coordinate system for this image is determined through star matching with the Two Micron All Sky Survey (2MASS) catalog. When many images meet the criteria to be a suitable "master" image, we select one image manually. We investigated the effect of our master frame choice on our final astrometric parameters and find that the choice of master frame does not significantly affect our astrometric measurements.

Reference stars are selected to be between magnitude 8 and magnitude 13 in the MEarth passband, are not blends or close binaries, and are unambiguously identified in the 2MASS catalog. Additionally, in order to avoid effects due to higher order plate corrections, we attempt to avoid using stars near the edges of the CCD. This is done by initially selecting only stars within 600 pixels of the target star (or in the few cases where there are multiple target stars, the average position of the target stars). If the total number of reference stars within 600 pixels is less than 12, then the radius of the circle is increased by 50 pixels at a time until at least 12 reference stars are selected. We always have at least 12 references stars in each field, and each star is weighted equally in our astrometric analysis.

Our astrometric time series are fit in an iterative manner, first fitting the plate constants for each frame, and then stellar motion parameters for each reference star. This is repeated three times so that the plate constants can converge toward a final solution. These plate constants are then used to fit for the motion parameters of the target star.

Each plate is fit through a least squares method with a six constant linear model based on the positions of the reference stars (the target star is excluded):

$$\begin{eqnarray} x^{\prime }_{i} &=& A_{i}x_{i} + B_{i}y_{i} + C_{i} \nonumber \\ y^{\prime }_{i} &=& D_{i}x_{i} + E_{i}y_{i} + F_{i} , \end{eqnarray} \tag{ 1 }$$

where x and y are the original flux-weighted centroid coordinates of the reference star in pixels, and A, B, C, D, E, and F are the plate constants. x' and y' are the reference star coordinates after the transformation, and are also in pixels. The linear plate constants allow for a different scale in the x and y directions, allow for translation of the frame in both directions, and corrects for any instrument rotation, as well as shearing motion in each frame. When removing the cameras on the MEarth telescopes for repairs or maintenance, disturbances to the data are likely. Our plate constants [A, B, C, D, E, F] remain very close to the identity matrix plus a shift, [1, 0, Δx, 0, 1, Δy] or a 180° rotation plus a shift, [ − 1, 0, Δx, 0, −1, Δy] over all images. The effect of higher order plate constants is mitigated by our selection method for reference stars, and including higher order terms doesn't increase the quality of our fit for stars with previously determined trigonometric parallaxes (see Section 3.5).

Once we have shifted and stretched the frame to align with the master frame, the coordinate system of the master frame (generated through star matching between the master frame and the 2MASS catalog) is applied to convert the x' and y' to α (R.A.) and δ (decl.). Then, we fit for both the proper motion for each reference star and the parallax for each reference star through the following method.

First, we remove the proper motion of each reference star since the time of the master frame:

$$\begin{eqnarray} \alpha ^{\prime }_{i,s} &=& \alpha _{i,s} - \mu _{{\rm RA},s} \Delta t_i \nonumber \\ \delta ^{\prime }_{i,s} &=& \delta _{i,s} - \mu _{{\rm decl.},s} \Delta t_i. \end{eqnarray} \tag{ 2 }$$

The subscript i denotes each image, s denotes the reference star, and the primed coordinate represents the transformation removing the proper motion since the master image. We then convert the stellar coordinates from R.A. and decl. to ecliptic longitude (λ) and ecliptic latitude (β) through a rotation of the coordinate system. Then, we remove the parallax motion at the image epoch and add the parallax motion from the master image:

$$\begin{eqnarray} \lambda _{0,i,s} &=& \lambda _{i,s} + \pi _{s}(P_{\lambda ,0,i,s} - P_{\lambda ,i,s}) \nonumber \\ \beta _{0,i,s} &=& \beta _{i,s} + \pi _{s}(P_{\beta ,0,i,s} - P_{\beta ,i,s}) , \end{eqnarray} \tag{ 3 }$$

where π is the parallax amplitude, and P_λ and P_β are the parallax factors in each coordinate for each star, s, and each image, i:

$$\begin{eqnarray} P_{\lambda ,i,s} &=& a_i \left(\frac{\sin (\lambda _\odot - \lambda _{0,s})}{\cos (\beta _0)} \right) \nonumber \\ P_{\beta ,i,s} &=& {-}1.0 \times a_i \left(\frac{\cos (\lambda _\odot - \lambda _{0,s})}{\sin (\beta _0)} \right) , \end{eqnarray} \tag{ 4 }$$

where a_i is the Earth–Sun distance in AU at the time of the image, i, or the master frame, 0, and λ_☉ is the solar longitude at the time of the image or the master frame.

Finally, one of eight constants is added to each individual star, for all images, depending on the side of the meridian the image was taken

$$\begin{eqnarray} \lambda _{f,s} &=& \lambda _{0,s} + G_{s,1,2,3,4} \nonumber \\ \beta _{f,s} &=& \beta _{0,s} + H_{s,1,2,3,4} , \end{eqnarray} \tag{ 5 }$$

where the subscripts 1 and 2 represent different sides of the meridian for the years 2008–2010 (before we changed the camera housing), while 3, and 4 represent the different sides of the meridian from 2011 September onward. Typical values for G and H are 0.1 arcsec. All data points are weighted equally, and our model is fit using the Levenberg–Marquardt χ² method. We note that in order to avoid degeneracies between the parallax and the meridian constants, G and H, it is necessary to obtain data on both sides of the meridian during the same phase of Earth's orbit, and our data collection strategy was adjusted in the middle of the 2011–2012 observing season in order to resolve this degeneracy. If this degeneracy was present in pre-2011 data for a field of view, then those data are discarded. However, as all observations prior to 2011 were taken at planet-hunting cadence only, this occurrence is very rare, as we observed each field during the entirety of each observing night. Finally, as the meridian constants are unique to each individual star, we note that the effects of any differential color refraction (DCR) can be partially corrected for through this method as well.

The plate constants, A_i, B_i, C_i, D_i, E_i, and F_i (one per image), and the stellar motion parameters, μ_{s, RA, DEC}, π_s, G_{s, 1, 2, 3, 4}, and H_{s, 1, 2, 3, 4} (one per star) are fit iteratively, while holding the other set fixed, until the solution has converged for the plate constants. If there are more than 12 reference stars available, reference stars whose motions are fit to be >1.0 arcsec yr⁻¹ or whose parallax place them closer than 100 pc are culled from the sample and the plate constants are refit with the remaining stars. We find only 23 reference stars in 17 fields fit these criteria. After these cuts, the median number of reference stars per field is 19 and the maximum number of reference stars is 90.

After the plate constants have converged, we fit the astrometric time series for the target star in the same manner as for the reference stars, fitting for the relative proper motion between the target star and the reference stars, the relative parallax between the target and reference stars, and the meridian constants (G_{1, 2, 3, 4} and H_{1, 2, 3, 4}). To avoid the effect of singular outliers, we refit our model after excluding points that lie outside of three standard deviations of the residuals for our initial fit. We perform an additional test on our data by refitting the parallax signal while holding the proper motion constant to the value reported in Lépine (2005). If the fit parallax amplitude changes by more than 2σ, then we discard this star from the sample. This test eliminated 41 stars from our sample, and we believe that this is principally due to a degeneracy between the parallax motion and the proper motion, which can be resolved in the future by gathering additional data at the proper phase of the year.

We estimate the internal errors in our trigonometric parallax measurement using a residual permutation algorithm, where we take the residuals from our best fit model, move them over one time stamp, add them back to our best fit model, and then refit the permuted data set for the parallax amplitude, proper motions, as well as our meridian offset constants. We apply this method only to our astrometric time series for the target star and not to the reference stars. This implicitly assumes that the errors in the frame constants (derived from the astrometric precision of the reference stars) are negligible compared to the errors in the astrometric precision of the single target star, which is valid due to the number of reference stars we have contributing to each plate solution. This method, while not allowing the flexibility of generating thousands of fake data sets (except for the minority of our stars that have over a thousand data points), has the benefit of preserving long time-scale correlated noise in our time series, and we find that our derived errors tend to be larger for stars that have fewer total measurements.

One concern with estimating our errors through this residual permutation algorithm is that for the systems which have less than 50 data points (for which we have 238 out of 1507 objects), there may not exist enough permutations to reliably estimate our error bars. In order to test this, we also estimate our error by refitting 1000 synthetic data sets generated by adding white noise to our observed data with a standard deviation equal to the standard deviation of our initial residuals. We find that the error bar we find from this method is comparable in magnitude to the residual permutation method, and therefore we elect to quote the error bar derived from the residual permutation method for our results.

Since our reference stars are all relatively bright stars, the reference frame itself exhibits a small parallax motion, as each reference star is also subject to the observational effects of trigonometric parallax. This effect causes us to systematically measure a smaller parallax angle and therefore a larger distance than if our reference image was static. We used the Besançon model of the galaxy (Robin et al. 2003)⁴ to estimate the parallax of synthetic populations of stars along the same sight-lines as our targets. For each target, we generate a synthetic star catalog oversampled by a factor of 1000, and select only stars whose apparent magnitudes are between 8.0 and 13.0 in the I-band (an adequate approximation of our non-standard filter for this purpose). We select random subsets of these stars that match our observed reference star magnitude distribution in order to estimate the average distance to a typical reference star and the associated uncertainty. Typical corrections for these stars are between 1 and 2 mas, but can be as high as 4 mas in certain directions. We note that the uncertainty in the absolute parallax correction is negligible compared to the uncertainty in the relative parallax measurement, and therefore only quote the error in the relative parallax measurement as our total error.

We release here a catalog of our results for each target star (see Table 2). If you want the best estimate for the distance to an individual star, use Column 18 and the error in Column 16.

The subset of the MEarth sample presented here includes 1507 stars for which we can obtain reliable results. This includes 240 stars for which we were able to locate a trigonometric parallax determination in the literature, many of which are from the compilation available in the Yale Parallax Catalog (167 stars; van Altena et al. 1995), Hipparcos (41 stars; Perryman et al. 1997), or Lépine (2005). A representative example of an astrometric time series with MEarth data (for LHS 64) is shown in Figure 1. For this star, we find a parallax amplitude of π_abs = 0.0412 ± 0.0017 arcsec, not significantly different from the previous determination in Harrington & Dahn (1980) of π = 0.0418 ± 0.0027 arcsec, as reported by van Altena et al. (1995). This data set is taken completely at astrometric cadence, demonstrating that our data collection strategy is sufficient to measure trigonometric parallaxes of our targets. We further note that we measure a parallax for GJ 1214 of 72.8 ± 2.4 mas, consistent with the recent determination of 69.1  ±  0.9 mas by Anglada-Escudé et al. (2013) and the historical determination of 77.2 ± 5.4 mas provided by van Altena et al. (1995). Therefore, we are able to accurately measure apparent stellar motion associated with trigonometric parallax, although in some cases with larger uncertainties than dedicated astrometric programs such as RECONS (Henry et al. 2006a), the Brown Dwarf Kinematic Project (Faherty et al. 2012), and the solar neighborhood project (Jao et al. 2011 and references therein). In Figure 2 we show the MEarth derived parallax motion compared to the values reported in the literature for all stars that had a previous trigonometric parallax measurement available. Literature trigonometric parallax measurements come from Monet et al. (1992), Harrington et al. (1993), Gatewood et al. (1993), van Altena et al. (1995), Tinney (1996), Ducourant et al. (1998), Benedict et al. (1999, 2000, 2001), Dahn et al. (2002), Reid et al. (2003), Pravdo et al. (2004), Henry et al. (2006b), Pravdo et al. (2006), Smart et al. (2007), van Leeuwen (2007c), Makarov et al. (2007), Gatewood (2008), Gatewood & Coban (2009), Lépine et al. (2009), Smart et al. (2010), Riedel et al. (2010), Khrutskaya et al. (2010), Shkolnik et al. (2012), and Anglada-Escudé et al. (2013). The scatter of the residuals when compared to the reported literature value is approximately Gaussian. Fitting this distribution as a Gaussian, we find a best fit width ≈15% larger than the sum of our errors and the errors reported in the literature added in quadrature (see Figure 3). We report our measured parallaxes and uncertainties for each target in our catalog.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

To determine whether the plates are modeled adequately by terms that are strictly linear in the coordinates, we refit the subsample of target stars with previously determined trigonometric parallaxes with a second order plate model. We found that the additional parameters ended up improving our derived value of the parallax relative to the previously determined value for only approximately half of the stars, and the remainder of the stars showed a marginal decrease in agreement with previous values. Furthermore, on average, the fit became worse, as the standard deviation of the residuals to the literature values increased by 2 mas. Therefore, we do not believe the lack of higher order terms in our plate model to be a significant source of error in our results, and that including these higher order terms causes us to fit astrometric noise rather than real trends associated with our detectors.

Another possible source of error is DCR in the atmosphere through which our measurements are taken. Since the reference stars (from which the plate constants are derived) are, on average, bluer than the target star, the relative effects of DCR between the reference stars and the target star could become important. Stone (2002) measured the effect of DCR in different optical bands as a function of color and found that the maximum DCR effect in the I-band for stars with a B − V of 2.0 is 12 mas, suggesting that the effect in our bandpass is probably small, as the B − V color of a typical late M dwarf is approximately 1.8. Nonetheless, we proceed to estimate the effect that DCR should have on our data.

The effective wavelength of a typical M dwarf target, accounting for telluric absorption, the filter bandpass, and the detector quantum efficiency is 850 nm, and is 840 nm for a typical reference star. We assume a typical target is a M5V star (Pickles 1998), and that a typical reference star is a G2V star. The index of refraction for air at 15°C, and standard atmospheric pressure at these wavelengths is n_840 nm = 1.00027482 and n_850 nm = 1.00027477 (Ciddor 1996). However, typical conditions at the MEarth observatory on Mt. Hopkins are significantly different than standard atmospheric conditions. The seasonal average temperature is 10°C and the typical pressure is 575 mm Hg. Additionally, the amount of water vapor in the air also affects the index of refraction of the atmosphere. For a typical relative humidity of 30%, at these conditions the partial pressure of water is 2.8 mm Hg (Ahrens 1994). Correcting the index of refraction for these effects using the methods in Filippenko (1982) and Barrell (1951), we find that the expected DCR between our reference stars and the target M dwarfs is typically 6 mas at airmass = 1.41, and 10 mas at airmass = 2.

Our model is capable of accounting for some of the DCR signal with the meridian constant parameters G and H. These constants are fit on a star by star basis, and therefore the "mean" DCR correction term for each star in each data set becomes merged into the meridian correction for each star. Any residual differential color effect will only be a second order effect and much smaller than 10 mas, below our threshold for detection.

To investigate whether any significant effects due to airmass or color remain in our data, we show the residuals in each coordinate direction (ecliptic longitude and ecliptic latitude) as a function of the hour angle at which the image was taken, in Figure 1. We find no significant directional offset between the residuals and the airmass where the image was taken, and conclude that the effect of DCR in the MEarth bandpass on our astrometry is negligible.

For nearby stars with large radial velocity, the effect of secular acceleration (a changing of a star's angular proper motion as a result of its changing solar distance) may become important, as our model assumes a constant angular velocity for each star. Barnard's star, one of the fastest moving stars in the sky, has a secular acceleration of approximately 1.2 mas yr⁻² (van de Kamp 1935). Since our astrometric model is fitting for the average proper motion over a maximum of a 4 yr time period, this can only result in a maximum systematic of ≈2.5 mas, below our detection threshold. Therefore, we ignore any effects of secular acceleration for all of our targets.

We further investigated whether the residuals in our derived parallaxes compared to previous results correlated with other external parameters. We find no correlation with the brightness of our target star, the average brightness of our reference stars, the color of the target star, the average color of the reference stars, or with intrapixel variation (evaluated by looking at our residuals as a function of sub-pixel position).

All of the stars in our sample have estimated distances from the Lépine (2005) piecewise linear relationship in V_est − J color. The V_est magnitudes for our targets all come from this catalog. This relation was calibrated from the 3104 M dwarfs from the LSPM North catalog (Lépine & Shara 2005) that had trigonometric parallax measurements, and has a mean error of 35% on the distance estimate. We note that most of the estimated V magnitudes in Lépine (2005) come from photographic plate measurements, and some have uncertainties as large as 0.5 mag. We note that high quality V photometry is available for some of the stars where photographic estimates were used by Lépine (2005); however, for the purposes of this work, we use the estimated V magnitudes compiled by Lépine (2005). In Figure 4, we show the distance modulus as expected from the photometric distance, using the calibration from Lépine (2005), compared to the value derived from the MEarth astrometry, as well as stars that have previous trigonometric parallax determinations available. In this plot, an equal mass binary would have an offset of 0.75 in distance modulus from the photometric distance modulus estimate. However, we find that that photometric distance estimates have a significantly higher scatter than trigonometric measurements, and that the typical scatter in the photometric measurement is large enough that identifying equal mass binaries through comparison of photometric estimates with our trigonometric parallaxes is not trivial.

Zoom In Zoom Out Reset image size

We note that previous estimates of the binary fraction among M0–M5V stars place the binarity fraction at 42% ± 9% (Fischer & Marcy 1992), 27% ± 16% (Gizis & Reid 1995), or 27% ± 3% (Janson et al. 2012), with a trend toward smaller binarity fraction for lower mass primaries (which the MEarth survey preferentially targets). Fischer & Marcy (1992) find that the binary distribution for M dwarfs peaks with companion object having an orbital period between 9 and 220 yr, and they used a range of detection techniques, including spectroscopy, speckle interferometry, and direct imaging. Gizis & Reid (1995) and Janson et al. (2012) relied on imaging and lucky-imaging techniques, respectively, and therefore are not as sensitive to tighter multiple systems. Regardless, there is likely to be a significant amount of contamination from target stars that are actually unresolved binaries. As the MEarth survey target list was designed with a distance cut-off, and it is harder to resolve close binaries at larger distances, it is likely that the contamination is due primarily to unresolved binary systems further than 33 pc masquerading as single stars estimated to be within 33 pc. Using only a photometric distance measurement means that the volume in which an equal mass binary can masquerade as a single star is 2^3/2 larger than the volume in which we aim to obtain our sample. Removing these stars from our sample before investing a significant amount of observing time to investigate whether these stars have transiting planets will make the MEarth survey more effective at finding planets. With significantly improved stellar distances, the limiting factor in distinguishing unresolved multiples from single stars is the quality of the available photometry that can be either compiled or gathered for these objects. We are currently working on calibrating the MEarth data to obtain accurate absolute optical magnitudes, and hope in the near future to use this data to determine which of our targets are likely binaries. However, for the rest of this paper we assume that all stars in our sample are single stars.

The MEarth survey is designed to look at nearby mid-to-late M dwarfs within 33 pc to the Sun, primarily drawn from the catalog compiled by Lépine (2005). However, we note that in designing this sample, we have introduced spectral type dependent metallicity biases, as the V_est magnitude has a dependence on metallicity as well as spectral type. The Lépine (2005) catalog was designed to be approximately 50% complete out to 33 pc with the goal of being mostly complete out to a distance of 25 pc. With our estimates of the trigonometric parallaxes for each star, we are now in a position to determine how complete the census of the solar neighborhood is out to these distances for the types of star in the MEarth sample. Assuming that on these small scales the effects of galactic structure are negligible, we can assume that the cumulative number of MEarth targets should increase linearly with the volume, with radius R as R³. In Figure 5, we show the cumulative number of MEarth targets out to 50 pc, with a best fit R³ line, fit to the cumulative number of stars from 5 pc to 15 pc. We omit the cumulative number of stars on scales smaller than 5 pc because there are not a sufficient number of stars to make a reliable estimate.

Zoom In Zoom Out Reset image size

In Figure 6, we plot the ratio of the number of stars to the number we expect to find for stars with V_est − K < 5.5 (M4.5 and earlier), 5.5 < V_est − K < 6.5 (between M4.5 and M5.5V), and for stars with a V_est − K > 6.5 (spectral types later than M5.5V). We plot both the sample of stars for which we measure parallaxes (presented here), and all stars that originally made the initial MEarth target selection from Nutzman & Charbonneau (2008), for which the majority MEarth has taken data. For stars that do not have a measured parallax from the MEarth data, we use an estimated distance from a spectroscopic measurement (if available) or its V − K color. Each color bin is fit to an R³ power law using the cumulative number counts between 5 and 15 pc, and the ratio of the number of stars we find within that distance to that expected from our fit is what we define as the completeness ratio. The sample of new MEarth parallaxes used to construct this plot has one known incompleteness: it is missing some stars earlier than M4.5 in spectral type at distances less than 15 pc because these stars were too bright for MEarth to observe efficiently. However, because we base our completeness estimates on the number of stars between 5 and 15 pc, we therefore underpredict the local density of M4.5 dwarfs and earlier and appear overcomplete at distances above 15 pc for these stars in the top panel of Figure 6. As expected, this artifact largely disappears in the bottom panel of Figure 6, when we include all stars regardless of whether MEarth measured a new parallax for them. For the full sample, we find that for stars with a V_est − K color less than 6.5, we are nearly complete out to a distance of 25 pc, but for redder stars, our completeness begins dropping off at the smaller distance of 20 pc. New searches for these smaller, redder objects, utilizing WISE data for example, may identify these missing objects in the near future. We do not find a clear correlation between completeness and Galactic latitude, which indicates crowding and confusion in the Galactic plane may not be the limiting factor in identifying the missing systems (see Figure 7).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

We can now estimate the masses and radii of our stars more reliably, aiding our understanding of the physical characteristics of individual M dwarfs in the solar neighborhood. Specifically, we can more reliably estimate the intrinsic brightness of our targets, and through previously published relationships, estimate the mass and radius of each star.

Delfosse et al. (2000) obtained an empirical relation between the masses of low mass stars and their luminosities in the near infrared, with the smallest scatter obtained in the K band. As each of the MEarth targets have precise K-band magnitudes as measured by 2MASS (Skrutskie et al. 2000), we can use these measurements with our trigonometric parallax measurements to obtain more reliable masses for each star. Similarly, by using the mass–radius relation for low mass stars obtained by Bayless & Orosz (2006), we can obtain more reliable radii for each star. Both of these are important for characterizing MEarth's sensitivity to planetary transits in these systems.

In Figure 8, we show histograms of our newly derived stellar masses and radii compared to the values derived by Nutzman & Charbonneau (2008), which used the photometric distance measurement and the same mass–radius relationship. For stars where the newly estimated mass is M < 0.5 M_☉, we find a median of the absolute value of the offset of ΔM ≈ 0.08 M_☉, whereas if we take the global sample, we find a median of ΔM ≈ 0.12 M_☉, as the initial MEarth sample was constructed to be limited to only the mid-to-late M dwarfs. We note that the stars that we assign higher mass values to are located further away than photometrically indicated and may instead be unresolved binary or multiple systems. Additionally, a systematic trend of the V_est − J colors being too red would also systematically shift our stars to higher estimated mass once we have obtained trigonometric parallax distances to them. Similarly, the typical change in the stellar radius is also ΔR ≈ 0.08 R_☉ for stars whose mass from MEarth astrometry was M < 0.5 M_☉ and ΔR ≈ 0.12 R_☉ for stars whose final mass was M < 0.8 M_☉.

Zoom In Zoom Out Reset image size

Importantly, the stellar radii of our targets determine our sensitivity toward transiting exoplanets in the system, as the transit depth is approximately proportional to the ratio of the areas of the stellar disk and planetary disk. Since MEarth's observing strategy is to obtain a per-pointing signal to noise ratio sufficient to detect a 2 R_⊕ planet transiting in front of the star at a significance of 3σ, an accurate estimation of the radius is essential to our ability to detect a transiting planet of a given size. Since large planets tend to be rare around small stars (Gaidos et al. 2012; Morton & Swift 2013; Dressing & Charbonneau 2013), obtaining the appropriate exposure time is essential for MEarth to accomplish its science goals. Additionally, the mass of the host star directly determines the planet's orbital period at a given separation, and directly affects the temperature of the planet as well. Therefore, characterization of a planet directly depends on the host star's physical parameters. As of the 2012–2013 observing season, MEarth is using these new radius determinations to determine which targets are observed in a given season and to set the integration times for these targets.

Aside from the transit depth of a detected companion being dependent on the stellar radius, we also note that obtaining accurate stellar parameters will also affect our understanding of any discovered planet's habitability. The boundary of the habitable zone is currently a topic of debate (see, for example, Gaidos 2013; Danchi & Lopez 2013; Kopparapu et al. 2013), but even in the most simple definition (distance at which the equilibrium temperature of a blackbody supports the existence of liquid water), the period of a planet in the habitable zone of the more massive M dwarfs in our sample have longer orbital periods than what can be easily detected with the MEarth observatory. Therefore, by shifting our priority to smaller M dwarfs, we are sensitive to smaller planets at all orbital distances, and increase our rate of recovery for planets in the habitable zone of the target star.

While the census of the solar neighborhood out to 10 pc is largely complete down to the mid-M dwarfs, there are several stars for which photometric distances places them within or near the 10 pc boundary, but for which there is no published trigonometric parallax. In the sample presented here, we find 37 stars whose trigonometric parallaxes place them within the 10 pc radius boundary. Of those 37 stars, 29 of them have previously measured parallaxes that confirm this distance. The rest of the stars have photometric V − K distance estimates that place them within 10 pc of the Sun, and we therefore confirm their proximity to the Sun.